Good afternoon, everyone, and welcome to the Kopin Corporation Third Quarter 2026 Earnings Call. Please note that this event is being recorded. At this time, I would like to turn the conference call over to John James, Investor Relations for Kopin. Please go ahead.
Good afternoon, everyone, and welcome to the Kopin Corporation Third Quarter 2026 Earnings Conference Call. During today's presentation, all parties will be in a listen-only mode. Following the presentation, the conference will be open for questions. This conference is being recorded today, and the earnings press release accompanying this conference call was issued earlier today. Before we get started, I would like to remind everyone that during today's call, we will be making forward-looking statements as defined in the Private Securities Litigation Reform Act of 1995. These statements are based on the company's current expectations, projections, beliefs, and estimates, and are subject to a number of risks and uncertainties that could cause actual results to differ materially from those forward-looking statements.
Potential risks include, but are not limited to, demand for our products, operating results of our subsidiaries, market conditions, and other factors discussed in our most recent annual report on Form 10-K and other documents filed with the Securities and Exchange Commission. Although the company believes that the assumptions underlying these statements are reasonable, any of them can be proven inaccurate, and there can be no assurances that the results will be realized. The company undertakes no obligations to update the forward-looking statements made during today's call. Kopin Corporation's Chief Executive Officer, Michael Murray, will begin today's call with an overview of Kopin's strategic progress and business developments during the third quarter and the period since. Following Michael, Kopin CFO, Erich Manz, will review the company's Third Quarter 2026 Financial Results. I would now like to turn the call over to Michael Murray. Michael, please go ahead.
Thank you very much. Good afternoon, everyone, and welcome to our third quarter 2026 earnings call. In the third quarter, our funded development programs continued to scale, and they are setting the stage for future product revenue as these programs move from R&D into production. Revenue grew 52% year-over-year and 43% sequentially to $18.2 million. Our loss from operations narrowed to $500,000 , a $3 million sequential improvement from the second quarter, and our operations generated positive cash flow. Erich will take you through those details, but I wanted to start with what is driving this performance. Our model is quite simple. We utilize government and partner funding to develop platform technologies, build the manufacturing capability for them here in the U.S., and convert that capability into multi-year programs that will, in turn, drive product revenues across the defense, medical, and AI infrastructure markets.
This quarter, every part of that model progressed positively. Today, I will cover three areas: our color MicroLED display program, our optical interconnect product line called Neural I/O and the work with Fabric.AI, and our defense franchise, including an update on first-person viewers and the Drone Dominance Program. Let me start with the color MicroLED. Building on three IBAS milestones we reported in July, including the successful achievement of more than 150,000 nits of full-color brightness in a single panel and the delivery of a U.S.-based bonding equipment structure to Westborough, Massachusetts in October. We announced an additional $18.6 million sole-source award from the Office of the Secretary of War through the U.S. Army Contracting Command.
Combined with our initial $15.4 million award, total IBAS funding for our color MicroLED platform technology now stands at $34 million, a continued validation of our technology and of the progress our team has made. The new award funds the design of two additional MicroLED displays, one for next-generation fixed and rotary-wing pilot helmets, and one for next-generation thermal weapon sights, widening the programs our technology can address from soldier-borne systems to weapon sights, helmet-mounted displays for aviation, among other applications. It also accelerates our effort to onshore MicroLED manufacturing in the United States and allows Kopin to compete on almost all the vertical market segments for microdisplays within our OLED and now MicroLED product offerings for multiple generations of production. We remain on track to transition to manufacturing MicroLED product in mid-2027.
During the third quarter, we demonstrated our color MicroLED display for Soldier Borne Mission Command, or SBMC, to the prime contractors and the SBMC program office, slightly ahead of schedule. The design is progressing as planned, and the results are quite impressive. Programs like SBMC could represent hundreds of millions of dollars in revenue for Kopin over the life of the program, and we recently received small prototype orders from SBMC prime contractors who want to evaluate our displays and capabilities for the program. Our manufacturing investments also earn support at the state level. In October, Kopin received two major manufacturing honors at the 2026 Massachusetts Manufacturing Mash-Up. First, we were awarded an $800,000 grant through the Massachusetts Manufacturing Innovation Initiative, or M2I2, for advanced equipment in Westborough that will help mature our MicroLED and Neural I/O manufacturing from technology readiness Level 4 to levels 5 and 6.
Second, Kopin was named Manufacturer of the Year by the Massachusetts Legislative Manufacturing Caucus, which is a high honor given the caliber of manufacturing companies across the Commonwealth, and a recognition that belongs to our engineering and manufacturing teams. I want to thank them for earning the recognition they so richly deserve. Recently, we also announced that we have collaborated with our town and state representatives to access a new 30,000 sq ft facility, which resides beside our current headquarters. The new facility will serve as our corporate headquarters, while our current facility will only support the new manufacturing and production lines due to the demands we are experiencing. Erich will provide more financial details on the facility and the $1.4 million tax credit associated with it.
These investments serve both of our strategic growth markets, since our U.S.-based MicroLED production line will support both our defense displays and Neural I/O transmitters. This creates a structural volume and absorption advantage that is very difficult for our competitors to replicate. That brings me to Neural I/O, where our program with Fabric.AI continues to move quickly. Since our last call, we confirmed the platform's first public demonstration at CES in January of 2027, and formally launched the joint technology steering committee that I lead. Fabric.AI has also engaged the global consulting firm, Kearney, to lead the commercialization strategy for their platform. On the technical side, we have begun initial transmission link testing, and we remain on schedule for our demonstrations at CES and SPIE Photonics West early next year, with the first-generation production silicon targeted for 2028.
We believe seeing Neural I/O in action will continue to generate significant and additional interest in the product line than we are already experiencing from potential customers at this point. In October, we published our second joint white paper with Fabric.AI called "The Data Center in a SerDes-Free World," which looks beyond the individual link to the entire AI system and has already been downloaded several hundred times. By replacing serialized electrical links with massively parallel MicroLED-based optical channels and extending high-bandwidth connectivity from approximately one to tens of meters, Neural I/O could expand a tightly connected compute domain from roughly 72 GPUs today, row-scale configuration spanning hundreds to thousands of connected GPUs. For a 10,000 GPU system, the paper models 1.3x to 1.6 x greater useful training throughput per GPU, roughly equating to 23%-38% less training time for the same workload.
By reducing communication, synchronization, and reliability losses, these are all modeled results, but create a tremendous amount of opportunity for the future. They do explain why our customers and the conversations with our customers are shifting from the cost of an individual link to the utilization of the entire data center and the total cost of ownership of the rack and the interconnect system as it pertains to power, water, and commissioning savings. Those conversations are translating into commercial momentum. Several additional AI infrastructure companies and leading semiconductor firms have signed non-disclosure agreements to review Neural I/O's progress, and we are actively bidding new solutions with several new potential customers. Our Neural I/O opportunity pipeline grew rapidly over the past quarter, driven by customer confidence in our ability to scale MicroLED manufacturing here in the United States.
Furthermore, we have already begun submitting several significant customer bids and expect several more after the demonstrations. We also continue to engage U.S. defense and government agencies on versions of Neural I/O that must be manufactured in the United States as per the FCC regulations on production lines like the one we have in our building here in Westborough. Turning to our defense business, which remains the foundational business of Kopin, orders from new European customers continue to gain momentum. In September, we announced a $2.4 million OLED microdisplay order through Kopin Europe from a new customer, a leading international thermal weapon sight manufacturer, with deliveries beginning in the fourth quarter over a six-month period, and the potential for follow-on orders as the program advances.
That program has just entered low-rate initial production using a Kopin OLED display, and this order affirms our decision to bring the full OLED manufacturing process in-house. We continue to expect our OLED deposition line to come online in early 2027. Our defense backlog is supported by programs with congressional budget support through 2030, and several of our sole-source and IDIQ contracts give customers room to order beyond current quantities, along with multi-year revenue visibility for Kopin. Turning to first-person viewers for drones. Last quarter, we said we expected Drone Dominance Program volume orders to begin in late August following the program's phase II evaluations and awards. The program itself has moved more slowly than expected. The Department of War announced its Gauntlet 2 winners in mid-September, with plans to purchase 60,000 drones from that pool of vendors.
The program moved onto its next gauntlet, which we expect approximately another 60,000 drones to be acquired. The program also added Gauntlet 2.5, which we did not expect in July, which is currently in flight testing as a result. The timing of FPV volumes are tied to the program and has shifted due to the additional Gauntlet phase to test gravity bombing drones. Kopin's partners, however, fared very well in both the recent deep strike and close-quarter combat scenarios, and Gauntlet 3 is roughly scheduled to commence in February 2027. What has not changed is a significant demand signal. More than a dozen new Sentinel FPV customers have placed small prototype headset orders for use in Drone Dominance Program and within other Department of War and U.S. government law enforcement agency use cases.
Drone manufacturers need trusted U.S.-built FPV OLED displays and optics, and our Dual Situational Awareness design remains a clear differentiator for battlefield use. We have also adapted our go-to-market strategy and approach to match how drone makers want to buy. At AUSA in October, we announced an expanded FPV evaluation kit lineup. Optical module concepts and modular components will let drone manufacturers and integrators alike design our optics into their own systems alongside our Sentinel FPV goggle. This broadens our opportunity beyond the outcome of any single program award. At AUSA, we also showcased our full-color DarkWAVE day and night see-through heads-up display. We continue to work with drone manufacturers both inside and outside the program, including production opportunities for 2027, and we will update you as those convert.
To summarize, in the third quarter, we grew revenue 52% year-over-year, narrowed our loss from operations by $3 million sequentially, generated cash from operations, and demonstrated our color MicroLED display for Soldier Borne Mission Command ahead of schedule. Since the quarter ended, we announced an additional $18.6 million IBAS award that brings total IBAS funding to $34 million, a new OLED thermal weapon sight order, earned state support and recognition for our domestic MicroLED and Neural I/O manufacturing, and moved Neural I/O closer to its public debut at CES. These platforms, our domestic OLED manufacturing, multiple new color MicroLED displays, Neural I/O, DarkWAVE, and Sentinel FPV optics are all designed to serve multiple customers in global markets rather than a single customer program. As our Westborough manufacturing initiatives come online, more of what we sell will be built on capability we control here in the United States.
That is what makes our growth more durable and forecastable. Turn the call over to Erich Manz, our CFO, to review our third quarter 2026 financial results in more detail. Erich?
Thank you, Michael, and good afternoon, everyone. The third quarter was an important quarter for Kopin, and in my view, a strong validation of what we've been communicating over the past several quarters. We have consistently stated that our visibility to our forecasted revenue was good, and that executing on production and development programs would drive meaningful revenue growth, that we would maintain discipline around our operating expenses as those programs scaled, and that this combination would move us towards profitability. This quarter, we delivered on those objectives. This quarter is a good illustration of how our model works as funded programs scale. Revenue increased $5.5 million from the second quarter, led by our Fabric.AI development program, while operating expenses grew at a slower pace, and our operations generated cash. Before I walk through the numbers, one note on comparability.
Our third quarter 2025 results included two significant benefits related to prior litigation matters, a $5.1 million reduction of the litigation accrual and an approximately $3.3 million reduction of accrued legal fees recorded within SG&A. Those items make a year-over-year comparison for SG&A, operating income, and net income less meaningful, so I will also reference our sequential performance where it is helpful. Total revenues for the third quarter ended September 26th, 2026, were $18.2 million, compared to $12 million for the third quarter ended September 27th, 2025, an increase of 52%, and up 43% from $12.7 million in the second quarter. The year-over-year increase primarily reflects research and development revenue from our joint development program with Fabric.AI and grant revenue under our U.S. government color MicroLED award. Product revenues were $8.8 million compared to $10.7 million in the third quarter of 2025.
Sequentially, product revenues increased 15% from $7.6 million in the second quarter. Non-product revenues were $9.4 million compared to $1.3 million a year ago. The increase primarily reflects $5.4 million of research and development revenue from our Fabric.AI program, as well as $3.1 million of grant revenue under our color MicroLED award. As a reminder, Fabric.AI is funding up to $15 million worth of development activities to successful demonstration. Through the end of the third quarter, we had received $9 million worth of that funding, including $4 million during this quarter. Together with our IBAS color MicroLED award, which now totals $34 million, these funded programs run over multiple quarters, and as they move from R&D into production, they set the stage for future product revenues.
Cost of product revenue for the third quarter of 2026 was $6.4 million, or 73% of the product revenues, compared to $8.4 million, or 79% of net product revenues in the third quarter of 2025. The improvement, including the reduction from 86% in the second quarter of 2026, was primarily attributable to lower scrap levels and certain favorable items that were one-time in nature. Although the improvement is positive, the benefit of these items is not expected to be fully recurring, and accordingly, the third- quarter cost ratio should not be viewed as representative of the company's anticipated cost structure in future periods. Research and development expenses were $5.6 million compared to $2.5 million in the third quarter of 2025.
Funded R&D increased to $4.5 million from $0.6 million, reflecting the work we are performing under our color MicroLED award and our Fabric.AI program, the same work that generates our grant and development revenue. Internal R&D decreased to $1.2 million from $1.9 million on fewer labor hours for internal programs. Selling, general, and administrative expenses were $6.7 million in the third quarter of 2026, compared to $1.6 million in the third quarter of 2025. The prior- year figure reflected the approximately $3.3 million legal fee reduction I mentioned. Sequentially, SG&A increased from $5.1 million in the second quarter, primarily due to higher incentive compensation, outside consulting, and legal fees, yet it declined as a share of revenue from 40% to 37% today. More importantly, we are gaining operating leverage as revenue scales.
Total operating expenses, meaning R&D and SG&A combined, were $12.3 million, up 28% from the second quarter, while revenue grew 43%, and declined 800 basis points to 68% of revenue, a reflection in our tight focus on SG&A spend. Excluding funded R&D costs, operating expenses were 43% of revenue for Q3 2026. Operating expenses were 50% of revenue and 57% of revenue, respectively for Q2 of 2026 and Q3 of 2025. Loss from operations was half a million dollars compared to income from operations of $4.5 million in the third quarter of 2025, which included the $5.1 million litigation accrual benefit and a legal fee reduction. Sequentially, our loss from operations saw a $3 million improvement.
Net loss for the third quarter was $3.4 million, or a loss of $0.02 per share, compared to net income of $4.1 million, or $0.02 per share in the third quarter of 2025. The quarter's results included approximately $2.7 million of net losses on investments, which largely reflect non-cash changes in the fair value of our strategic equity investments. Turning to the balance sheet, as of September 26, 2026, we had cash and cash equivalents of $25.8 million, up from $24.9 million at the end of the second quarter. Total cash, restricted cash, and marketable securities were $51.3 million, up from $50.3 million, and included $25.5 million of restricted cash, of which $24.2 million collateralizes the supersedeas bond for the litigation matter appeal.
Our operations generated approximately $2.4 million of cash in the third quarter, supported in part by funded, the development funding from Fabric.AI, reducing net cash used in operating activities for the first nine months of the year to $2.7 million, more than covering the $1.4 million of capital expenditures in the quarter, as we continue to build our domestic MicroLED and OLED manufacturing capability. For the remainder of 2026, we expect capital expenditures of between $1 million and $1.5 million, and we believe our current liquidity is sufficient to fund operations beyond the fourth quarter of 2027. On the BlueRadios matter, there have been no changes to our accrual, and we continue to expect no further developments in our federal appeal until approximately mid-2027.
Turning to our outlook, with $41.5 million of revenue through the first nine months of the year, an increase of 34%, we continue to expect a strong end to the year, along with meaningful progress towards sustainable GAAP profitability. Looking further out, our confidence in 2027 continues to build. Our funded programs are expanding, our domestic manufacturing investments are coming online, and our customer engagements are broadening across defense and AI infrastructure, all of which position us for continued growth next year. As it stands currently, we have good visibility into our revenue pipeline for 2027, and approximately 50% of our revenue is already in our backlog. Finally, as we recently announced, we are also taking steps to position our operation for future growth while maintaining our cost discipline.
In coordination with the state of Massachusetts, and in connection with our commitment to our location for the next 10 years, we have secured additional office space adjacent to our existing facility. This arrangement involves only a minimal increase in operating lease costs and will allow us to dedicate the additional space created within our current facility to manufacturing expansion as our demand grows. With that, I'll turn the call back to Michael for closing remarks.
Thanks, Erich. Before we take questions, a few closing thoughts. Since our last call, the U.S. government expanded its IBAS commitment to our color MicroLED technology, Neural I/O moved from architecture towards hardware demonstration, and our defense franchise continued to win new business, and our operations generated cash. What differentiates Kopin has not changed. We are the only company in the world manufacturing four types of microdisplays, and we invented the bidirectional microdisplay. We are the sole source provider on several Department of War programs of record, and our technology is now being applied to some of the fastest-growing segments of defense and AI infrastructure.
Coupled with the recent IBAS awards, manufacturing awards, and the investment received from the federal and state government, our defense partners, customers, and investors should feel confident in our ability to translate these new research and development contracts into significant, profitable, and sustainable product revenue in the near future. With that, operator, we can open the call for questions.
We will now begin the question-and-answer session. To ask a question, you may press star then one on your telephone keypad. If you are using a speakerphone, please pick up your handset before pressing the keys. If at any time your question has been addressed and you would like to withdraw your question, please press star then two. At this time, we will pause momentarily to assemble a roster. The first question comes from Jonathan Siegmann with Stifel. Please go ahead.
Good evening. Thanks for taking my question.
Hi, John.
Hey. IBAS, that was great news in the quarter, but that was after the quarter, so it is not in backlog. I was just wondering, first of all, is it appropriate to add that to the 35 that you reported, thinking about what you would be pro forma? Then, the run rate for that in 2027, how much of the original tranche is remaining, and how does that step up in 2027? Thank you.
So first question, it is additive. So, that would be roughly in the terms of $45 million-$47 million of backlog. Of the original IBAS award for 2027, roughly speaking, $5 million-$7 million, I believe, will remain and be consumed in 2027 from the first IBAS award.
Thank you. Then the nearest term opportunity for incremental orders is the Drone Dominance Program. Is that what we are hearing correctly from your comments?
I think that is correct. We are expecting additional Drone Dominance Programmatic awards. We are also working on some thermal weapon sight awards for their current customer base, as well as some more display awards from our current customer base, and heads-up displays for aircraft. So, we still see a fairly significant amount of opportunity for awards in Q4, and that will round out our backlog for 2027. We also expect to see awards for continued development on our Neural I/O chipset, and that chipset would be a monolithic design, so it would be a fairly significant endeavor for the company.
Thank you very much.
You bet, John.
The next question comes from Jaeson Schmidt with Lake Street. Please go ahead.
Hey, guys. Luke Fingerson on for Jaeson Schmidt here. Looking at product gross margin, came in at 27% this quarter, about 13% last quarter. Obviously, you said had some benefits this quarter. How should we be thinking about this going into Q4 and beyond?
Yeah, that will come back down to what we had seen for the prior normalized run rate. The 14% that you saw before was probably artificially low, but we will be in the low 20% on the production activity.
Okay. Thank you. Super helpful there. Yeah, that is it for me. Thank you.
The next question comes from Austin Moeller with Canaccord Genuity. Please go ahead.
Hey, guys. This is Eddie on for Austin. Thanks for taking the questions. My first question would be if you guys could clarify what the next steps are for Fabric.AI as they kind of work to develop hyperscaler relationships in its sales channel for Neural I/O interconnects?
I think that is a question for them. I think that is a question for them, actually. My interpretation and view is we are there to support them with their customer endeavors. We are working with them on our own customer endeavors, and they have hired a great firm called Kearney to help them with their go-to-market strategy. I would not want to speak to their customer lineup at this point.
Okay, got it. Thank you for that. My second question would be, have you guys had any conversations with the DOW about the executive order from last July that allows government data centers to essentially be built on military bases and federal land, and how you guys can kind of support that infrastructure?
Great question. We have. We have met with several four-letter agencies and three-letter agencies within the government to make them aware of the capability that Kopin has with our Neural I/O platform and how we can build that technology here in the U.S. A bit too early to talk about some of the programs that we are pursuing right now, but I can tell you that our Department of War and U.S. government agency opportunity pipeline for Neural I/O is well over several hundred million dollars.
Okay. Thank you so much.
Thank you.
The next question comes from Josh Sullivan with JonesTrading. Please go ahead.
Hey, good evening.
Hey, Josh.
What should we be preparing to see at CES in January and then at Photonics West? What kind of demo should we expect to, or deliverable, should we be thinking about that we might see?
Well, we have a few surprises, but I think you can expect to see a demonstration around bandwidth. We have already stated that we expect to show a pathway to at least 1.6 terabits. So that would be one. Two, showing the ability for us to transmit and receive data using this system, using a MicroLED. I can say that the demonstration will be limited due to the TIAs that we have to use because it is not a fully monolithic design, Josh, yet. But when it becomes a fully monolithic design, we will be able to show what we are capable of. But the rest, you will have to show up and see.
I hope to be there. Then I guess, what do you think the customer engagement step-up is going to be once that demo happens? I mean, you are obviously already engaged now, but as you get a real-world demo, are you anticipating a big step-up, or do you already think you have interested parties in the backlog right now?
Yeah, we already have interested parties. I can fully affirm that we are actively bidding on opportunities where companies that want either a CPO or an NPO type of application are coming to Kopin. Even some of our potential competitors, or co-competitors, if you will, companies that we thought might be competitors to us, would like us to work with them on building MicroLEDs. These companies have realized that we are very good at it. We are building them here in the U.S., and we have a tremendous amount of skill in this area. Most of the companies that we compete with are optical companies that are based in telecom. We are coming from the display side and have a view of how to create pixels that are very uniquely small and tightly compacted. So, we are very much engaged.
We have a very significant pipeline of opportunities already. We think after the demonstration and after we show folks that we are able to complete this technology in 2027 and 2028, that our pipeline will grow exponentially after that.
Got it. Maybe just one on the defense side. You have an international order here. We think about Theon, and we think about some of your other European or international relationships. How is that market coming together, and maybe what should we expect over the next year?
Yeah, great question. Seeing a much more involved European defense organization that is rallying around what is going on in the global environment. Said differently, we are seeing a significant amount of opportunity in Europe, mainly around MicroLED and OLED display technologies. There are several helmet-for-aviation opportunities that we are working on, several thermal weapon sight programs that we are seeing opportunities on, and our partnership with Theon is going well.
We are still working on DarkWAVE with them. We expect that we should have some customers lined up in 2027 for DarkWAVE through Europe, through Theon. Yeah, Europe is coming along quite well. If you recall in Q1, we also received several orders from new customers in Q1 as well. I think in terms of this year, and I am off the cuff here, we have got about $8.5 million of orders from European defense companies this year alone.
Great. Thank you for the time.
Thanks, Josh.
The next question comes from Christian Schwab with Craig-Hallum. Please go ahead.
Great. Thanks for taking my questions. I guess it wasn't clear to me the Westborough expansion, when additional operational capacity would come online. Is that mid-2027, which I thought I heard you say in the prepared comments, or is that more as-needed basis?
Sure. We're moving the executive team and finance teams out of 125 North Drive in early January, and we'll be adding around 10,000 to 15,000 sq ft of additional manufacturing capacity throughout 2027 to house mostly Drone Dominance volumes. The additional capacity would be mid-year 2027.
Great. That is a great segue to my second question. Given 12 prototype orders moving a little bit more slowly in first-person view, do you have a range of revenue outcomes or orders that you would anticipate in calendar 2027 from that program?
Yeah. The good news is we were very conservative, Christian, in our first round of revenue figures. So, we are sticking by the revenue for next year because our current customer base has already received their orders. Through Gauntlet 1 and 2, we still think revenues for next year in drone goggles will be in the range of around $10 million of revenue for us. The additional Gauntlet 2.5 was a surprise to us, and I think to many. However, it is going to slow us down a little bit, but we were very conservative originally, if you recall, in what we thought our order book and our revenue was going to be next year for drones. So, we are still in the game there with the current thinking.
Fantastic. My last question, you talked about production silicon and Neural I/O coming in 2028, I think I wrote down. Is that when we should expect first commercial revenue from Neural I/O, or is that more of a 2029 event, or too early to tell?
Our goal is 2028 for commercial revenue. We are looking to move to a monolithic IC design as soon as possible. The change that has happened, Christian, is both, I believe, Kopin and Fabric.AI, through our work with Kearney, have gained a tremendous amount of confidence in our capability to achieve the technical goals that we wanted to achieve with this design. I think starting slow and getting good before we go fully into a monolithic design was the right thing to do. But now that we have the confidence in the technology team and the approach, we want to move to a monolithic design as soon as possible, and enter this market with the best technology. I think the industry is going to be very interested and surprised with what we show them in CES and Photonics West.
We want to move to monolithic as soon as possible, and also achieve productization revenue in 2028 for both companies.
Fantastic. No other questions. Thank you.
Thanks, Christian.
The next question comes from Martin Yang with Oppenheimer. Please go ahead.
Hi. Thank you for taking my questions. First question on micro OLED. Once you start domestic production, is the goal to switch to 100% domestic sourcing, or is there If so, do you have a target date or year in mind when you achieve 100% domestic sourcing? Thank you.
We believe we'll have full domestic sources in 2028. We'll be fully depositing OLEDs in 2027, Martin. The difference being our digital backplane technology is still built outside of the U.S. We have activities And by the way, to be clear, not built in China. None of our OLEDs that we're shipping to U.S. companies are built in China, to be clear. But our new deposition system that we're expecting in Q1 of 2027 will produce the vast majority of OLED. The only caveat being is our digital ASIC. We may still manufacture that in Taiwan for a period of time until we can transition out of Taiwan to the U.S. for that digital ASIC. But that is the laggard of the system. That could happen in 2027, late 2027 to 2028.
Got it. Thank you. A question on profitability. I think the press release mentions you are progressing towards that goal. Is the goal still to achieve GAAP profitability in Q4? If not, then what's our target to achieve that profitability?
Yeah, Martin, that still is the plan right now. We do expect with the revenue projection that we have for Q4, and our spending profile right now, that GAAP profitability is achievable.
Do you think once you pass that threshold, do you think it is sustainable throughout 2027?
That is the model we are putting together right now, yes.
Thank you. That is it for me.
Thanks, Martin.
The next question comes from Alex Fuhrman with Lucid Capital Markets. Please go ahead.
Hey, guys. Thanks very much for taking my question. Congratulations on the success you have had here recently. Wanted to ask Michael about Neural I/O and some of the benefits that you are seeing from some of the early testing. Looking at the white paper you put out a month ago, it looks like you teased out some early findings on increases in throughput and how you could potentially be training, I think you said 23%-38% faster for the same workload. Crosses my mind a lot of smarter investors out there are really focused on looking for solutions that can help to reduce power and water use- in data centers for AI training.
I think especially after the recent election results, that is top on investors' mind. Can you talk a little bit about how some of those efficiencies you are seeing on training time and throughput, how that can translate into reduced resource consumption, and any traction you are seeing from that?
You bet. I will talk at a high level. Happy to follow up with a more technical detail review. At a high level, Kopin has spent a significant amount of time on Capitol Hill, both with the Senate and Congressional leaders on AI infrastructure, and the threat of China being able to penetrate our interconnects. Being able to have a backdoor into training modules and models through interconnects is a serious issue. That is number one. Number two, working with the Senate and the Senate Intelligence Committee focused around how we can bring AI closer to the battlefield will allow quicker decisions, save more lives, have more effective battle outcomes, is a critical area for the Department of War. That means you have to lower power consumption. You cannot drag around a generator with you to bring AI closer to the battlefield.
Being able to reduce power consumption and water use in data centers is critical for the Department of War. Using a MicroLED is not an area where we see any pushback from the Department of War. What they want to see is the robustness of the capability, the ability for it to transmit data above 1.6 TB, and a meantime between failures that is better than what they currently have. I am pleased to report that we are able to show all of those things, and we will be showing those things in January. One of the things that we have been tracking is using the next generation of GPUs has an inherent issue, and that issue is the interconnect between the GPUs as they are using large language models.
The interconnect in the middle is slowing them down, and that copper in the middle, this copper SerDes, is slowing the large language models down and creating a tremendous amount of heat, which creates power consumption, which also creates water consumption. You take the heat and the bandwidth issue out of the middle, and now you have a more effective system. That is what this MicroLED platform is going to do, and we will be demonstrating that in January at CES and Photonics West.
Okay, that is really helpful. Thank you for that explanation, Michael.
No problem. Great question.
This concludes the question-and-answer session. I would like to turn the conference back over to Michael Murray for any closing remarks. Please go ahead.
Thank you very much, operator, and thank you for listening today. We appreciate it, and we look forward to speaking with you again in our next call. Take care.
The conference is now concluded. Thank you for attending today's presentation. You may now disconnect.