Yesterday we announced another new product, so I thought I'd spend a little time today trying to explain what these new products are about, what they're doing for our markets and market opportunities. So, does the clicker work? Oh. Okay, so, obviously the Safe Harbor, I'm not going to read this. Everything I'm going to share with you is based on our best understanding of markets, competition, etc., subject to risk we spell out in the 10-K that we encourage you to read. So, is the clicker working? Okay. So, I'd like to start with maybe a review of 2024, and you can see, this is based on Q, the first three quarters' actuals, and the midpoint of our guidance in Q4 . Based on that information, or based on that data, revenue increased by about 20%. This is in the environment of roughly 3%-5% WFE growth.
You can see the flow-through down to earnings was quite strong, roughly 40% increase in our earnings over the prior year, and quite exciting for us because we've been driving such attention on getting control of the balance sheet, especially on inventory, which those of you who've been following us for several years have seen the inventories constantly going up. Now we're bringing those down, unlocking a significant amount of cash, and cash generation is expected to be around 25%, so very strong there. Looking, let's say the last four years, you can see WFE roughly around 3-ish%. Our growth, our CAGR over that same period is around 8-ish%, a little bit higher, up and down, decimal points there, and you get to about a 70% outperformance for WFE over the last four years.
But also those of you know that I don't usually talk about correlation to WFE because we're not highly correlated to WFE. What's really exciting for us and for our growth is the new products, and when we introduce new products and the outsize impact that can have on our business, given the size of our business, so last year, 2024, and mostly H2 , we announced four new products, and we did two acquisitions, so that's what I'm going to cover, so before I get into the products themselves, I wanted to talk a little bit about the markets that are driving our business over the next couple of years, and I think fundamentally one of the most important things to understand about Onto Innovation is the fact that we have this synergistic portfolio of products that apply to a broad range of the market overall.
What that means is that we can ride different waves of growth as they occur. A couple years back, several years back, 5G was the big wave of growth, and we were able to apply inspection, metrology, specialized metrology, all into that market to take advantage of it. Now it's AI, and we can see that in the AI packaging space, which we'll talk about in a little bit, both on the memory and on the 2.5D on the logic side. We see Gate- All- Around and the transition away from FinFET architectures to Gate- All- Around architectures, and of course Power Semiconductors. Each of these markets, we're serving with a multitude of products, so it outsizes our growth opportunities in those markets.
So, there's been a lot of focus and emphasis on AI packaging, and we narrowly define that as basically memory, high bandwidth memory, and CoWoS or 2.5D logic packaging specifically for AI use cases. What's really driving this adoption and the growth and the enablement of AI right now, it's the trends that I tried to show, we tried to show here in this slide. Traditional packaging has always been around packaging a die, and so there's a lot of advances. As the I/O counts went up, die sizes shrank, driving interconnect technology down, and, you know, package that die. So, copper pillars, etc., fan-out technology. The shift now, so that's continuing, it's actually accelerating. The shift now is around packaging multiple die, so heterogeneous packaging. So, not just as you can see from the 2D packaging picture there, where die are packaged and connected to an interposer.
Now you're seeing multiple die being packaged together, either in a 3D stack like HBM, or in 2.5D, or side by side, and then that entire thing is packaged. That increases complexity quite a bit, and you can see some metrics there, but it's also what's driving our opportunities. So, not only is the volume going up, but the complexity, the process control intensity is going up, and you can see here, we project the Dragonfly growth to be about 158% year over year, and we're seeing opportunities to bring in our films metrology into this space, and you can see that growth has been even stronger year over year, and you can see roughly the size. Obviously, Dragonfly is the bulk, but the films metrology is growing as well.
We expect that growth to continue, so as we look out over the next several years, this trend, this, these dynamics of heterogeneous packaging is going to continue. We estimate our current SAM to grow. We didn't spell it out there, but to grow looks like 10%-15% or so over the next few years. More importantly, because of the complexity, so that's the volume part, but because of the complexity of the process, there are new opportunities for us, new challenges customers need to solve, and we develop these products and technologies that you can see listed there to solve those, essentially adding another $300 million or so to our SAM, to our opportunity in AI packaging. Yeah, well, actually I will talk just briefly. So, the four products here you can see, this is specific to AI packaging.
We've talked about most of them before, so the subsurface technology, looking for chips and cracks buried into the wafer itself. The unpatterned technology, that's brand new, that's from the Lumina Instruments. We'll talk about that at the end of the material. And then there's been a lot of confusion around the two products at the bottom, the 3Di bump metrology and the EchoScan, the void detection metrology. I'm going to talk a little bit over the next two slides about why we're approaching the market from both perspectives, looking at both 3D bumps as well as hybrid bonding applications. So here's a roadmap, kind of illustrates what I said earlier about the constant drive for smaller, denser interconnects, and you can see that reflected in the well-published HBM roadmaps.
And there's something, and you can see the interconnect count per wafer is also increasing quite dramatically, which again adds to the complexity of the processes. But what's happening around eight to five microns is customers are looking at their roadmaps, and there's two dynamics happening. One, continuing customers are looking at continuing to extend the bump roadmaps. We're seeing bumps, on our 3DI technology, less than five micron in size. Current technology, you can see there HBM4, less than 10, right? So significant improvements in the bump technology that people are looking at, customers are looking at. In addition, of course, the well-publicized transition to hybrid bonding, which would eliminate bumps.
Now the challenges, we expect that both technologies are going to be in play for several years, and there's challenges with each that need to be solved for the customers to move their roadmaps further, and that's what these two tools are designed to do. So, going into the 3DI, the bump technology and the bump roadmaps, as these bumps were getting smaller and smaller and customers were looking out at their ability to measure the bumps, they started to get concerned that the incumbent technology would not have the legs and the ability to meet their roadmaps, and they wanted to see, you know, a different technology. Their concerns were around the use of white light for the measurement. The problem with white light is you have a wider beam, it's harder to focus in a very small spot, and it scatters.
You know, there's multiple wavelengths there, so different surfaces, different surface roughnesses cause the light to scatter. You can collect less of it on your detector. Our approach has always been to use laser-based technology, and the customers were aware of that, and they see that as an advantage for these smaller, denser bumps. We're able to focus a much smaller spot, we're able to reduce the noise from white light and light scattering, and get a stronger signal. As a result, we believe this tool will be faster, as well as more capable for much, much smaller bumps. So, for the roadmaps for the future, and you can see here we've made some progress.
We started delivering a system, I believe, in October-ish, and we've got deliveries to memory manufacturer, to a Tier 1 OSAT, as well as a customer who's actually manufacturing bump process equipment, so that they have it in their lab to develop next-generation bump technology. So, I think there's a nice broad appeal for the product and good progress so far. Now, EchoScan, this is actually more of a revolutionary technology. So, today, and I'll explain what's happening today. Customers right now, if they bond wafers together, they're concerned about voids in those, in those bonds. Right now, the way they measure those voids is they take a stack of wafers, they submerge them in a bath of water, they take some sound waves, run it through that, look for voids to come back. Essentially, it's a 30-minute process per stack, and they can only see 10-micron voids.
When they pull that out, they then have to take it to a spin rinse dry and hope that those 300 million interconnects, they're all clean, the water's not stuck in there, no particles got in there, no delamination occurs. There's a lot of concern around yield with this. What customers want is, A, 10 microns on a 5-micron pitch, useless, right? So, they need to see 1-micron defects. That's, that's table stakes. In addition, they want to eliminate the water, the source of yield loss. So, they want this done in atmosphere. So, leveraging some of the acoustic technology that we've developed over years and some acquisitions we've done, we're now releasing the first acoustic scanning tool. Basically, it's an inspection tool that can measure 1-micron voids across a wafer in atmosphere, hopefully matching the same throughput.
So, we're going to be delivering the first tool this quarter to a customer that we've been working with for several, several quarters now. We've done a lot of demos for multiple customers. We expect other customers to follow. And you can see the SAM here. I think it's somewhat conservative, but it really depends on the sampling rates. So, a customer's going to have to measure every single wafer that's stacked, or every single wafer that's bonded, or only do it at the end. That's something they're going to have to figure out, and that would change the sampling rate, and maybe, you know, if they're measuring every one, that would increase the SAM. Okay. So, that's all on the packaging side, and so there's a lot of, you know, excitement there, of course.
What we're seeing is that, at least when it comes to HBM, the amount of growth we're seeing in HBM and the fact that the HBM wafers actually take up more of the capacity for the DRAM manufacturing, you can see here it's projected to be over 10% of DRAM, worldwide DRAM capacity, being taken up by the growth in HBM. Take that, combine it with the slow recovery of the bit demand growth. Remember, it all fell off a cliff several years ago, slowly recovering HBM. What we're seeing now is that, as I mentioned in my last earnings call, a recovery in DRAM spending, and I think that was met with some skepticism, but as of yesterday, we've got a $69 million VPA, most of it is orders, and, that's all to support expansion for DDR5.
What's really interesting here about this, though, is it reflects an expansion of the optical metrology ecosystem that we offer, so when we look at this order, it includes the OCD, where we've always been very strong, but it also includes an expansion of our integrated metrology, so broader adoption of our integrated metrology, as well as our films metrology, and that story is similar in the Gate-All-Around node, so we've always talked about Gate-All-Around being a very strong node for, or transition for, Onto Innovation. Especially in the OCD, we saw an increase in capital intensity, and we always said, you know, it'll depend on how customers make the final decisions as they're really ramping the HBM versus what they're doing in pilot and R&D, and we're starting to see that play out.
So, if we normalize the FinFET, a FinFET line with a Gate-All-Around line at, let's say, 100,000 wafer starts per month, we see a 20% increase in capital intensity for OCD, and that's being reflected in the data we're seeing right now. In addition, and perhaps a bigger surprise, is the strong adoption of the films business, and there you can see nearly 300% increase over FinFET. Obviously, we had very little films business several years ago, but very strong adoption, and those bars, those two charts are normalized, so they're on the same scale, so you can see that our films business is actually starting to grow and catch up. I mean, it's not there yet, but it's more than half, at least for Gate-All-Around, the OCD business, so very, very strong momentum there, and so we talked about films a couple of times.
We talked about growing adoption in the advanced packaging space. We talked about growing adoption now in the DRAM and in the Gate-All-Around, so in advanced node space. I thought it would be important to remind everyone about that optical metrology ecosystem. About 2021, we announced a new metrology suite with more advanced OCD capability, pushing the limits of OCD down below the 3nm node, where everyone thought X-ray was going to have to take over, and a new integrated metrology tool, the AI Impulse 5, as well as our first introduction of the Iris platform.
If you recall, there was a lot of, that was a focus on common films, and our intent was to first see if customers would adopt the IRIS films, and then, because they were pushing us very hard to go after critical films as well, we said, first, show us you're committed, show us the adoption on the films platform, then we'll invest and expand. And so, I'm very, very happy to announce that now we have launched, as of, you know, yesterday's press release, the Iris G2, which is a single platform that will cover both common and critical films, now opening up the entire films SAM to us, which is about $1 billion, and, we'll be delivering that first tool in the first quarter. So, I've talked a lot about the films business, and I haven't really quantified it.
So, here you can see, year over year, the films adoption essentially doubled. So, momentum's been building in multiple markets, including Power Semi, to the point where now it's about a $100 million business, estimated in 2024. We're also qualified now at four of the top five semiconductor manufacturers with films. Hence the momentum and hence the investment in the IRIS G2, also as a result of demand from the customers. And again, what they really wanted was a single platform to unify all their films, products, and strategy. We believe this new tool will be roughly 30% faster than the current tools, than the incumbent technology, matching the precision. So, a significant 3x improved precision over our IRIS platform, which will match performance of the current incumbents, but be much faster. So, better cost of ownership. Again, we're shipping this tool in Q1 .
I know I'm hitting you with a lot of information, but hopefully it's adding clarity and not confusion. So, now shifting of gears yet again, Power Semiconductors. So, this is a market that everyone, if I, and actually I should have looked, all the faces, the eyes will roll, Power Semi, it's dying, what are you talking about? 2023, we said that we had a record year. We said 2024, we'd be kind of flattish to that record year. And you can see from the tagline here that actually 2024 was a record year, a fairly strong one with a very good surge in Q4 , about, it's double digits. So, 10% plus increase over 2023. So, it's not a volume thing, right? Because you're not seeing a lot of customers expanding their volumes, expanding their lines, but they're very focused on driving yield improvement.
They want to drive more profitability out of their lines, so when they do get orders, they do start to grow and ramp, they'll make a lot more profit, invest in more capacity, and it'll be a nice self-fulfilling cycle there. So, you can see, because we also get a lot of questions about, is it all silicon carbide? I mean, where's your focus? And it's really across the board, Power Semiconductors, across the board. And you can see there the silicon carbide, GaN, and silicon power. Our growth rate about 60% overall over the last four years. That market has grown around 8%. Of course, we started again from a very small base, but it might be interesting to look at the mix. Our mix is still pretty reflective of the overall Power Semiconductor mix. Probably a little bit stronger in silicon carbide, but across the board, largely matching.
What we see moving forward is actually an acceleration here. I know there's a lot of concerns around EVs and the different applications for Power Semi, but in my view, this is a trend that is not going to reverse. There's going to be, you know, waves of adoption. We had one, there's less now. I think new battery technology is going to help make another wave of adoption for EVs. I think there's going to be innovations, just like in semiconductors, that drives this market forward. But the move towards electrification of everything, I think is, that's a trend that's here. We're very happy that our portfolio of products is being widely adopted by these customers and adding to our growth opportunities. We talked about the portfolio, we talked about products enhancing our growth opportunities.
The acquisition of Lumina Instruments, or Lumina Technologies, was a very exciting one. Originally focused on compound semi market, it's a technology that brings in scatterometry, so a capability for inspection that we don't have currently in our portfolio. That enhances our ability to solve other applications in unpatterned wafer inspection, and even in packaging for glass panel, which we haven't talked a lot about today, but certainly something we're focused on from the stepper perspective, but we also see from a CoWoS perspective the opportunity to use this technology for glass carrier inspection. So glass carriers are used for some of this 2.5D, combining the different die, then flipping, driving interconnects, the die are all removed, the glass is cleaned and then reused. But if there's any residue, that's going to cause a yield issue.
If there's any deformities or defects, that's a yield issue, so having this type of inspection, which is unique with the PrimaScan, that can use multiple lasers to inspect both the top surface, the inclusion defects, so embedded defects, and the bottom surface simultaneously is a huge value for our customers, so everyone who I've talked to about this product line, very excited. They actually come up with new ideas for how they could apply it for crystalline defects, some other materials analysis that we didn't even realize, so I think this is going to be a pretty good growth opportunity for us. You see the SAM here is expected to be about $250 million over the next couple of years, and you can see also that compared to the incumbent technologies, current technologies will be about 30% more sensitive and 50% faster, so very, very compelling value proposition. Okay.
You're going to miss the why invest. In conclusion, yeah, right. Why invest in Onto Innovation? I covered a lot of different materials. Some of it's been talked about, but maybe not fully explained to people. I think really one of the big differentiations that we have as a business and as a business model is this broad portfolio, this synergistic portfolio that works well together, enhances the value by working together, and how it applies across the semiconductor value chain from bare wafer manufacturing all the way out to advanced packaging. I think another big thing is the, because of that, the positions we have are very strong market growth drivers right now. We mentioned several of them. Gate-All-Around, AI packaging, DRAM we talked about, that's going to drive some growth as well as Power Semiconductors. Really good, good position there.
In addition, without just waiting for, and that's part of that business model, without just waiting for the tide to rise as if we're just tied to Power Semi, when's power going to come back, then we'll explode. We have the ability to combine the different technologies, work with our customers, and add to our growth by adding to our SAM and expanding our SAM. And you can see that roughly about a billion SAM is added through those new products between the IRIS and the AI packaging, about a billion, which is roughly 30% increase in our total market opportunity. So, fairly good growth there. Underpinning this is a strong and improving financial model, which when you flow it all down is generating cash that of course fuels additional M&A.
And for a company of our size, you know, $100- 200 million starts to add up quite, quite quickly. So, when we talk about another $1 billion in SAM and we think about, you know, even just getting 30% of that market, that adds another $300 million to our business. When we look at nearly $800 million in cash and the different acquisitions we can do, that adds another $100-150 million in potential revenue opportunities. So, it starts to really grow the business and it's mostly in our control. And I guess to leave you with, you know, since AI was the big opening story, here I asked my IR team, Mark Sgriccia team, to come up with a Gen AI image for semiconductors and waves of growth.
I think there's still some work to be done, so they're going to need some, they're going to need some more processing power here. But, all right, with that, Charles, we can take questions if there are any.
All right. Thanks, Mike, for the, for the presentation. Didn't expect that we're going to get such a comprehensive product update, market update from Onto. I was thinking probably more of a fireside, but, but you know, maybe I'll start with a couple of questions then, then open it up to the audience since we have some time. So, Mike, I think the first question, right? One of the things about packaging, you guys have been very unique within the process control equipment space, strong presence in packaging.
Ever since second half 2023, I believe everybody started to notice, right, that your two peers who traditionally focused on the front end start to talk increasingly more and more about packaging. I think there's no better, I mean, statement about your strong position when it comes from your competitors who are not there, but they want to get in there. Yeah, so definitely I see there are a couple of things, right? You're already strong with the 2.5D CoWoS. You have, you have a product footprint in hybrid bonding. But I think there's one other thing you touch upon a little bit as the panel-level packaging. Going back a few years, the panel was considered as, okay, probably it's too expensive, and by the way, where do you get a volume? Because panel is big, supposedly, right?
But now AI has changed the things, and you see that the GPU chips are getting bigger and bigger, right? And that's now on TSMC's roadmap. So, would you mind talking a little bit more about panel and what products do you have there and what's the timing of the ramp you're expecting as of today?
Good question. So, I think for a panel, it's not just the die that's increasing. It's what I talked about in one of the earlier slides. It's the package size is increasing. So, the combination of bringing multiple die together and packaging them together, that's becoming much, much larger. And that roadmap is going to continue. So, trying to put, let's say 240 by 240mm and lay that out cost-effectively on a 300mm wafer, you know, not going to happen. It's too expensive.
You're going to waste too much real estate, you know, circle and square, not fit well together. But going to panel, large square surface, you can put these larger packages together, process them more effectively, and, you know, basically save cost. So, the transition to panel I think is clear. The challenge has been, like you brought up AI, several of those, interconnects are smaller than what existing panel technologies can really handle or handle in the future. So, the transition to glass and panel I think is going to be an important combination, allowing for a much more stable substrate that allows for higher resolution printing, better, smaller interconnects. I think that's another big transition. What we do in this space, we have a very unique stepper. We've talked about that over the years. So, we have very high resolution, wide field optics.
Most of the other technologies are going a different approach using direct imaging. There's challenges with direct imaging. We can go into that at a different time. So, we have the unique stepper capability, but then we augment that with our Firefly. And recently our Firefly was upgraded, the G3 was upgraded with now the Dragonfly optics. So, historically they had different optical paths. Now they're the same. So, all the advances we're doing on Dragonfly, which are very, very fast, we're being able to apply to the Firefly for free. So, that's very powerful. And now the Firefly supports all of the metrology options that the Dragonfly had, which turns out to be more important to our customers today than the inspection was, which surprised us. So, their struggles are less on defect issues, but more on metrology, the actual printing, the quality of the process.
So, that's the two products. Then the Lumina Instruments. So, the PrimaScan product line is going to help. One of the big concerns with the transition to glass is the glass quality, quality of glass, any inclusion defects. You're going to process it all. And if you miss any of these defects, you still have yield issues. PrimaScan looks like it's going to be able to solve a lot of those process control concerns around glass and the glass substrate. So, kind of incoming inspection. So, all three of those areas. And we just published a paper around our StepSmart technology. So, leveraging our software to do feed forward, feed back from the stepper and the Firefly to improve the overall litho process. So, I think that combination of products is again part of our value proposition, part of our strategy and going to open up the door.
So, you also asked about when will panel ramp? That's. I would say, we've already delivered one kind of hybrid tool for both glass and advanced IC substrates. That customer, if they stick to their plans, has a very aggressive plan to move to glass. We have another customer who presented at our PACE Applicate Grand Opening, who is already engaged in glass panel processing, trying to ramp into volume and has plans to do that, you know, get customers qualified. He's qualifying now over the next year to two years, two years max, probably next year. So, I think we're right at that cusp. If I had to bet, I would say it's more of a 2026-2027 opportunity, but we'll see.
So, we have maybe four more minutes and maybe I just open it up to the audience. Yes, please.
So, right now the litho tool is being applied to advanced IC substrates. And so the big market for that is enterprise servers. And they're kind of in the toilet. So, they expanded quite aggressively. Actually your old company was in the short term, yes. We definitely, we still have a backlog of steppers. Customers have been moving those around because they have a lot of our steppers idle. But I do think the transition to glass is going to be an important one. And we're already there with the tool that we've shipped and has actually signed off already. So, I think we're well positioned there. I think that transition will begin R&D orders in 2025, 2026 pilot. And I think that'll be good. When enterprise servers come back, that's, that remains to be seen. That's relatively low.
I think we're 10%-15% now, have been since half our customers got put on an entity list. So, we've felt that pain already.
Any other questions? Yes.
On the last call, you spoke about HBM.
So, I think what I said was we don't see the match between the CoWoS expansion and the HBM expansion. And our expectation was because Samsung had added so much capacity and wasn't qualified that NVIDIA was essentially trying very hard to qualify Samsung for obvious reasons. Pricing benefits don't leave all that. And what I said was whenever that decision is made, if it's made for Samsung, you know, probably no big increase because they have a lot. If it's not, then the other two are going to scramble to try and pick up.
Now, a couple of weeks ago, maybe last week at CES, I think Jensen Huang kind of publicly said, "Samsung not qualified." So, that's now in stone, and yes, we're seeing the others scramble like I thought. Is it going to be, you know, a big, huge, massive amount? Well, we don't know yet, but we are seeing that dynamic play through. We are seeing, you know, the other two quickly trying to get some tools, but the magnitude of which we're going to have to see. I showed in the slides; we think that the AI packaging is going to continue the growth, or at least the demand is going to continue for the next couple of years. So, that's our view based on how rapidly customers are expanding and not sitting on idle capacity. They're having to expand yet again and yet again, yet again.
So, I think, you know, for us, 2025, 2026 should still be good growth years, but quarter to quarter and competitive dynamics, all that is, you know, adds some variability.
All right. We're probably at the end.