Hello and welcome to Coherent's Markets Day. I will now turn it over to your host, Paul Silverstein.
Good morning or good afternoon, or good evening to those of you who are joining us from Europe or Asia. My name is Paul Silverstein, Vice President of Investor Relations and Corporate Communications at Coherent. It is my pleasure to kick off today our communications market deep dive presentation. This will be the first in a series of planned market deep dives with a goal of providing greater insight into our various market opportunities, including size, growth rates, underlying key trends and drivers, and our competitive position and competitive differentiation. We decided to start this series with our communications market, given that the rapid emergence of AI represents our most prominent trend, and communications remains our largest market, along with the attending significant investor interest therein.
In our June fourth quarter of fiscal 2023, communications accounted for 47% or almost $540 million of our $1.24 billion in total revenue. Our business spans both the datacom and telecom markets, which accounted for approximately 60% and 40%, respectively, of our fourth quarter communications revenue. Joining us today in the order in which they are presenting are the following Coherent executives: Dr. Sanjai Parthasarathi, our Chief Marketing Officer; Dr. Lee Xu, Executive Vice President, Datacom Transceivers; Dr. Beck Mason, Executive Vice President, Telecommunications; and Dr. Julie Sheridan Eng, our Chief Technology Officer. Following the conclusion of the four presentations, which are prerecorded, sell-side analysts will have the opportunity to ask questions of Sanjai, Lee, Beck, and Julie in a live Q&A session.
That said, for investors on the call, should you have any questions, please feel free to email me your questions at Paul.Silverstein@coherent.com. I will be happy to ask them. The presentations and Q&A session will be available for replay on Coherent's website for 12 months, and please feel free to reach out to me following the event. I would welcome any feedback that you may have. I'd like to remind everyone on this call that we will refer to forward-looking statements, including all statements the company will make about its future financial and operating performance, growth, strategy, and market outlook, and that actual results may differ materially from those contemplated by these forward-looking statements.
Risk factors that could cause actual results and trends to differ materially are set forth in the investor presentation, the company's annual report on Form 10-K for the fiscal year ended June 30, 2023, and in other SEC filings of the company. Coherent disclaims any obligation to update any forward-looking statements, which speak only as of their respective dates. It is now my pleasure to turn the microphone over to Sanjai.
Hello, everyone. I'm Sanjai Parthasarathi, Chief Marketing Officer of Coherent. Today, I would like to welcome you all to our first Markets Day. We are eager to share our excitement regarding our communications market opportunity and our positioning and competitive differentiation in this market through our current and future innovative technologies and products. Let me begin with our total addressable market opportunity across all of our markets. Our four diversified markets are industrial, communications, electronics, and instrumentation. Our current TAM across all of these markets is over $64 billion. Over the next five years, we expect this TAM to double. We further segment our markets by defining verticals within each market. These verticals include precision manufacturing, semiconductor capital equipment, display capital equipment, and aerospace and defense within the industrial market.
Our precision manufacturing vertical consists of a broad portfolio of optics, lasers, and turnkey manufacturing systems, all related to laser materials processing for applications that range from fading jeans to precision welding of electric vehicle batteries. Our semiconductor capital equipment vertical serves the wafer fab equipment market with a broad range of advanced materials, optics, and lasers, from wafer chucks to diamond windows for extreme UV lithography. Our display capital equipment market is centered around enabling display manufacturing processes such as laser annealing for OLED displays and display cutting. Our aerospace and defense business consists of optics, materials, and lasers for applications in intelligence, surveillance and reconnaissance, contested space, and directed energy. Our communications market, which is the focus of this event, includes our datacom and telecom businesses. We will come back to these shortly. Consumer, automotive, and wireless electronics are our verticals, which comprise our electronics market.
Within our consumer electronics vertical, we enable advanced sensing applications for smartphone and wearable devices, including AR/VR headsets, through our broad portfolio of lasers, optics, and materials. Silicon carbide substrates and devices for electric vehicles and energy applications are the focus of our automotive vertical, along with lasers and optics for autonomous and driver assistance systems. And finally, we segment our instrumentation market into life sciences and scientific instrumentation verticals. Within life sciences, we serve the biotechnology equipment market, PCR test equipment, for example, as well as the medical laser market with our broad portfolio of optics, lasers, thermoelectric materials, and subsystems. Before we jump into our communications market, I'd like to review our revenues in fiscal 2023 by segment, by market, and region. Fiscal 2023 was a record year for our communications, electronics, and instrumentation markets.
While fiscal 2023 was a reset year for our industrial market owing to the macroeconomic environment, we did have a record year for our semiconductor capital equipment vertical, driven by growth in the front end of the line wafer fab equipment applications that enable the next generation nodes. Within communications, both our Datacom and Telecom businesses generated record revenues. All three verticals within our electronics market also generated record revenue. Within our automotive vertical, we are especially excited about the rapid growth in silicon carbide electronics that is enabling the electrification of transportation, first cars and light trucks, and then trains, ships, and planes in the next decade. Let's turn now to the communications market, which is the focus of today's session. Owing in part to the fact that about 3.6 billion people are still not connected to the internet, the demand for bandwidth remains insatiable.
We believe that 90% of the traffic created on the internet was created in the last 2 years, and we expect the half-life to continue to decay exponentially. The metaverse, powered by immersion technology and IoT devices, generating data and communicating with each other, are two examples of applications that drive the demand for bandwidth and visual communications that drive the demand for displays. It is clear that optical networking infrastructure already is critical to enabling current and future applications that drive increasing demands for higher bandwidth and spectral efficiency, and for lower latency and power consumption, too. Bandwidth and latency demand of new applications continue to drive the trend toward high-performance optical infrastructure, which in turn is driving ongoing compression in optical technology upgrade cycles. The rapid emergence of artificial intelligence and machine learning already is accelerating this trend, despite just beginning to emerge.
Its impact on optical infrastructure upgrades likely will be increasingly prominent over time. In short, high-speed optical connectivity already is of critical importance to these networks, and it will be of still increasing importance in the future. Especially exciting to us, as the world's largest supplier of optical components, modules, and subsystems, is the increasing network content of optics as the need for higher bandwidth and lower latency increases. This is true for both wireline and wireless networks within and outside of the data center, within the traditional access, and not just the edge and the core of service provider networks, and within cable TV and broadband access systems, switches and routers, in addition to traditional DWDM optical systems.
The datacom and telecom markets for us represent a combined served available market of $17 billion in calendar 2023, which we estimate will demonstrate a 13% compound annual growth rate over the next 5 years, increasing to $31 billion in calendar 2028. As a key driver of this growth, we continue to see internet content providers, or ICPs, ambitiously building out their optical network infrastructure, with their capital expenditures outpacing that of traditional communication service providers. They are investing not only within the data centers and in the data center interconnects, but also in metro, long haul, submarine, and even satellite networks, all of which are enabled by coherent innovations. Before we dive into the details, let's review how we segment the datacom and telecom verticals.
The chart on the right shows the various parts of a communications network: submarine, long haul, metro, and access networks and data centers. Data center interconnects, which are networks that connect data centers together, are classified under Telecom. All the connectivity within the data center or data center interconnects are classified under Datacom. We have been the market leader in Datacom transceivers for three decades. We are vertically integrated, and we have all the leading-edge technology platforms, including gallium arsenide and indium phosphide-based lasers, as well as integrated platforms such as silicon photonics. We have climbed the experience curve, and Lee will talk more about our leadership position in the industry. Let's start with Datacom and with the most important application, AI connectivity. AI connectivity has recently emerged as an explosive driver of Datacom transceivers and a key enabler of the new long-term trend in the demand for high-speed transceivers.
Mainstream networking topology is giving way to a new topology for AI systems. Traditional CPU-based compute platforms are increasingly being replaced and enhanced by accelerated compute devices. This fabric of GPUs connected to each other with high-speed interconnects form an accelerated compute cluster for machine learning. All of these high-speed interconnects are enabled by datacom transceivers. While we've been selling into AI applications for a while following the advent of generative AI in early 2023, we embarked on a detailed sizing of the AI market opportunity for datacom transceivers. This chart shows our latest view of the size of the overall datacom transceiver market, including, in particular, the part of the market driven by AI applications.
We expect the AI segment will grow at an almost 50% CAGR over the next 5 years, driven by 800 G, 1.6 T, and eventually 3.2 T transceivers. Historically, high-speed upgrade cycles have proven to be quite long. They have also proven to involve overlapping line rates for over a decade. For example, we are excited by the ramp of 800 G, which will be the bulk of the AI transceiver market for the next 5 years. We expect to lead in 800 G, and we expect to see the early launches of 1.6 T deployments begin as early as late 2024. The laser is at the core of any transceiver, and it is a key enabling component for higher speed optical communication that imparts the functionality of the network that it enables.
In fact, it enables all high-speed transmission. We have a variety of laser platforms, including our vertical cavity surface emitting laser or VCSELs for short reach applications, our electro-absorption modulated laser, EMLs, and our latest DFB Mach-Zehnder for longer reach. We have segmented the Datacom transceiver market by the laser type and by AI and traditional applications. VCSELs are important for AI, as AI fabrics typically consist of relatively short links. The first wave of deployments for AI will be enabled by 800G transceivers, driven by our 100G VCSELs and EMLs. VCSEL-based transceivers are most applicable in GPU clusters, also known as level zero connectivity, which typically involve physical server-to-data center switch or data center switch to data center switch interconnects that are less than 100 meters in length.
EML-based transceivers, due to their longer reach capability, feature in level one interconnections, which are typically up to 2 km in length. There are also CW laser-based silicon photonics transceivers, which are best suited for applications up to 500 m in reach. While silicon photonics-based transceivers also will likely compete with VCSELs in the short reach segment of the market, we believe VCSELs will prevail due to their low cost and low power consumption. The second wave for AI will be with 1.6T transceivers, enabled by 200G per lane lasers. At 200G per lane, optical link budgets become challenging due to increased dispersion and noise, driving the need for advanced lasers. Our award-winning 200G per lane EMLs support links up to 2 km.
Our new innovative DFB Mach-Zehnder laser, which we expect to ship in time for the market demand in calendar 2025, support links up to 10 km. These advanced lasers, along with our high-power CW lasers for silicon photonics, enables us to provide a complete portfolio of all the required transceivers for success in this market. We are also developing 200 gig VCSELs, which, once available, we expect will be the mainstay of the short reach market. Technology choice ultimately depends on performance requirements and costs, and these vary from application to application. As a vertically integrated manufacturer with the broad technology portfolio that includes a range of transceivers based on VCSELs, EMLs, and CW laser silicon photonics, we expect to play a leading role in all these emerging applications. Here, we showcase our laser opportunity for these transceivers, segmenting again by AI and non-AI and by laser type.
As we saw on the previous slide, AI applications will require various laser types. We expect the AI portion of the laser market will increase from 22% today to over 60% of the market by 2028. We are now getting ready for the first wave of demand with our high-speed optics, with our lasers, both for our own internal consumption in our datacom transceivers and for external sale to third parties in the merchant market. We also make all the other components that go into these next-generation transceivers, namely the optics, receivers, and ICs. As a completely vertically integrated supplier with the largest manufacturing scale in the industry, we believe that we are in a great position to continue to reinvest in our product portfolio and to continue to lead the market. Now, let me move on to the telecom market.
We have the broadest telecom optics portfolio in the industry. Our solutions are in play from the moment you pick up your phone and you open an app. An example would be the 5G RF filters in the phone that are enabled by our unique pulsed excimer laser deposition platform. The second example would be our silicon carbide substrates that underpin gallium nitride on silicon carbide devices that power the RF amplifiers in the base station. The third would be front-haul transceivers that are at the base of the cell phone tower that convert electrical signals into the optical domain. Once you are in the optical domain, our broad portfolio can fulfill all of the telecom network requirements from access, metro, long-haul, submarine, and space-based links.
In addition to our end-to-end portfolio, we are completely vertically integrated down to the materials level, which gives us a cost advantage and also enables us to push the performance limits of our higher performance products. The optical network has been on a disaggregated path for a few years now. Disaggregation promises to help operators reduce cost and broaden the supply base through standardization and multi-vendor interoperability. Service providers and data center operators would like to be able to build telecom optical networks using products from a mix of vendors, the same way the web scalers build data center networks. Disaggregation is a relatively new trend in telecom networks, and this trend is gathering momentum. Pluggable coherent transceivers for these disaggregated networks are slated to grow at a 20% CAGR over the next five years.
This is a market where we are making great strides, leveraging our vertical integration with our industry-leading indium phosphide platform to offer uniquely differentiated products. Customers have been leveraging our innovative products to create new markets and applications. Beck will talk more about those innovations and the outstanding customer reception for the same. Finally, I would like to thank and recognize our customers and partners for their continued support of our innovation and our market-leading products. With that, I turn it over to Lee.
Thanks, Sanjai. Hello, everyone. I'm Lee Xu, EVP and General Manager of Datacom Transceiver Business Unit at Coherent. I'm proud to share with you our Datacom Transceiver business history, strength, and future outlook as we address the explosive growth, the surge, and sustained demand that Sanjai spoke about earlier. I'm excited to showcase our AI-related products, our vertical integration, which is one of the hallmarks of our technology differentiation, and how we are preparing to fulfill our customers' explosive demand with operational excellence, comprised of both our diversified and scalable manufacturing strategy and our focus on exceptional quality. Our Datacom Transceiver business started as Finisar, which was founded in 1988 and acquired by what, what was then II-VI in September 2019. We pioneered pluggable transceivers and enhanced our position through various acquisitions and internal development to offer the broadest portfolio of pluggable transceivers in the industry.
The depths and breadth of R&D, along with our vertical integrated production model, are key competitive differentiators, which provides advantages in terms of cost, quality, visibility, and time to market. Our R&D and manufacturing prowess together translate into stronger and deeper strategic customer engagements, as opposed to others that can offer only transactional customer relationships. This slide shows our Datacom revenue trend for the past 18 years. Historically, we have consistently enjoyed a leading market share, as reported by a number of external industry market analysts. I would like to take the opportunity to thank our customers, suppliers, and sales and distribution channels for their historical and ongoing trust, partnership, and collaborations. This shows that our market leadership has been consistent and durable across a protracted period of time, and the industry upgrade cycles.
You can see how we have led every major historical upgrade cycle, spanning 10G, 100G, 200G and 400G ramps. In fiscal year 2023, our 200G and 400G revenue was over 60% of our total Datacom sales, among the highest in the industry. Since the acquisition of Finisar four years ago, we have demonstrated steady revenue and profit growth. We're now well-poised to lead the next Datacom transition, namely 800G and eventually, but not far off, 1.6T upgrade cycle, driven initially by the rapid emergence of AI. Thanks to close collaboration among our various development teams, spanning lasers, ICs, detectors, passives, integration, new product introduction, and manufacturing operations, supply chain, all of which are under one roof.
We have had design wins for all meaningful 800G volume slots, and we have already started volume shipment of some of the major 800G products. The opportunity on hand this year, and over the next few years, is the disruptive growth of AI, as Sanjai just highlighted. In the next two years, most of that revenue growth will be from various 800G modules, primarily driven by AI-related deployments while still very early, we're optimistic that AI will prove to be an exceptional driver of our datacom revenue growth for many years to come. We began shipping 800G transceivers in low volume in the third quarter of our fiscal 2023.
But based on our design in work in fiscal year 2022-2023, and our orders in hand, we have already established a strong market position and are confident that we will grow faster than the market. Among the factors underlining our confidence, are existing customer relationships and the substantial orders on hand for our 800G transceivers, the depth and breadth of our 800G transceiver portfolio, and our product development, including our 1.6T development, that we expect to commercially ship before the end of fiscal year 2025. Our industry-leading 200G per lane lasers constitute the key enabling components of those 1.6T and eventually 3.2T transceivers. During the rest of this presentation, I want to show you that we're ready to take the leadership position enabling this revolution.
Before I go into more details of our business, please allow me to take a few minutes to give you our overview of the ecosystems and our business strength. We provide optical transceivers to all major companies in datacom communications. As measured by revenue, hyperscalers comprise the first and largest segment of our datacom customers. Our second-largest datacom customer group are traditional communications or networking equipment system manufacturers, also referred to as OEMs or ODMs. While over time, the transceiver market has become quite competitive, mostly with pure-play integrators who source and procure all the key building blocks from third parties, such as Coherent. Customers continue to value our differentiation. A prominent source of differentiation is provided by our technology leadership, both in terms of our high-speed laser platforms, which I will come back to, and Julie will discuss in more details, and our vertical integration.
Regarding the latter of these, the vertical integration, we're the only supplier who is completely vertically integrated down to the material level. This includes various types of lasers, detectors, passives, optical component, and integrated circuits. The advantage of vertical integration is that our lasers, ICs, and other component designers closely collaborate with our integration engineers, which accelerates our time to market, time to quality, and time to volume. This is becoming even more important at the higher data rates. This is a trend that will persist. Our operational excellence, scalability, and diversity in manufacturing provides a third source of competitive differentiation. With my 25 years of transceiver business experience, I can say humbly that the significant technical challenge of building one datacom transceiver in a lab is dwarfed by the challenge of building millions of them with flawless excellence in quality and with industry-leading cost.
We do appreciate that our ecosystem is complex and interwoven. For example, some of our large customers have their own transceiver designs and contract manufacturers. Some of our suppliers also sell competitive transceiver products. We're able to maximize our profitability and drive our roadmap by serving the entire ecosystem at multiple levels. For example, we offer components such as VCSELs, EMLs, passive optical components, and the materials to our transceiver competitors too. We believe we offer the best products money can buy at all levels of the value chain, and we're grateful for our customers' appreciation. In such a competitive field, we have maintained an overall number one leadership position in the transceiver market for over the past 15 years, and we believe that our financial performance, in terms of a gross margin percentage and free cash flow relative to revenue, have been consistently better than our nearest competitors.
This achievement comes out of and reflects our strength built up over the years. The technology endowment that we have is the result of three decades of leadership of Finisar, in combination with the complementary assets and similar mindsets of legacy II-VI, including the current leadership of our company. One of our strength is our broad portfolio of offerings. All of the customers, in particular our OEMs, appreciate the one-stop shop, which provide volume and associated purchasing power. Access to this large and deep customer base has allowed us to enhance our scale and, in many cases, to lower our cost. Let me move on to our manufacturing operations. One of the key strength is our assembly and testing scale in our two MAG factories in Ipoh, Malaysia, and Wuxi, China. We have excellent shared quality and IT systems in both factories.
They both have long track records of making many advanced, high-quality transceiver modules for demanding customers. Both factories have ramped up quickly to produce significant transceiver unit volumes, and they have the capability to support future volume expansion. We have recently opened a datacom R&D design center in Penang, Malaysia, to be not only first to market, but first to volume for launch of new products. One key advantage of our factories is that we have sophisticated automation, test development, and process development capabilities. These capabilities, plus our new product introduction systems and engineering teams, enable us to very quickly scale up most advanced products. That capability was honed and proven in many previous ramp-ups. It is an important factor in our success in maintaining our market-leading position in the transceiver industry.
Coherent is unique among large transceiver suppliers in that we have world-class 3-5 device development teams and fabs in Zurich, Stockholm, Sherman, Texas, and Fremont, California. These fabs provide devices both for our own internal transceiver production and to external transceiver suppliers. We also buy from external sources for the purpose of supply chain continuity and resiliency. For 800G generation, our internal EML and VCSEL chip lasers are ready for volume shipment. The transceiver-level R&D teams in California and Shanghai are a huge strength for us. The teams developed every major flavor of 800G transceivers and secured design wins with key customers. We have already started ramping production and shipment. We expect our production and shipments to accelerate as we progress throughout fiscal year 2024.
In the first half of fiscal year 2024, mainly due to external component shortages, we expect to ship no more than 20% of our expected 800G revenue for the full fiscal year. We expect our 800G volume, output, and revenue to be significantly higher in the latter half of fiscal 2024, with sequentially higher volume and revenue in each quarter. This slide shows our product roadmap beyond the current 800G modules. We have been closely engaged with major customers, working collaboratively on specification, technology choices, and development schedules. The key here is the 200G per lane technologies, not only for lasers and optics, but also for the required electronics. We're in good shape for 800G and the future 1.2T, 1.6T and 3.2T generations from our R&D perspective.
We expect to establish clear market leadership in 1.6T. We started that development over a year ago and have made significant progress. This development is going to be more challenging, but it also plays to a lot of our company's overall strength. First, the time to market for this generation is even more crucial for our customers as they are vying to release the next generation of AI systems as early as possible. We believe our close R&D engagement with them provides us with competitive advantages. Secondly, different customers have different architectural designs, so different technologies are needed to address different needs, including a variety of lasers spanning silicon photonics, EMLs, DFB Mach-Zehnder lasers, and VCSELs. We believe that we have a clear lead in many of these technologies.
We're rallying all of the related resources to gain a strategic advantage, and we will differentiate our transceiver designs and laser technologies in this new generation of transceivers. We talked about the huge demand surge for 800G and of our work toward meeting that demand. We also talked about the new product development beyond 800G. We plan to maintain our share of the non-AI datacom transceiver market and expect to gain a significant share of the AI market. Even though it's going to be very competitive, we have had a good start in our qualification, our order booking, and the initial shipment of 800G transceivers. We're confident that we can gain significant share in calendar 2024, and we'll build on that in calendar 2025.
We're investing strategically to ramp our capacity and drive down our cost, and we're investing heavily in the next generation of 1.6T products. We're optimistic as to keeping and growing our share, both in transceiver and in merchant market of the optical components over the next 5 years, driven by the AI tailwinds. In summary, the current AI surge and what we expect will be strong, sustained growth for many years to come. It provides us with an enormous opportunity. We're ready with our design wins and our capacity preparation, and with great customers who are the leaders in the AI market. We're well-poised to lead the market in the 800G, 1.6T, and 3.2T upgrade cycles. Thank you for your attention. Beck Mason will now address our opportunity and positioning in the telecom market.
Coherent has been in the telecommunications business for more than 30 years, and over that time, has grown that business both through internal innovation and through the acquisition and integration of companies with critical enabling technologies. Our technology roots date back to 1990 and include foundational IP developed at AT&T Bell Labs, Corning, Nortel Optoelectronics, Alcatel Optronics, Pirelli, IBM, and others in Europe and Japan. Our acquisitions include Marlow, which brought industry-leading solid-state thermoelectric cooler technology, Photop, with its broad array of vertically integrated passive optical solutions, Faraday Rotators from Integrated Photonics, the telecom pump laser, amplifier, and signal monitoring business from Oclaro, wavelength selective switch platforms from both CoAdna and Finisar, and indium phosphide lasers, and finally, optical transceivers and related transmission components technology from Finisar.
Telecom fiber optic networks cover a wide range of applications, from broadband access networks at the edge that connect homes and businesses to the internet, through fiber-based aggregation networks deployed by cable multi-system operators and traditional communication service providers. They support mobile networks through wireless front-haul and backhaul links, metro and long-haul core networks that interconnect data centers at central offices and colocation hubs, and then finally, subsea networks that connect across the oceans and around our coastlines. Coherent produces the products that transmit and receive the information over these networks, enabling hundreds of separate optical transmission channels using separate individual wavelengths to be carried over a single fiber, with aggregate capacities into the several tens of terabits per second. We also produce all the transport components that are used to add, drop, route, and switch these individual channels over the network.
Our pump lasers and fiber amplifier products regenerate the signals and extend their reach over thousands of kilometers. In long-haul applications, our monitoring solutions enable the performance of these channels and the health of the fiber infrastructure to be managed effectively. Modern telecom networks are configured as a mesh or a series of interconnected rings. You can think of each one of these connection points as a node, where signals can be amplified, redirected, and added onto the network or dropped from the network. We categorize the products that support this in two ways. First is transmission products that convert the electrical data from internet switches and routers into the complex optical waveforms used to carry the data over the fiber. Second is transport products that manage the switching, routing, and amplification of these signals within the network.
Coherent has one of the industry's broadest portfolios of technologies and products that power optical transport networks from the edge, where subscribers are connected through to the metro and core networks that interconnect them to the data centers that make up the cloud. Our products benefit our customers by simplifying their user experience and enabling them to reduce both capital and operational expense, while paving the way for the evolution of future networks and increased traffic that we expect them to carry. In all applications, customers value high performance, high reliability, smaller size, lower power consumption, and lower cost. Because of the realities of network operations and economics, we see a different emphasis on each of these attributes, depending on which section of the network is being considered. So, for example, in access networks, cost, power dissipation, and physical size play a significant role.
In core networks, spanning hundreds to thousands of kilometers, performance is key. In submarine and space networks, high reliability becomes a factor of primary importance. Global internet traffic continues to grow at a healthy pace, driven by video, wireless, web services, machine-to-machine communication, and many other applications. Over the past five years, we have shown steady growth in our telecommunications business revenue as we keep pace with the relentless demand for increased capacity in the global network. By the end of this year, there will be 29.3 billion network devices supporting 5.3 billion internet users worldwide. Overall, internet traffic is growing at 24% per year, and capital spending in the telecom markets that we serve is projected to grow at 14% per year for the next five years.
Coherent's telecom revenue has grown significantly over the last 10 years and was just short of $1 billion this past fiscal year. We have a broad portfolio of optical transceivers for telecom applications, and these span from compact, low-power solutions with direct detect receivers for edge access aggregation applications, delivering optical links to homes and businesses, all the way up to very high-performance coherent transceivers that can operate over thousands of kilometers and carry data rates as high as 800 gigabits per second on a single wavelength. Our strength is in the depth of expertise we have in the critical enabling technologies for these transceivers. We make the fundamental indium phosphide photonic integrated circuits that generate precise laser wavelengths required for dense wavelength division multiplex systems, and that encode and decode complex modulation waveforms onto these wavelengths, enabling them to carry very high-capacity data across long-haul networks.
In addition to this, we have a deep capability in RF analog IC design, enabling us to produce components for amplifying the electrical signals used in both the modulation and detection of the optical signals. We have also developed advanced digital signal processing solutions that perform the complex calculations to convert the analog waveforms used in the optical communications domain into digital data that can be sent to the Ethernet switches and routers used to direct traffic through the network. Along with these fundamental building blocks, we have advanced packaging capabilities for integrating the photonics and electronic circuits, for coupling the light from the photonic components into the optical fiber, and for controlling the environment of the critical components.
Our subsystems teams then integrate these photonic components into compact, pluggable transceivers that plug into the faceplate of the optical switches and routers that sit at the edge of data centers and telecom central offices. An important building block in a coherent transceiver is the digital signal processor. This device converts digital data from a switch or router into the complex analog modulation signals that are used to transmit data over a fiber optic network. It also converts the received signals at the other end of the link back into digital data and compensates for any signal impairments that occur over the fiber optic links. It is a very significant element in driving the overall performance of a coherent link, and it represents a significant portion of the cost of a coherent transceiver.
We partner with leading vendors in the industry to leverage their DSP solutions for some of our products, and we design and develop our own custom DSP solutions with our internal design team. This enables us to cover a wide range of solutions. We focus our designs on advanced process nodes, which enables the lowest power, smallest die size, and ultimately, the lowest cost solution. This combination of external partnership and internal capability lets us focus our engineering efforts where we can best differentiate with unique approaches that expand our market and enhance our ability to deliver disruptive solutions. Our transceiver products today cover multiple applications, from edge access to metro core, data center interconnect, and long haul. We support multiple form factors, including CFP2-DCO solutions at 100, 200, and 400 Gbps.
Our QSFP-DD solutions for 400 Gbps data rates are available in both ZR, which is a high-volume, short-reach, 80-kilometer solution for data center interconnect, and ZR+, which is a higher performance, longer-reach solution suitable for metro applications covering several hundreds of kilometers. One of our most exciting new products is our 100G QSFP28, which is an ultra-low power, 100 Gbps solution for both edge access and metro core applications. This part is the only coherent transceiver available today in the QSFP28 form factor. The solution is one quarter of the size of competing solutions in the CFP2-DCO form factor and less than half the power dissipation. The product is enabled by our internally designed low-power DSP, along with our compact, widely tunable semiconductor laser and high-performance silicon photonic-based coherent optical subassembly transmitter/receiver.
Our vertical integration enables us to achieve best-in-class cost, power, and performance for this solution. At the high end, our newest products are the 800-gigabit-per-second QSFP-DD and OSFP coherent transceivers for 800G ZR and ZR+ applications. These will be used in the next generation of data center interconnect applications and in telecom networks. Beyond this, we are actively working on the next generation of solutions that will deliver 1.6 terabit per second of data on a single wavelength. In addition to our transmission capability, we have one of the broadest portfolios of technologies for integrated subsystems in the industry. Our products cover all aspects of amplification, routing, switching, and monitoring in the network. Our amplifier technology is based on our industry-leading pump laser capability and delivers best-in-class solutions for receiver preamplification, transmitter, booster applications, and low-noise line amplification.
It includes both C and L-band solutions with erbium and Raman capability. For routing and combining DWDM channels, we have fixed mux and demux solutions, along with reconfigurable optical add-drop multiplexers based on our industry-leading LCoS WSS technology. For submarine applications, our ultra-low power, high reliability, liquid crystal-based WSS solutions are deployed in situations where reliability and low power dissipation are critical. To help support these capabilities, we have the industry's most advanced solutions for monitoring and control. Our optical channel monitors can track hundreds of individual wavelength channels within the network to ensure their performance. For more demanding applications, our coherent optical channel monitor technology is unique in the marketplace due to its ultra-high resolution capability, enabling it to deliver features beyond what traditional OCMs are capable of.
Finally, in order to monitor and ensure the integrity of the fiber infrastructure, we have developed Optical Time-Domain Reflectometry tools that can scan the fiber real-time and identify locations of breaks or excessive loss without disrupting live traffic. Increasingly, we combine many of these features together into integrated subsystems that support full node functionality for our customers. Typically, this integration is done in deep partnership with these customers to ensure compatibility with their network management software. One of the newest and most interesting products in the transport space is the Pluggable Optical Line System. With the introduction of IP over DWDM approaches using coherent transceivers plugged directly into switch and router ports, there has been an increased demand for a compact, efficient solution to combine and amplify these signals.
Our POLS product delivers a solution that enables the amplifier to be plugged directly into a transceiver port on a network switch or router, and when combined with a passive optical multiplexing cable, can serve the dual function of multiplexing the signal, the optical wavelengths from the various transmission channels, and then amplifying them onto the network. We continue to advance our technology in a number of critical areas. In the wavelength selective switch space, we are scaling to solutions that enable both C and L bands channels to be managed within a single device. This more than doubles the bandwidth capability of our competitors' solutions. On the amplifier side, we are leveraging our higher power pump capability to deliver more efficient amplifiers and to broaden their spectrum coverage so that more channels can be supported on a single fiber.
This increases the total information carrying capacity of the fiber and lowers the cost of the network. We have new optical channel monitoring technology that enables a single device to cover both C and L bands together, cutting the cost in half for monitoring C plus L networks. And finally, on the pump laser front, where we continue to lead the industry in the highest power per pump emitter, we are now transitioning to designs where multiple pump lasers are integrated into a single package, yielding much more compact and efficient amplifier designs. At Coherent, our strength is not just in our R&D and our technology, but also in our manufacturing capability. We have vertically integrated manufacturing that spans all the way from the base device and component level to the module and subsystem level. Coherent is a world leader in compound semiconductor wafer fabrication technology.
For our telecom products, we have ndium phosphide and gallium arsenide wafer fabs in Zurich, Switzerland and Stockholm, Sweden, respectively. They provide the photonic chips used in our telecom products and our enablers in our datacom products, as you heard in Lee's presentation. Our internal photonic packaging production facilities rely on highly automated robotic assembly to ensure consistent quality and performance of our products, and to enable high volume production with reduced labor and overhead costs. Our telecommunications pump lasers are built in Shenzhen, China, and the Philippines. Our photonic components are manufactured in our automated facilities in Wuxi, China. Our transceiver products are also built in Wuxi, China, and in our factory in Ipoh, Malaysia. We manufacture passive optics, along with amplifiers, line cards, and other subsystem products in Fuzhou, China, and we also have manufacturing locations in Vietnam, Thailand, and several other locations.
Our focus is on having geographic diversity of supply to ensure business continuity in the face of potential supply disruptions and to provide flexibility in country of origin to meet tariff and security-based sourcing concerns. We have an experienced manufacturing technology development team that is expert in assembly and test automation. Their capability enables us to automate most of the manufacturing stages of our production. The use of internally developed automation helps us minimize our capital expense and customize our solutions to best fit the requirements of our products. It also ensures more consistent product quality, less waste, and better manufacturing efficiency and cost. We pride ourselves on having one of the best quality reputations in the industry. We have a strong focus on continuous improvement in product quality, and we drive a zero-defect culture approach.
An example of this is the reliability of our liquid crystal WSS platform, which has a mean time to failure of more than 4,000 years, based on 8 billion cumulative hours of device operation. Coherent is one of the largest suppliers in the telecommunications component market, and we leverage that scale to keep ahead in product design, manufacturing technology, and product quality, so that we can bring superior value and quality to our customers. The breadth of our technology portfolio in both transport and transmission applications allows us to bring to market vertically integrated solutions that are higher in performance, more cost-effective, and better optimized to work together in telecom networks. We win based on the skill set of our people and the capability of our technology, our innovative culture, and our commitment to delight and enable our customers.
Whenever you're ready.
Hi, I'm Julie Sheridan Eng, Chief Technology Officer of Coherent. Today, I'll focus on our internal component technology that is critical to our success in the data communications and telecommunications markets, including the exciting new opportunity presented by the rapid rise of artificial intelligence and machine learning that Sanjai and Lee talked about. As Lee and Beck also mentioned, the speed of transceivers is critical to the performance of the network. Network upgrades and architectural changes to address artificial intelligence and machine learning are driving us and the industry to introduce higher speed transceivers at a faster pace than ever before. This requires tremendous innovation in high-speed components, especially semiconductor lasers. Over the years, we've made strategic investments that give us a unique level of vertical integration at scale.
We not only design and manufacture our transceivers internally, we also design and manufacture many of the components in our transceivers, including lasers, detectors, integrated circuits, and passive optical components. When designing a new transceiver that requires a new component, we may either internally design and manufacture that component or source it from one of our valued development partners. Key factors driving our decision as to what to develop internally and what to develop with suppliers include our drive to profitably grow and generate targeted returns on our investments, as well as time to market and other strategic considerations. Why have the capability to design and manufacture these components internally? Having the ability to internally design and manufacture critical components, including semiconductor lasers, matters for several reasons.
First, larger customers prefer to work with a small number of suppliers, therefore favoring suppliers that have a full complement of products and technologies in their portfolio. Second, by being in control of the semiconductor laser technology, we control the pace of development of our products, which we can leverage for time to market and can ultimately affect both revenue and market share. In addition, we are fast to innovate and better able to differentiate our products. Finally, specific to AI and ML, the market is growing very quickly, and our customers value our semiconductor laser manufacturing capacity that we can bring to support their aggressive ramps, which is part of our strategy to win. Of course, our semiconductor laser manufacturing capability also provides our customers with diversity and security of supply.
Both capacity and diversity and security of supply are very valued by our customers and should translate to revenue growth and market share. For datacom, including both dedicated AI/ML applications and traditional datacom applications, 800G and 1.6T transceivers require 100G per lane and 200G per lane lasers. The type of laser used is determined by the data rate and the length of the fiber link. Generally speaking, interconnects in the AI and ML portion of the network are less than 50 meters, while interconnects connecting top-of-rack switches to leaf spine switches are up to 500 meters, and interconnects connecting switches to routers or routers to routers are between 2 and 10 kilometers. Telecom network connections range from 10 kilometers to multiple hundreds or even thousands of kilometers. Each of these distances and applications are best served by different laser technologies.
For link distances less than 100 meters, including many links in the AI and ML portion of the network, vertical cavity surface emitting lasers, or VCSELs, are used. Our VCSELs are based on our gallium arsenide technology platform. VCSELs are generally the lowest cost, lowest power consumption solution, and accordingly are the lasers of choice for less than 100 meter connections. We're currently experiencing significant demand for VCSEL-based transceivers for AI and ML applications. Having multiple 6-inch gallium arsenide VCSEL fabs in the U.S. and in Europe, Coherent is one of the highest volume manufacturers of VCSELs in the world, and the only Datacom manufacturer in the world with an integrated 6-inch wafer platform. We recently had a press release announcing over 200 billion VCSELs shipped for consumer and Datacom applications.
Our 100G per lane VCSELs are experiencing a high volume production ramp to support 400G and 800G transceivers. Regarding the capability of the technology to support the next generation of transceivers, we are nearing the physical limitations of directly modulated lasers, and there has been significant industry-wide debate as to whether it will be physically possible to create a 200G per lane VCSEL. Most industry participants agree that if it is possible, a 200G per lane VCSEL is valuable, and that achieving 200G per lane VCSELs will require significant changes in VCSEL device design and fabrication at Coherent, we've been hard at work to address these 200G per lane VCSEL design and manufacturing challenges.
We have just demonstrated our first device results on a new VCSEL platform that we believe can lead us to 200 G per lane VCSELs. We believe 200 G per lane VCSELs represent a significant differentiator in the AI and ML market, which can help us grow our revenue and our market share. Turning now to distances greater than those that can be supported by VCSELs, single mode semiconductor laser devices are used. These lasers are made from indium phosphide materials. Coherent has multiple indium phosphide fabs in the US and in Europe. Our ondium phosphide technology platform is one of the very few in the industry that has been field-proven at scale, with more than 200 million Datacom lasers deployed in the field over the past two decades. Our lasers have been qualified and deployed by virtually every network OEM and hyperscaler in the world.
We leverage that deep experience to develop lasers to support single-mode 800G and 1.6T transceivers. For link distances greater than 100 meters but under 2 kilometers, silicon photonics-based transceivers may be used. While indium phosphide generally has better electro-optic performance, silicon photonics can offer advantages by allowing the integration of passive optical components. At Coherent, we have an in-house silicon photonics design team to develop the needed silicon photonics-based components for our transceivers. As is common in the silicon industry, these components are designed in-house and manufactured in Tier 1 foundries. All silicon photonic products require an indium phosphide Continuous Wave, or CW laser, to generate the light. We have developed and are manufacturing in-house indium phosphide CW lasers.
We are making these lasers commercially available in the market, and we will leverage them for our own transceiver designs, such as our silicon photonics-based 800G DR8 transceiver that we demonstrated at the European Conference on Optical Communications, or ECOC, in September 2022. For 500-meter to 10-kilometer links, indium phosphide-based electro absorption modulated lasers, or EMLs, may be used. We manufacture 100G per lane EMLs to support 400G and 800G transceivers, such as our EML-based 800G DR8 transceiver, which we also demonstrated at ECOC in 2022, where we showed it interoperating with our silicon photonics-based 800G DR8 transceiver. Also at ECOC 2022, we introduced our 200G per lane EML, for which we received the Lightwave Innovation Reviews Award.
As we look forward to 200G per lane transceivers, VCSELs, when feasible, silicon photonics, and EMLs will continue to be used. However, achieving 10-kilometer reach at 200G per lane is a significant challenge. For that application, we're excited about our newest laser technology, called the DFBMZ, which stands for Distributed Feedback Laser with Mach-Zehnder. This is an indium phosphide CW laser, monolithically integrated with an indium phosphide Mach-Zehnder modulator, which is a very advanced photonic integrated circuit. This laser technology will enable 1.6T transceivers with up to 10-kilometer reach. This is state-of-the-art in 200G per lane laser technology and is another way we are differentiating our products using our internal component capability. Eventually, Datacom links are connected up to the Telecom network.
Telecom networks require extremely complex lasers and detectors, and we have one of the industry's broadest portfolio of telecom products and technologies. This is the result of decades of innovation and investment. As with longer-reach datacom applications, both indium phosphide and silicon photonics technologies are valuable for different telecom transceiver products. We're particularly proud of our indium phosphide photonic integrated circuit, or PIC, platform. Using this platform, we design and manufacture extremely complex devices, such as indium phosphide tunable lasers with semiconductor optical amplifiers and Mach-Zehnder modulators for tunable transceivers, and also narrow align with tunable lasers with dual in-phase and quadrature, or IQ, modulators, with on-chip amplification and dual IQ detectors for coherent optics transceivers. These are highly complex indium phosphide PICs, for which Coherent is one of the most experienced and highest volume suppliers.
Due to this integration capability, our 400G ZR+ product was recognized at ECOC last year with the ECOC Exhibition Industry Award for Optical Integration. One of the advantages of having internal indium phosphide and silicon photonics capability in-house is the ability to utilize this expertise to develop and introduce novel products. At the Optical Fiber Conference in March of this year, we unveiled the world's first 100G pluggable coherent transceiver in the QSFP-28 form factor, which we call 100G ZR. This transceiver plugs into existing switches and routers, enabling customers to upgrade their network from 10 gigabits per second to 100 gigabits per second over links stretching up to several hundred kilometers without replacing network equipment. We received the Lightwave Reviews Award for this product.
We were able to introduce this novel product well ahead of others in the market because we had the internal component capability to drive our own roadmap. At Coherent, we also manufacture our own gallium arsenide and indium phosphide photodetectors that receive the light and convert it back into electrical signals. In addition, we also have a fabless in-house integrated circuit design team, designing laser drivers, transimpedance amplifiers, and DSPs that are manufactured in Tier One silicon foundries and used in our transceivers. We also design and manufacture a number of passive optical components, such as isolators and circulators, that are used in our transceivers. All of these platforms contribute to differentiating our products and bringing value to our customers. It's important to note that all these component capabilities can be used for our transceivers that support Ethernet, InfiniBand, and proprietary links such as NVIDIA's NVLink.
In fact, our transceivers for AI and ML are protocol-agnostic, meaning the same transceiver can support Ethernet, InfiniBand, or NVLink. This means our internal components support all of these applications as well. There has been significant industry discussion about linear pluggable optics, or LPO, half linear pluggable optics, or HALO, as well as near package optics, or NPO, and co-packaged optics, or CPO. It's important to note that these different implementations can be thought of as packaging and architectural partitioning differences compared to traditional retimer pluggable optics. But the underlying component technologies remain largely the same. For this reason, our internally designed components can support LPO, HALO, NPO, and CPO, and when these implementations emerge, we will be ready to support them with our internal components. There is no one-size-fits-all when it comes to transceiver and laser technologies.
With our broad range of technology platforms, including gallium arsenide, indium phosphide, and silicon photonics, we have the luxury of choice for any given application. We look at the detailed requirements for the application and choose the best technology, considering power consumption, optical specifications, reliability, cost target, and market window of the application. We do detailed simulations and measurements and make objective decisions based on these results. Because we have all the technologies in-house, we understand the trade-offs between technologies and can choose the best technology for the application on an application-by-application basis. I began my talk by describing our strong position with our complete portfolio of lasers that are capable of 100G and 200G per lane, which enable 800G and 1.6T transceivers for artificial intelligence and machine learning.
Having the full set of lasers really matters for a few important reasons. By being in control of the laser, we control the pace of development, and in addition, our semiconductor laser manufacturing capacity offers our customers strong support for their aggressive manufacturing ramps and also provides them with security and diversity of supply. With our entire portfolio of 100G and 200G per lane component technologies, we will be able to support 800G and 1.6T transceivers, which we believe will be in high demand for the remainder of this decade. Our components support Ethernet, InfiniBand, and proprietary links such as NVIDIA's NVLink, as well as traditional retimed transceivers, LPO, HALO, and eventual NPO and CPO.
As the leader in transceivers for over two decades, we expect to continue to innovate, to remain a technology leader, and to deliver volume solutions to our customers to enable data centers and telecommunications networks that meet their needs. We're excited to be part of the hardware solution that is helping to bring forth the power of artificial intelligence and machine learning. Thanks for watching us as we reframe what this industry is all about as we go forward. Now, let me turn it back over to Paul.
Thank you, Julie. Michelle, if we could open up the line to questions from the analysts, and when those questions are done, I'm going to pose questions that I received via email from investors. Thank you, Michelle.
This concludes our presentations. We will now begin our Q&A session. To ask a question, please press star one one on your telephone and wait for your name to be announced. To withdraw your question, please press star one one again. Please hold one moment while I compile the Q&A roster. The first question comes from Joseph Cardoso with JP Morgan. Your line is open.
Hi, thanks for the question. This is Joseph Cardoso from JPMorgan on for Samik Chatterjee. You know, first, just thanks for putting together this presentation. Really interesting and useful information. So just diving into my first question here, you know, you outlined that datacom and telecom SAM both growing low to mid-teens through 2028 respectively, you know, which I guess was a bit of a surprise, just given that most investors expect datacom growth to outpace telecom over the near and long term. So maybe you can just dig into the assumptions underlying the telecom SAM you presented, and what the main drivers or areas of investments that you see unfolding over the time horizon that get you to that low to mid-teens growth rate. And then I have a follow-up. Thank you.
Thanks, Joe. I'll ask Sanjai to respond to that.
Okay. So thank you for the question. We see both markets growing over the longer term, the five-year term. In telecom, there are areas such as coherent transceiver growth and some amount of disaggregated systems growth. So we are seeing that growing at a much faster clip than the more traditional part of telecom. So that's our assessment of the market.
Got it. And then maybe more on the R&D investments that you guys are doing around DSPs. You know, specifically, can you elaborate around your thoughts relative to how strategically important it is for you to invest in this in-house? And then when you think about the roadmap here, are you looking to develop it for specific use cases or applications and then leverage merchant for others? Or are you eventually looking to bring the solution completely in-house? Curious around your thoughts there. Thanks.
Yeah, I'll ask Beck to respond to that. Go ahead.
Sure. Thanks for the question. Yeah, we see that DSP technology is really important for our coherent transceiver product lines. Today, we have a mixture of both internal developments, where we are developing our own solutions, really where we want to have differentiated applications, or we don't see it well served by the merchant market. We have strong partnerships with merchant DSP suppliers, and we expect to continue those as we go forward. And so you can look for us to have a hybrid model, where either for very unique applications or very high volume applications, we would develop it internally, and we would also use a blend of external suppliers. As you know, DSP developments are extremely expensive, and this helps us kind of leverage and get best return on investment for our R&D.
Joe, I'm gonna go ahead and pick up on your question and ask Beck to expand, because I think a lot of people understandably would say, you know, Coherent, as in capital C, Coherent, given that you've got Ciena and Huawei and Nokia with their own internal long-standing DSP programs, and you've got the Ciena now Cisco and then finally now Marvell and NEL, in terms of merchant players, a very strong lineup. Folks have been shipping for years. Why should any of you folks think that we're gonna be successful? Beck, do you want to respond to that?
Yeah. I mean, I think our first solution right out of the gate has been very successful for us. We have a deep knowledge of the whole technology ecosystem in a coherent transceiver. I think that knowledge helps us really differentiate on our DSP capability. Our first solution in the 100 gig space is a really purpose-built DSP that's come in at a really differentiated power and performance capability. And it's so compelling that we have many large tier one customers that have their own internal DSP capability that are looking to source that solution from us. So it just shows you that we actually can differentiate and bring a really important value into that space with our own DSP solutions.
So we're not coming in with just a cookie-cutter solution?
Exactly.
All right. Joe, anything else?
Thanks for the color, guys.
Michelle, next question.
Please stand by for our next question. The next question comes from Tom O'Malley with Barclays. Your line is open.
Hey, guys. Good morning. Good afternoon. Thanks for taking my question. Mine is just on the handoff of speeds within the data center. On slide 24, you guys show the progression of hyperscaler growth at 200 and 400 G through fiscal year 2023. But on your earnings call, you guided September down pretty significantly and kinda commented on higher speed datacom rates being down. And then if you look down at the 800 G ramp that you have in Q1 to Q2, it's not that material of a revenue contribution. I just wanna understand, can 800 G ramp fast enough for you guys to see growth in that high-speed datacom bucket for the fiscal year 2024? Thank you.
So Tom, let me before I ask Lee to respond, I just want to clarify something. I don't think we've guided. We certainly didn't guide down 800 gig. So I'm not sure on what basis you made that statement, but I'll ask Lee to respond. Go ahead, Lee.
So thanks for the question. And we do have, on 800G, we do have a very significant backlog now, and we're going to ramp up progressively through the next four quarters. And the customers on 800G covers all of the customers with high volume demand on 800G. And we do see that we're going to have quite a significant growth on datacom transceiver business in fiscal 2024 versus 2023.
Helpful. Then my second one is just, I thought it was really helpful on slide 26, you kinda show your product portfolio from a transceiver perspective, and then kind of mention that you cover all major protocols, including Ethernet and InfiniBand. In the AOC products specifically, could you talk about if there's any share differences or where your share may stand between in Ethernet and in InfiniBand? Is there any material difference between the two of those? Thank you very much.
We view really the AOC as one form of a transceiver, right? Because it's, it does have, the fiber directly linking to end transceivers. And we sell to InfiniBand market, and the Ethernet market, both markets, in a significant volume. And we have significant share in both product families.
Yeah, Tom, as I think you're well aware, transceivers, ours and others, they're protocol agnostic with the exception of Fibre channel. And so whether it's Ethernet, InfiniBand, or NVLink, or any other proprietary protocol or standard, our transceivers can handle all of that. And I think we just answered your commercial question. Anything else?
That's it. Thank you, guys.
Please stand by for our next question. The next question comes from Meta Marshall with Morgan Stanley. Your line is open.
Great, thanks. Super helpful presentation. I guess I just wanted to get a sense of, you know, what are some of the bottlenecks to your production and being able to kind of meet some of these, objectives that you laid out in the deck? And just kind of secondary to that, what do you see as the greatest bottlenecks in kind of, the larger ecosystem if they're not kind of yours? I think, you know, GPUs are kind of understood, but maybe outside of the GPU ecosystem, you know, what are some of the hurdles within the ecosystem to 1.6 terabit adoption? That would be helpful. Thanks.
I'll ask Lee Xu to respond. Go ahead, Lee Xu.
So, thanks for the question. In the short term, the 800G ramp up is limited by the DSP availability or the DSPs ramping up. And, that bottleneck can be quickly eliminated. So that's why we're saying that we're going to have significant ramping up over the next few quarters. And, we have, as what I mentioned earlier, we have significant backlog, and most of which we're going to deliver in a timely fashion for our customers. And, as for the 1.6T development, that involves, you know, brand-new technologies in terms of a 200G per lane optics and 200G per lane electronics. And that's going to take a few years to mature.
Our first product with the 200 G per lane optics is going to be released a little bit over a year from now. The broader deployment of 200 of 200 G per lane or 1.6 T technology will take a few years.
Lee Xu, just to be clear, you're saying first initial shipment within the next year or so, volume shipment in terms of commercial acceptance and volume, going to take another couple of years thereafter?
That's right.
Okay. Mita-
Perfect. Thank you so much.
Are you good?
Yep. Yep, thank you.
Please stand-
Lee, I'm gonna, I'm gonna ask you to expand. Michelle, before you go on to the next, I'm gonna ask Lee to expand on Meta's question, because the question I've gotten quite often over the past several weeks, understandably, is: to the extent that we have bottlenecks, if we can't ship, is there a risk that we lose that revenue to other competitors?
I don't think so. As what I said, we get or keep getting significant booking on 800-gig transceivers. So far, when we get these orders, we make commitment on the shipment dates. So far, we can ship within the next year for the majority of the booking and backlogs. We don't think that we're losing business to anybody.
All right. Okay. Michelle, go on. Thank you. Thank you, Lee.
Please stand by for our next question. The next question comes from Sidney Ho with Deutsche Bank. Your line is open.
Thank you. Thank you for the very informative session. So I have a couple of questions on the datacom transceivers. One more on the technology side and the other one on the business side. First one on technology: Given the strength of silicon photonics and datacom transceiver market that you expect over the next few years, I think that's on slide 15, can you speak to the importance of co-packaged optics within that segment? I think it, another slide shows that it's gonna start shipping around end of calendar 2025. What is Coherent's competitive advantage in that technology, and can you maybe size the opportunity within that segment?
I'll ask Julie to respond to that. Good question, Sidney. Go ahead, Julie.
Hi. Yeah, thanks, Sidney. Yes, so as you mentioned, we do have, or you asked about, we do have an in-house silicon photonics design team. As you probably know, for silicon photonics, the silicon photonics portion generally does the receiving, it does the modulation, and it does some passive optics integration, but it doesn't generate light. So you have to have an indium phosphide CW laser to generate the light, and we are one of the few suppliers of that device. So that's a differentiated advantage that we have there, in addition to just the fact that we do have a very capable silicon photonics design team. Then as things, and actually, that indium phosphide CW laser is not so easy to make, actually, and there aren't that many people that are able to do that.
And so that is a significant differentiator, actually. And then as we look forward to potentially Co-Packaged Optics, I mean, Co-Packaged Optics, you can really just think of it as a repartitioning of the transceiver. And so again, it has a CW indium phosphide laser and silicon photonics die. So we can still provide that same thing for Co-Packaged Optics if you know, as long as our customers are interested in deploying that. As far as deployment of Co-Packaged Optics, I mean, one thing I've mentioned before, but I'll say again, is that it's very important to understand how much our customers love pluggable optics. They love pluggable optics because it's a multi-vendor ecosystem. It allows you to pay as you go, right?
It's also very easy to change in and out, so it gives them a lot of flexibility. Definitely we do see in the long term, people going to co-packaged optics, but the trade-offs are a lot. As it goes to co-packaged optics, it may be a couple of years. I don't know, Sanjai, if you have a comment on the timing of the market, but-
Yeah, I mean, I would say the 1.6T transceiver is a pluggable transceiver, so we know that. So that's gonna last for another four, five, six years. And beyond that, the jury is still out.
Yeah. Yeah, but anyway, as I said before, all the components that we have can be put in a pluggable transceiver, or they can be put in a little, you know, a chiplet that could go into co-packaged optics.
Great. Thanks. That's very helpful. Yeah, I have the other business question I have is, as I look at the AI-related demand that's going to accelerate over the next few years, how does that change your revenue distribution for datacom transceiver? I think in the past, you talk about one-third hyperscalers, one-third top 20 customers, and one-third kind of like enterprise. How is that split going to change over the next few years? And along the same lines, can you speak to the financial implications of a change in revenue contribution from the different end markets and customers? Thank you.
All right, before I ask my colleagues to respond, Sunny, let me just remind you and everybody else on the call that we did say this past quarter, that for Datacom, web-scale was two-thirds over Datacom revenue. You're right that historically it had been one-third. This past quarter, the web-scalers were two-thirds of total revenue. Now I'll ask my colleagues to respond.
What we think is that with the AI search, that our high, the content or the proportion of the high-end, higher speed transceivers is going to get more. The sales to the hyperscaler and people who they're actually selling to hyperscalers, is going to get a bigger portion of our overall sales.
Yeah, again and again, Sunny, it's already two-thirds of total. And it stands to reason that as AI becomes a bigger and bigger piece of the pie, that revenue will shift in the direction of hyperscalers. And I think that's true for everybody-
Okay.
Not just us.
Okay. Thank you.
Thank you, Sunny.
Michelle?
Please stand by for our next question. The next question comes from Atif Malik with Citigroup. Your line is open.
Hi, thank you for doing this event, and thank you for taking my question. My first question is for Sanjai. Sanjai, if I look at slide 13 and 14, in slide 13, you're saying that the non-AI portion of the transceivers will come down next year. And then in slide 14, you're indicating that the 400 gig will start to get an uplift next year. So, I want you to reconcile what is it driving that 400 gigs pick up next year? Is that AI related, or is it something else? And then I have a follow-up.
Yeah, okay, thanks for the question, Atif. So on slide 13, what we are showing is the—when we talk about AI growth, we include everything. It's not just 800G. The majority of it is 800G, but there's 400G included as well. So the AI piece grows at a 47% CAGR. The non-AI piece grows at 10, and then that's how we arrive at a combined 18.
On the second part of your question on 400G, we think that on the traditional networking, so say the non-AI portion, we think there's a little bit of a digestion that's going on in 400G at 200G and 400G over the next 6-12 months. And then it'll pick up again after that digestion phase. We pass that phase, then the growth picks up again. That's our view of the market today.
Thanks, Sanjai.
Very helpful. As my follow-up, in response to an earlier question from Meta, you guys talked about DSP being a supply bottleneck for you. A leading DSP player, Marvell, seems to have shipped even more DSPs than they expected a quarter ago. I guess my question is, are they shipping more to your competitors, or is the hiccup more internal, some internal DSP program that you have in-house?
Well, thanks for the question. This is Lee. So when we say it's a bottleneck, for a short term, that means relative to our customer, well, our customer want us to ship, it's a bottleneck. But we are-- we have been increasing our, kind of 800G shipment, in a very fast pace, from Q4 to now Q1 and Q2. For example, this quarter's shipment is more than double of what last quarter's shipment for 800G. So, you know, their, Marvell's shipment, actual shipment portion of that is certainly coming to us.
Yeah, and Atif, just to be clear, as we go throughout the year, we're expecting and as we said on the earnings call, we're expecting 800G shipments in revenue to increase meaningfully throughout the year. With each passing quarter, as we get more and more of the particular issues get resolved, we expect our ability to ship and ability to recognize revenue to increase meaningfully. And it will be a back half-weighted year for 800 gig. Notwithstanding that, in the first quarter, we'll ship meaningfully more than in the fourth quarter of last year. In the second quarter, we'll ship meaningfully more than in the first quarter, et cetera.
Thanks, Paul.
Thanks.
Please stand, please stand by for the next question. The next question comes from Christopher Rolland with Susquehanna. Your line is open.
Hey, guys, this is Matt Myers on for Chris. I wanted to dig in a little bit into your vertical integration strategy. Can you guys talk about what you still outsource here for on the datacom side? I know you guys talked about PAM4 transceiver, PAM4 DSPs, being outsourced, but do you have your own TIAs, laser drivers, analog CDRs, linear direct drive? Any of these, are you interested in doing yourselves? How often do you use these products, and what do you need to use across your lineup for your various products here?
I'll ask Julie to respond to that. Thank you.
Yep, thanks for the question. Yeah, as you know, so in the datacom transceivers, you have the DSP, you have laser drivers, lasers, photodetectors, TIAs. And as you pointed out, generally, for datacom, we are using merchant DSPs. And we have the capability to do almost all the lasers and photodetectors in-house, although sometimes we do use external vendors. We, you know, our in-house R&D, we prioritize what we work on, and we can't work on everything at the same time within the budget we're willing to allocate to that R&D and be successful at it. So we do purchase things from the outside market as well as use our own internal capabilities, and that mix may change over time on any given product. We do also like to second source where we can, just for security of supply.
So sometimes you may see us buying things from the outside market just to make sure that we have a second source. As far as the laser drivers and TIAs go, we do have a strong capability in-house to design those, but we do also use a significant number of externally purchased from OEM merchant vendors, just again, as a mix. And what we do is we prioritize where we spend our R&D in-house by sort of two simple factors. The first is a simple business case analysis, where it's going to have the biggest financial impact on the company.
The second is strategic considerations, like the one that Beck mentioned, where we were able to introduce a product well ahead of everybody else in the market because we had the component capability, and we chose to design a differentiated product that required new components, and we chose to invest in those components ourselves. So we'll continue to have this hybrid model, and we leverage the strengths that we have in-house, but we do have strong, strong relationships with suppliers as well.
Thank you, Julie.
Got it. That's really helpful. I guess just another quick follow-up here. Is there a way to think about an attach rate from a GPU to a transceiver for you guys?
I'll ask Lee to respond to that. Lee or Sanjai.
Attach rate transceiver to GPU.
Yeah, in general, this is, you know, every configuration is different. But if you look at the way the high-end type of applications, let's say typically 8 GPUs, they're about 8 modules, 8 high-speed modules for that application. But on the lower end, the applications that are there are less number of modules per CPU, so.
On average-
Hopefully, that gave you some idea.
Would you say on average across systems, it's greater than one, less than one? How would you think about that?
The high-end ones are sort of one is to one, so I would say it's, it would be less than.
Thank you.
Sure.
Please stand by for the next question. The next question comes from Ruben Roy with Stifel. Your line is open.
Thanks very much, and thanks all for the very informative presentation. I think this question picks up on some of the comments that Julie was just making, and maybe Sanjai can comment as well. But was wondering, as you, you know, sort of move into more complex technologies, faster data rates, et cetera, and as you make these decisions between internal components and outsourcing, et cetera, historically, the margin structure of modules and transceivers specifically has been sort of volatile. And certainly for Finisar, I think that would be the case, and I would love to know if you disagree with that.
But as you look ahead, as the technology gets more complicated, difficult to manufacture, et cetera, if you could spend a few minutes on how you're thinking about margin structure, just qualitatively, in new technologies versus, you know, sort of the older technologies. And Julie, I think you did mention, you know, there—this has been a difficult competitive landscape to participate in. How do you see that evolving as you get, you know, to 800 and beyond, up to 1.6? Thank you.
Ruben, I'll ask Julie to respond, but I'm sure Julie would agree, transceivers has always been a competitive market. It's. That's not new and different. It's not something unknown to us and unknown to our predecessor, Finisar. It's always been competitive, and we assume it always will be, but we've, we've done a pretty good job in that context. But let me ask Julie to respond.
Right. Yeah, I've worked in transceivers almost my entire career, and I always say it's not an industry for the faint of heart. But we have been number one for over 15 years, and we have no intention of that changing going forward. We feel like we're only getting stronger. But to be more specific, I think what you see is, as you go through hearing about what we're saying, what other companies are saying, I think you see the Datacom supply base of transceivers getting more concentrated under fewer vendors as the data rate goes up. It's just really hard. It's hard to make these components, it's hard to make these transceivers, and also the pace with which we have to go to the next data rate is really fast.
In the older Ethernet days, you know, we might be first to market, and then there'd be 10 other guys that come in after a couple of years that make it even more challenging. Now, by a couple of years, you got to be on the next data rate. That kind of pace and the fact that the technologies are getting harder and harder, generally, well, definitely benefits us. I think you'll see. I'm hopeful that the just compression of the total number of transceiver vendors that can attack this space with the complexity and on the time frame that the technologies have to change is going to be fewer. You're going to see more concentration, which generally, roughly speaking, is helpful to margins. That being said, it's always a competitive market, as you pointed out.
I forgot the second part of the question. What is it?
Ruben, can you repeat the second part of your question?
That was, that was really the bulk of it. I, I was just trying to see if you could give us an assessment of how you're, you know, thinking about margin structure of 800 and up transceivers versus sort of, you know, older generation technology.
Yeah, I-
Ruben, we have said publicly that, as a general proposition, as a general proposition, the latest and greatest, in this case, 800-gig, measure 1.6, with the benefit of volume and assuming no yield issues, should have higher ASPs and should have higher margins than previous generations. That said, all transceivers tend to be below corporate average. This is not a new revelation. This has always been true. We've spoken about in the past. To be clear, as a general proposition, I'm going to let Julie speak.
Yeah, I could just add one more comment, which is, it's largely why we invest in our own components, because if you don't invest in your own components, it's basically an assembly and test business, right? But as we invest in our own components, then that's margin we're not paying to other people, and that's what allows us to have the margins that we do have on transceivers. So it's a, it's a, you know, important part of our strategy, actually, is to do those components ourselves and also do the manufacturing in-house ourselves.
Got it. That, that's really helpful. I appreciate it. And, Paul, if I could ask a quick follow-up of back. Just wanted to see if we could talk a little bit about how you're thinking about, coherent pluggable adoption in telco. You know, where are we in that timeline? And, you know, how do you see that evolving?
Yeah, I think what we're seeing is a gradual shift in the market away from historical line card-based transmission solutions, more and more to coherent pluggables. The capability of coherent pluggables continues to improve. You know, our latest generation of 400G ZR products give you much greater reach. And I think the other trend that's driving it is the increasing push towards disaggregation at the CSP level, which is making coherent pluggables the right solution to kind of grow that. So, you know, longer term, we see coherent pluggables obviously being a much larger than 50% portion of the overall port market in coherent systems.
Very helpful. Thank you.
Thank you, Ruben.
Please stand by for our next question. The next question comes from Dave Kang with B. Riley. Your line is open.
Thank you. Thanks for the presentation. My first question is on slide 19, where there's a traditional versus a disaggregated networks, and the ROADM. Our ROADMs appear to be replaced by a couple of arrows. Just wondering if you're implying that ROADMs are going away, and what do these arrows represent?
Yeah, I mean, maybe we'll do that as a two-part. I'll talk about ROADMs first. First of all, I think the ROADM market continues to grow and be very, very healthy. We see increasing ROADM applications. You know, we have a set of very low port count edge ROADMs that are starting to proliferate into edge access networks, which is a big growth. We even deploy ROADMs in undersea wet plant applications in submarine, which is interesting. Overall, we're projecting a very steady and healthy growth for the ROADM and WSS market over the next five years. I think it's proven to be a very capable technology.
What's interesting is that the customers are still unlocking more of the fundamental capability of the ROADM physical layer as they deploy more sophisticated network operating systems and network management systems. So yeah, I'd say in disaggregated networks, the ROADM will play an important role and continue to be a vital part of the network.
Got it. My follow-up is on 800-gig. Just wondering who you view as main competitors and what is the initial take on the market share for this market?
I'm going to ask Lee to respond to the first part, and, and I'll take the second part of the question. Go ahead, Lee.
Okay. Our main competitor, what we see in our customers, are it's mainly InnoLight.
So Dave, on the second part of your question as to what do we see as our share, you know, recognize as we said it publicly, we only started shipping 800-gig in the March quarter or fiscal third quarter. And so since share is obviously a backward-looking concept, I would argue, no disrespect, but our share today is somewhat irrelevant. The real question I think you and everybody else wants to know is, what will our share be in the future, and how does that translate to revenue growth? And what's the associated attending margins of that revenue growth? And as we've, I think we've indicated in terms of the revenue we have in our guidance related to 800-gig, the revenue that we didn't put into our guidance, that's several hundred million, as well as what we shared with you today.
Clearly, demand, measured by orders, measured by customer engagement, is surging from what had been a very low level when we first started shipping, to an already meaningful, increasingly meaningful level. We think our share going forward will be far more meaningful than our share looking backwards. Once again, we've only been shipping for two quarters now, for third quarter shipment. We also think given our technology roadmap, our current, the breadth and depth of our technological and product solutions, that we do expect to gain share, meaningful share over time. Lee, anything you want to add to that?
Yeah. I want to add, overall, we are the number one datacom transceiver leader, right? And we have, overall, the Ethernet InfiniBand market. We have about 20% share. That's covering all the data rates. And, as for the higher data rates of 800G and the future 1.6T, we expect to get significantly more share than that.
Got it.
Okay, thank you.
Thank you. Just one final question is, sticking with competition, InnoLight appears to be your main competitor. Them being a Chinese vendor, is there pressure on North American hyperscalers to go with American supplier rather than Chinese supplier?
We do see some of our customers express they're more willing to work with us using our parts.
I think if I could add also the fact, Lee, as Lee mentioned, that we have a huge manufacturing footprint in China, but we also have a huge manufacturing footprint in Malaysia. So we offer our customers that diversity of supply as it, you know, as their wish, basically, you know, to address that. And also not just being a, also being a U.S. company, the majority of our IP is U.S. and European-owned, and that's also increasingly becoming important to our customers.
Got it. Thank you.
Thank you.
Please stand by for our next question. The next question comes from Chris Grenga with Needham & Company. Your line is open.
Hi. Thank you very much for the detailed presentation and for taking the questions. This is Chris Grenga on for Jim Ricchiuti. Just, to what extent do you see China contributing to data comm transceiver AI application growth? And do you ascribe any probability, or is there any concern that government export restrictions on AI-enabling technology could broaden to encompass some of these components? Thank you.
I'll ask Lee to respond to that. Thank you, Chris.
Okay. So we sell to all the major AI infrastructure building companies in China. But what we do see is that their usage of AI or demand at the consumer level is about one generation behind their U.S. counterparts. So, you know, we are always having higher share in more advanced product than previous generation products.
Chris, to be clear, I think what we just told you is that the current AI-driven demand for 800 gig is not currently coming from the Chinese.
Got it. Thank you very much. And, could you perhaps discuss a bit more how the 200 gig per lane VCSEL could the significant differentiating factors that are related to that technology? And, it looks like that technology is kind of factored in, you know, beginning in fiscal 2026. Just curious, you know, what visibility you have into the development of that you feel confident putting that projection in fiscal 2026. Thank you very much.
Thanks, Chris. I'll ask Julie to respond to that.
Sure. So, you know, any directly modulated laser technology, well, all technologies, but in general, the directly modulated laser technologies have gotten tougher and tougher as we approach higher and higher speeds. I mean, there was a period of time when people thought you couldn't make directly modulated lasers at 14 G, actually, right? So we are pushing the physical limits of laser technology and including VCSELs. But what we do see is for the AI, it is a very cost-effective, power consumption effective solution. And then the lion's share of AI links are less than 30 meters so that's why we felt if, if someone can make a 200 G VCSEL a reality, that there-- we believe there will be adoption of that. That being said, it's very challenging 'cause we are approaching physical limitations of these type of lasers.
So we do have some initial results that we're excited about, but we also feel like we have a lot of work still left to do, which is why we've said that it would probably be 2025 or beyond, for that.
Thanks very much.
Thank you, Chris.
This concludes the Q&A session.
Michelle-
I will now turn it over to-
Yeah, Michelle, so we're in an effort to be transparent, and we were trying to be as transparent as possible, and I'm also cognizant of that on these things, you could get attention fatigue. But we're gonna err on the side of lengthening the session, and I've got a number of questions that have come in from the investment community via email, that I'm gonna pose to my colleagues on behalf of those investors, and I hope it'll be appreciated by all of you. So first off, looking at slide 10, telecom SAM is doubling while CSP CapEx is relatively flat. What is driving that disconnect? The mix shift referenced on slide 9 away from platform toward pluggable optics, or is there something else driving that differential? And I'll ask Beck to respond to that.
Sure. I mean, I think the big thing driving that differential is the ICP CapEx spend. And what's been happening over the past several years is actually the CSP CapEx has been relatively flat, but ICP CapEx spend has been growing steadily. And last year, ICP CapEx spending, as a percentage of total telecom market, actually surpassed CSP CapEx spending, and that was a very interesting dynamic, and it still continues to grow at a pretty fast rate. And so, largely, as we dial forward, that's gonna be the big driver of growth overall in telecom CapEx spending next few years.
Beck, if I may, I'm gonna put on my not-so-old sell side hat and ask you, is also part of the answer that within the construct of total CapEx, that the future continues to bend towards optics? That optics is on the receiving end of more and more spend, whether you think about 5G or eventually years away, 6G, and we think about having access shift in the fiber, et cetera, et cetera.
Yeah, very true. So I think the other big dynamic for us, which is important, is that the value that the optics represents in telecom networks is increasing as a percentage of the overall spend, right? And if you dial back several years ago, optics was less than 50% of the total value, and now it's north of 60% and maybe approaching 70% of the total value deployed the net. And that obviously benefits us. It drives our market opportunity up over time. And I think the other point that maybe is alluded to in the question is, trend towards disaggregation also helps to broaden our market opportunity.
Great. Thanks, Beck. All right, I'm gonna keep going down the list. There, there are a handful of questions for those of you who want to hang on, and this is archived, by the way. In slide 13, it looks like the non-AI portion is slightly down in 2024 versus 2023. Slide 14, the 400G and below speeds are up in 2024 versus 2023. What will be the difference between the 2023 to 2024 takeaway from these charts? Are you guys counting 400G and below as non-AI or 800G and above as AI? Sanjai, I'll ask you to respond to that.
Yeah. Yeah, thanks, Paul. I think we kind of touched upon this a little earlier. On slide 13, that chart includes all AI. So it's not just 800G, it includes 400G and 800G and beyond. And so what this tells us is the traditional 400G, 200G non-AI applications have slowed down. There has been a digestion phase that's going on that we think will pick back up in the out years.
Great. Thanks, Sanjai. All right, I think this next one's for you, Beck. What is the current mix of transceivers that use internal DSPs versus external provider DSPs? I assume the individual asking the question is referring to our telecom business. As we said before, in datacom, we source those PAM4 DSPs from third parties. We don't make those. The other parts of this question is, and where can that mix be in 3-5 years? Would you ever start to sell DSPs externally? If so, how big could that market be for Coherent, and what would the margin profit profile look like? Go ahead, Beck.
Yeah, there's a lot of parts to that.
Mm-hmm.
First, I wanna say that we are just now introducing our first solution, based on our internal DSP, and we expect that to begin ramping into volume early next calendar year. That's a very important product for us, in a differentiated area. Over time, I expect that the mix of transceivers that we make with internal DSPs will increase relative to where it is today. Obviously, I don't know if I wanna speculate on an exact number of where it's gonna be in 3 to 5 years, but I would say that our internal DSP solutions represent a material portion of our total DSP consumption in the next 5 years, and it'll be an important aspect of our vertical integration strategy for the product.
Thanks, Beck. Sanjai, I think this next one's for you. I think we previously addressed this, so I'm not sure how much more you can say, but I will ask the question on behalf of the individual that posed it, which is: It is clear that there is, there's is 2 CPU per server going to 8 GPU per server, a 4x increase in the number of transceivers needed, or a 4x increase in the number of connections the transceiver needs to enable? Is there a way to break down the 40%-50% CAGR into unit versus ASP?
Okay, well, we touched upon this a little earlier, Paul, but let me try. So every configuration is different, so it isn't a one-size-fits-all in terms of these architectures. But if you take a high-end application, let's say, for example, that has 8 GPUs, those 8 GPUs need, typically that requires 4 optical high-speed—sorry, 8 high-speed optical modules, so sort of a 1-to-1, and that's on the high end. But on the lower-end applications, the ratio is less than 1, so more CPU, less modules. That's our best understanding today.
Okay. This next one is for Lee and Julie. So Coherent has all protocol support, but there's a difference in the opportunity market share. But is there a difference in the opportunity market share between InfiniBand versus Ethernet? Julie, you want to respond to that? Go ahead.
Sure. Yeah, as we said, and you alluded to, our modules can support all of the protocols, and because they can, sometimes we don't even know what protocol they get used for. We definitely see, initially, the InfiniBand and proprietary links are growing very, very, very rapidly. But there are plenty of people that are planning to use modified Ethernet for these applications, so the mix will probably change over time.
And as a reminder, commercially said in print and we've said verbally that we've got relationships with virtually-
Yes
All of the major. I'm sure there are some deployments we don't know about, but we have relationships with virtually all of the major web scale, cloud, and companies in those ecosystems.
Right, and the people who sell into the web scalers. And from our perspective, we don't have, we view all the opportunities are good opportunities for us. We feel we're engaged with the right people, we have the right products, and it's not that if one wins versus the other, that has a material difference to us.
Julie, the emphasis has a related question: Is the opportunity greater with the build/ramp of training versus inference models, or is the demand/requirement for Datacom transceivers the same?
Yeah. So, training is, you know, very GPU-intensive, so it requires a lot of compute resources, and that means there's a lot of optical interconnect between those compute resources. Whereas the inference is generally not as compute-intensive, so the training is beneficial for sure, for optical build-out. Yes.
All right. The next question, I think we already answered in part, and, and I'll answer the rest of it, which is: What are the components that are short relative to 800G? We mentioned TAM for DSPs. We've mentioned previously test sets. Those are two of the big issues. Not the only issues, but two of the biggest issues. Anything anybody else wants to add? Okay. Moving on, just a couple more, for those of you who want to hang in. My apologies. Here we go. Sanjai, I think this next one's for you or for Lee. Level zero versus level one, not layer zero, but level zero versus level one analysis on page 12. Couple of questions. Level zero would seem to be the more net new opportunity for optical transceivers, AOCs, given transition from copper.
Would we agree, and how is Coherent positioned versus NVIDIA, Fabrinet, AOI, et cetera, in tier zero AOCs as a supplier competitor? I think we touched on this earlier.
We did, yeah. I mean, the level zero, yes, the answer is absolutely the, that is a lot of the opportunity that we see, the short reach interconnects between the GPUs, which we call level zero. So that is. But that is not the only opportunity. There is also associated with these level zero interconnects or level one interconnects. We also factor that in when we size the market. So those are more traditional Ethernet today.
If I could just add that up to 50 meters in that level zero is very much favors the VCSEL-based solutions, which is one of the, you know, an area that we're very, very strong.
All right, and two related questions. Level one, which seems to be the more traditional transceiver market, would you expect to see growth here despite drop off of lower speeds?
So when we factor in the AI, when we size the AI opportunity, it, we include level zero for sure, but we also include level ones that are associated with these level zero buildouts. So some of it is already included in the AI piece. So yeah, I mean, the answer is absolutely, they're both important opportunities.
All right, one more. Sanjai, this is for you as well. It's going to be a market question. In terms of TAM outline on page 15, how do those TAMs relate to some of the current data points we see? Example, Fabrinet to something, I can't think, it's garbled, but $200 million plus with NVIDIA in September quarter. AOI talking about I apologize. AOI talking about $200 million plus per year with Microsoft.
I mean, the chart on 15 includes, as I said, it includes not just 800G, but also 400G. We believe the 800G opportunity, 800G alone in calendar 2023 is a $600 million opportunity, which is gross to about $2.5 billion in calendar 2024. So, that's our view.
Great. Thanks, Sanjai. And that's all the questions. I think I got through them all. If I missed, if somebody sent me something that I missed, I apologize, but please feel free, or whether or not you send me a question, I'll say it again, please feel free to reach out to me or my team. I would be happy to try to help to the best we can. And for that matter, we welcome feedback. And I mean that. I would appreciate any feedback as to how we could be communicating better, both events like this, around our earnings call, and throughout the quarter. Thank all of you for joining. Thank you, Michelle.
This concludes today's conference call. Thank you for participating. You may now disconnect.