Operator
Ladies and gentlemen, thank you for standing by. [Operator Instructions] I would now like to turn the conference over to Dan O'Neil. Please go ahead, sir.
Daniel O'Neil
Good afternoon. Thank you for joining our earnings call for the fourth quarter of fiscal 2025. Today, I'm joined by Bill Brennan, Credo's Chief Executive Officer; and Daniel Fleming, our Chief Financial Officer.
During this call, we will make certain forward-looking statements. These forward-looking statements are subject to risks and uncertainties discussed in detail in our documents filed with the SEC, which can be found in the Investor Relations section of the company's website. It is not possible for the company's management to predict all risks nor can the company assess the impact of all factors on its business or the extent to which any factor or combination of factors may cause actual results to differ materially from those contained in any forward-looking statement. Given these risks, uncertainties and assumptions, the forward-looking events discussed during this call may not occur, and actual results could differ materially and adversely from those anticipated or implied. The company undertakes no obligation to publicly update forward-looking statements for any reason after the date of this call to conform these statements to actual results or to changes in the company's expectations except as required by law.
Also during this call, we will refer to certain non-GAAP financial measures, which we consider to be important measures of the company's performance. These non-GAAP financial measures are provided in addition to and not as a substitute for or superior to financial performance prepared in accordance with U.S. GAAP. A discussion of why we use non-GAAP financial measures and reconciliations between our GAAP and non-GAAP financial measures is available in the earnings release we issued today, which can be assessed using the Investor Relations portion of our website.
With that, I will turn the call over to our CEO. Bill?
William Brennan
Thanks, Dan. Thank you for joining our earnings call for the fourth quarter of fiscal '25. I'll begin with a review of our results, and then I'll provide highlights for our outlook into fiscal '26. Dan Fleming, our CFO, will follow with a detailed discussion of our Q4 and fiscal year '25 results and then provide our outlook for the first quarter.
In the fourth quarter, we delivered revenue of $170 million, a 26% sequential increase and up 180% year-over-year. Our non-GAAP gross margin was 67.4%. For fiscal '25, Credo achieved revenue of $437 million or growth of 126% year-over-year. Our non-GAAP gross margin from fiscal '25 was 65%.
I'm proud of Credo's achievements in fiscal '25. Record-breaking revenue and profitability were fueled by surging demand for our innovative, reliable and energy-efficient, high-performance connectivity solutions. Our long-term commitment to customer-driven innovation paid off significantly in fiscal '25. Quarterly revenue nearly tripled from Q1 to Q4, validating our foresight and our ability to capitalize on a predicted inflection point. Our agile approach strengthened partnerships with hyperscalers amid a rapidly evolving AI landscape.
As a pure-play high-speed connectivity leader, Credo delivers a growing portfolio of differentiated solutions for global data center operators, currently supporting port speeds from 100 gigabits per second to 1.6 terabits per second. Our innovation spans 3 tiers, advanced series technology, cutting-edge integrated circuit design and comprehensive system-level solutions. These innovations are seamlessly integrated with our PILOT software platform.
PILOT is an acronym for Predictive Integrity, Link Optimization and Telemetry. PILOT offers an industry-leading user interface, robust debugging tools and advanced telemetry tailored for large-scale deployments. This holistic innovation strategy enables Credo to deliver copper and optical connectivity solutions that surpass industry standards, providing unmatched functionality, reliability and energy efficiency.
While achieving a remarkable revenue ramp in fiscal '25, we continued to build the foundation for sustained growth. Looking ahead, we anticipate increasing customer diversification across copper and optical connectivity for Ethernet, PCIe, UALink and other emerging applications in both scale-out and scale-up AI networks. Following the significant revenue inflection in fiscal '25, we're energized by the expanding opportunities and total addressable market that lie ahead.
I'll now review our business in more detail. Regarding our Active Electrical Cable product line in Q4, our AEC revenue maintained a steep growth trajectory. As anticipated, our customer base diversified with 3 hyperscalers each contributing over 10% of our revenue, strengthening our market position.
When we pioneered the AEC market years ago, we recognized the compelling advantages over both traditional Direct Attach Cables or DACs and laser-based optical solutions, especially at data rates of 50 gig per lane or higher. AECs have extended the viability of copper connectivity, becoming the de facto standard for intra-rack applications. Compared to DACs, AECs deliver superior signal integrity, advanced features and a more versatile form factor.
Now AECs are gaining traction as a robust rack-to-rack solution for distances up to 7 meters, offering over 100x greater reliability than laser-based optical modules, virtually eliminating link flaps and significantly improving energy efficiency, which are both key enablers for best-in-class AI deployments.
Credo's system-level approach has driven substantial competitive advantages by owning the entire solution stack, SerDes IP, retimer ICs, system-level design, qualification and production. Our approach positions Credo for significant innovation and time to market advantages. As data center architecture has evolved rapidly, we foresee a continued shift towards curated system-level solutions.
We're enthusiastic about the ongoing expansion of the AEC market. For instance, our recently demonstrated PCIe Gen6 AECs at GTC promise the same compelling benefits for AI scale-up networks as deployments transition to rack-scale architectures. Our growing traction with hyperscalers is evident. With strong customer forecasts and new design wins and qualification, we're confident in sustained AEC revenue growth.
I'll now discuss our progress in the optical market. Fiscal '25 was a standout year for Credo's optical business. We closed the year with strong momentum, expanding customer diversity across lane rates, port speeds and applications. We achieved our revenue growth targets, delivering 50 gig and 100 gig per lane optical DSPs to a broad base of optical module customers. In Q4, we secured a significant DSP win for an 800-gig transceiver with initial deployments expected at a U.S. hyperscaler in fiscal '26.
At the OFC conference in San Francisco last month, Credo's latest optical innovations drew widespread attention from industry leaders. We unveiled our ultra-low power 100 gig per lane optical DSPs built on 5-nanometer technology. This family, including full DSP and Linear Receive Optics or LRO variance, sets new industry benchmarks for power efficiency.
In collaboration with an optical module partner, Credo demonstrated an industry-first 800-gig optical module, the total power consumption of roughly 9 watts powered by our Lark LRO DSP and single-mode optics. We achieved error rates comparable to full DSP solutions, earning significant interest from hyperscalers prioritizing power efficiency for AI deployments.
We also showcased our 3-nanometer 200 gig per lane optical DSP, supporting port speeds up to 1.6 terabits per second. With leading signal integrity and power efficiency, this solution positions Credo to drive the industry's transition to 200-gig lane speeds in the coming years. Looking ahead, we see a dynamic and growing market for optical connectivity, where reliability and energy efficiency are increasingly critical.
Credo is poised to deliver system-level innovations that provides substantial advantages to our partners. Credo's optical business demonstrated robust execution in fiscal '25. With our growing differentiation and an expanding system-level market opportunity, we anticipate accelerated revenue growth in the years ahead.
Turning to our retimer business. In Q4 and fiscal '25, our retimer business delivered robust results, reinforcing our leadership. Retimer revenue growth was fueled by our 50 gig and 100 gig per lane Ethernet solutions, offering advanced features like MACsec encryption, gearboxing and software-enabled functionality tailored to diverse customer requirements. Our customer base now includes leading AI server vendors alongside traditional switching clients, reflecting the growing adoption of our solutions in AI-driven architectures.
For fiscal '26, we anticipate strong growth driven by the continued shift to 100 gig per lane solutions and increasing demand for system-level expertise and software capabilities to address hyperscalers' complex AI-optimized architectures. Our recently launched PCIe Gen6 retimer family, led by the Toucan retimer, achieved full compliance at the PCI-SIG Workshop, showcasing superior performance and interoperability. Demonstrations at GTC, OFC and most recently at COMPUTEX with 2 leading ODM partners further validated our capabilities.
Customer momentum for our PCIe retimers is accelerating, positioning Credo to secure design wins in calendar '25 with production revenue expected in calendar '26. Our competitive edge lies in leveraging core SerDes technology, a customer-centric approach and system-level innovation to deliver differentiated latency, reach and power efficiency. Additionally, our PILOT development telemetry software platform drives faster time to market, improved yields and enhanced system monitoring, providing clear advantages based on market feedback.
In summary, fiscal '25 marked a pivotal year for Credo, achieving record revenue, profit and market adoption of our innovative connectivity solutions, hitting the inflection point we anticipated. Our operational and customer-facing teams executed flawlessly to deliver these results.
Credo pioneered a market that transformed how hyperscalers connect switches and servers. We continue to innovate with recent product announcements reflecting customer-driven solutions. These advancements position us to capture significant opportunities in the global AI infrastructure investment wave, fueling our next phase of growth.
Reflecting on our journey, Credo has navigated successes and challenges with relentless focus on delivering world-class products. This resilience defines our DNA and is our greatest strength. Thank you, Team Credo, for your dedication. I'm excited for what lies ahead. We see growing demand for high-speed connectivity solutions across our hyperscaler customers to power advanced AI services. a trend we anticipate continuing for the foreseeable future.
Customers require reliable, power-efficient, high-performance and tailored solutions to support their diverse architectures. Credo meets this demand with a differentiated portfolio of copper and optical connectivity solutions customized for customers, built on our core SerDes IP, tiered innovation strategy and system-level approach.
With that, Dan Fleming, our CFO, will now provide additional details, and we'll then take questions.
Daniel Fleming
Thank you, Bill, and good afternoon. I will first provide a financial summary of our fiscal year '25, then review our Q4 results, and finally, discuss our outlook for Q1 and provide some color on our expectations for fiscal year '26.
Revenue for fiscal year '25 was a record at $436.8 million, up 126% year-over-year, driven by product revenue that grew by 157%. Gross margin for the year was 65%, up 257 basis points year-over-year. Our operating margin improved by 2,500 basis points as we continue to generate considerable top line leverage, driven by growth in our products while growing operating expenses considerably slower than revenue. That step-up in profitability flowed through to the bottom line as we reported earnings per share of $0.70 for the year, a $0.62 improvement over the prior year. In fiscal year '25, Credo not only delivered the dramatic product growth, which we had forecast, but we also demonstrated the earnings power in our business model.
Moving on to the fourth quarter. In Q4, we reported revenue of $170 million, up 26% sequentially and up 180% year-over-year and well above the high end of our guidance range. Our product business generated $165.9 million of revenue in Q4, up 26% sequentially and up 276% year-over-year. Notably, our AEC product line, again, grew healthy double digits sequentially to achieve new record revenue levels once again.
Our top 3 end customers were each greater than 10% of revenue in Q4. As a reminder, customer mix will vary from quarter-to-quarter, and we continue to make progress in diversifying our customer base. We continue to expect that 3 to 4 customers will be greater than 10% of revenue in the coming quarters and fiscal year as hyperscale customers continue to ramp more significant volumes and as we expect to begin to ramp 2 new hyperscale customers in the second half of fiscal year '26.
Our team delivered Q4 non-GAAP gross margin of 67.4%, above the high end of our guidance range and up 355 basis points sequentially. Our product non-GAAP gross margin was 66.5% in the quarter, up 354 basis points sequentially and up 1,289 basis points year-over-year, primarily due to increasing scale.
Total non-GAAP operating expenses in the fourth quarter were $52 million, at the high end of our guidance range and up 19% sequentially, primarily driven by head count. Our non-GAAP operating income was $62.5 million in Q4 compared to non-GAAP operating income of $42.4 million in Q3, up demonstrably due to the leverage obtained by achieving 26% sequential top line growth.
Our non-GAAP operating margin was 36.8% in the quarter compared to a non-GAAP operating margin of 31.4% in the prior quarter, a sequential increase of 538 basis points. Our non-GAAP net income was $65.3 million in the quarter, a record high compared to non-GAAP net income of $45.4 million in Q3. And our non-GAAP net margin was 38.4% in the quarter, well above the high end of our long-term net margin model of 28% to 33%.
Cash flow from operations in the fourth quarter was $57.8 million, up $53.6 million sequentially due to cash collection driven by the significant sequential product ramp. CapEx was $3.7 million in the quarter, driven largely by purchases of production equipment. And free cash flow was $54.2 million, an improvement of $54.6 million from the third quarter. We ended the quarter with cash and equivalents of $431.3 million, an increase of $52.1 million from the third quarter. We remain well capitalized to continue investing in our growth opportunities while maintaining a substantial cash buffer. Our Q4 ending inventory was $90 million, up $36.8 million sequentially.
Now turning to our guidance. We currently expect revenue in Q1 of fiscal '26 to be between $185 million and $195 million, up 12% sequentially at the midpoint. We expect Q1 non-GAAP gross margin to be within a range of 64% to 66%. We expect Q1 non-GAAP operating expenses to be between $54 million and $56 million. We expect Q1 diluted weighted average share count to be approximately 188 million shares. These expectations are based on the current tariff regime, which remains fluid.
We were pleased to see fiscal year '25 play out as we expected. The rapid shift to AI workloads continue to drive new broad-based customer engagement, and we executed well to deliver the sequential growth we had forecast throughout the year.
As we begin fiscal year '26, we expect revenue to exceed $800 million for year-over-year growth in excess of 85%. We expect non-GAAP operating expenses to grow at less than half the rate of revenue from fiscal year '25 to fiscal year '26. And as a result, we expect our non-GAAP net margin to approach 40%.
And with that, I will open it up for questions.
Operator
[Operator Instructions] Your first question comes from the line of Vivek Arya with Bank of America.
Vivek Arya
For the first one, on the revenue side, I was hoping you could perhaps quantify how large were the 3 10% customers, especially the largest one. And thanks for giving the fiscal '26 revenue outlook. That suggests kind of a modest kind of sequential growth through the year, but I think you mentioned that you expect other new hyperscalers to come on board in the second half. So if you could just talk through what the puts and takes are as you look at kind of shaping the year. What could be the upside drivers and downside risks from here through fiscal '26 as you get more customers onboard? And then I had a follow-up.
Daniel Fleming
Sure, Vivek. This is Daniel Fleming. So as you know, last quarter in our earnings call, we said that we had expected 3 to 4 10% end customers in the coming quarters and fiscal year. So we reiterated that again in our prepared comments, and that's exactly what we saw in our Q4. So the largest customer was 61% of revenue. No surprise who that was. And we also had a 12% and an 11% customer. And they were the same as -- the same customers that we saw in Q2 past that 10% threshold.
So kind of addressing your second point, we expect to continue our diversification throughout fiscal '26. And in addition to these 3 customers, we do expect to have 2 additional hyperscalers in the second half of fiscal '26 in addition to those three 10% customers. It's a little difficult for us to map out exactly how that will chart through the year. But right now, if you just kind of apply a linear projection, that's probably as best as we could project at this point in time. But we're certainly excited about our continued revenue diversification across several hyperscalers as our innovative solutions are more broadly adopted across the industry.
Vivek Arya
All right. And for my follow-up on gross margin, you mentioned scale as a reason for gross margin expansion. But I think for Q1, you're guiding gross margin to kind of get back to trend, 65%, but you should be getting more scale benefit. So is there anything else in mix or customers that [ are there ]? So just conceptually, how should we think about the gross margin trajectory for the fiscal year? And for extra credit, if you could also tell us about how you think about EBIT margin because there, you are above your longer-term model.
Daniel Fleming
Yes, we're certainly seeing the benefit of scale. We saw that in Q3. We saw it even more so in Q4 as that scale continued to grow. So it's a fair question that you asked. We were up 355 basis points from Q3 to Q4 gross margin-wise. And that puts us above the high end of our long-term model. But having said all that, our gross margin expansion won't always be linear as we continue to increase scale. And there will always be some differences in product mix from quarter-to-quarter. So we guided Q1 at 65% at the midpoint. But more importantly, we're not setting a new long-term model for gross margin, but we're clearly entering a phase right now where gross margin will be at or above the high end of that long-term expectation.
Vivek Arya
And anything on EBIT, Dan? They're also above target.
Daniel Fleming
Kind of the same. You would expect that to follow. Kind of the improvement or expansion in net margin should fall right through to EBIT. So you could do that math pretty -- in a pretty straightforward fashion. Our CapEx, the one additional piece I'll give you is if you look at all of the different tape-outs and leading nodes that we have planned over the coming fiscal year, our CapEx might be maybe double what it was this last year. So that would be the last piece of the equation you would need to come up with your own EBIT.
Operator
[Operator Instructions] Your next question comes from the line of Tore Svanberg with Stifel.
Tore Svanberg
Bill, Dan and Dan, awesome results, great execution. Bill, I was hoping you could talk a little bit more about some of the use cases with your large customers that are ramping, especially on the AEC side. So there's clearly an 800-gig upgrade cycle going on. I guess you're still shipping some 400 gig. And especially as we think about the 2 new customers that are coming out -- coming on second half of the year, if you could talk to the use cases for those as well, that would be really helpful.
William Brennan
Sure. So I think we see the use cases being pretty consistent with what we've talked about in the past. There's 3 basic areas. There's -- and our main business today is in connecting servers with switches. The first business that we ramped was really front-end connections. And that is what we think about traditionally when we think of the network.
Second area is now with back-end networks. And specifically with scale-up networks, this is the majority of the product that we're shipping today. Really, if we look at all of the customers combined, I would say it's far greater than 50% of the shipments that we're seeing scale-out back-end networks for AI.
The third area is really within disaggregated chassis. And so we see that being a growing part of our business, and that exists in both the back-end and front-end networks. What we're seeing is our largest volume right now is actually 50 gig per lane AECs, but we see the trend moving quickly to 100 gig per lane. Hopefully, that gives you the color you were looking for.
Tore Svanberg
Yes. No, that's great. And maybe as a follow-up and sort of related to that, because you mentioned your largest volume is still 50 gig, I assume that means it's more 400. But where are we sort of then in the inflection point to 800 gig? Is this sort of still very early days for 800-gig upgrades? And I assume it will be very different by customer, right? Some customers probably haven't even moved to 800 gig for your AEC products there, right?
William Brennan
Right. Yes, each one of our customers, we kind of view as a different market in and of itself. They all have different strategies, and there's many different ways to achieve the networking objectives as it relates to AI clusters, as it relates also to the network in general. And so we see that really towards the end of fiscal '26, we see it -- the transition for 50 gig to 100 gig overall broadly for our business will really start to happen in a bigger way. But again, we've got some customers who are already there. 100% of the shipments that we're making are 100 gig per lane. So it's hard to talk generally about it given the fact that each one of our customers will have a different strategy.
Operator
Your next question comes from the line of Quinn Bolton with Needham and Company.
Quinn Bolton
Guys, let me offer my congratulations as well. I wanted to come back to Dan to the gross margin and the step-down in the July quarter. Wondering if you could just talk to us about the tariff risk. I think you've got BizLink and Foxlink, which manufacture your AEC cables, both located in China and wondering if tariff risks and tariff costs are having any impact on that gross margin in the July quarter. And then I've got a follow-up.
Daniel Fleming
Yes, that's not -- we don't expect there to be significant tariff risk-impacted gross margin percentage at this point. So that's not what's driving potentially a minor reduction in gross margin percentage Q1 to Q2 -- or Q4 to Q1. But just thinking of tariffs in general, maybe I'll let Bill comment on that a little more broadly.
William Brennan
Sure. So I think what we've seen over the last quarter since our last call, tariffs and the overall macro economy are continuing to evolve. We're obviously monitoring the situation closely, and we're working very closely with our customers. And ultimately, we're trying to be as flexible as we can in delivering the best solution for each customer.
Over the past year, we've talked about the efforts that we're making to diversify geographically. Really happy to say that in the kind of worst-case scenario, within months, we could be out of one geographic location and into another. And so I feel like our team and our customers, we're trying to take a mindset that's dynamic. And ultimately, we feel like this thing is going to be more well understood over the next 3 to 6 months. And we're going to try to be as flexible as we can to deliver solutions in the most efficient way possible.
Quinn Bolton
Got it. And my follow-up, Bill, just you've talked about your fourth and your fifth hyperscale customer ramping. It sounds like kind of more second half of the year. You've got 3 10% customers. You're saying there could be 3 to 4. I'm wondering if you think either of the 2 hyperscalers that ramp this year, would you expect them to ramp so quickly that one of those could be a 10% customer on a quarterly basis by the end of fiscal '26? Or would you expect a more modest ramp from the 2 new hyperscalers this year?
William Brennan
I think all of the hyperscalers have the potential to be a 10% customer long term. If we look at more of a short-term outlook, it's very hard for us to get specific with that. I will say that one of the 2 additional customers that we've talked about, the ramp looks like it's going to happen towards the middle of the year. So we're getting some clarity that, that ramp is planned to be a little bit sooner than we expected.
The additional customer beyond that, it's also firming up, but it looks more towards the second half when that's going to happen. Again, really hard for me to say in that time frame how large these 2 additional customers will be. But long term, I surely believe that both of them could be a 10% customer long term.
Operator
Your next question comes from the line of Tom O'Malley with Barclays.
Thomas O'Malley
The first one, to your commentary about a majority of your business being scaled out today. If you look at the number of connections, like you're increasingly seeing a lot of these links are going to be in the scale-up architecture. And you're seeing UAL, scale-up Ethernet and then NV Fusion that will come out in different methods for connecting those nodes. Could you talk about your play in the scale-up architecture?
And do you think that you need to be embedded deeply in one of those protocol/standards to have success? Like to date, it feels like your customer wins are very much one-on-one. But as those standards grow and those protocols grow, like do you feel like that's where you will get next wind of this company, is when you get aligned there? And just any comments on the NV Fusion, please.
William Brennan
Sure. Before we jump into NVLink Fusion and scale-up Ethernet or SUE and UALink, first of all, there's a large market for PCIe. And so that's Gen5 today moving to Gen6. And then longer term, when we talk about these other standards, these are all going to be 224-gig SerDes. So at the -- in the near term, we expect that we're going to increase -- we're going to establish revenue and really increase that revenue base in the PCIe Gen5 and Gen6 time frame. And then after that, we're going to be flexible in the sense of offering Gen7 or doing products that would be universal for all of the standards that we mentioned.
At the physical layer, Layer 1, these are all similar SerDes. And we think that if we talk about AECs specifically, we think that for the 224-gig per lane AECs, we'll be able to support Ethernet SUE, UALink and also NVLink Fusion. So we think generally that the announcement from NVIDIA is good for the market. Open standards are good, and that will open up opportunity for the market in general, including Credo.
Thomas O'Malley
And that's helpful. As a follow-up, I wanted to kind of dive in on some comments you made in the preamble on the optical side, just talking about the success of the execution this year and then thoughts on growth into next year. In terms of your opportunities at higher speeds, are you seeing more traction with customers today? It seems like there's a lot of diversification going on. And just any evidence points of that success in the market thus far.
William Brennan
Yes. Our progress with optical continues. So we felt great about what the team was able to accomplish in fiscal '25. And we're looking again to double or even beyond double our optical revenue in fiscal '26. So the majority of our business today is, believe it or not, 50 gig per lane. and we've got several designs that are in flight, and we're seeing traction now in revenue for 100 gig per lane designs. That is going to continue. Our share increase for 100 gig per lane optical DSPs will continue.
I think when we saw each other at OFC, I mean, you could sense the interest that we saw in the demonstrations that we made there. Probably most impressive coming out of that show was showing an 800-gig LRO DSP built into a module that was being demonstrated, showing great error rates. And the key point was at roughly 9 watts. So there's an increasing focus on power efficiency, especially in AI networks, AI clusters. And so my feeling is that we're going to experience a lot of success in the 100 gig per lane market in the next 12 to 24 months.
And of course, beyond that, 200 gig per lane is coming. I think we're in the camp that says it's going to come a little bit more slowly than predicted as it always seems to happen. But also at OFC, we demonstrated what we believe is an industry-leading 200 gig per lane optical DSP solution, again, leading error performance and setting probably a new benchmark in power efficiency for that 1.6T market. We're going to come out with both full DSP and the LRO variants at the same time. And our expectation is that we're really well positioned in that market as that develops.
Operator
Your next question comes from the line of Vijay Rakesh with Mizuho.
Vijay Rakesh
Bill and Dan, congratulations on a great quarter, I guess. Just a couple of quick questions. I'm at the airport. It's a little noisy. But can you talk to what are the -- what's the traction you're seeing on optical DSP? I think you talked about 3 customers going to 5. Just wondering if they're all full DSP or LRO. And when do you expect those to ramp to become much bigger, I guess?
William Brennan
Sure. So I feel like we're really well positioned right now. If we talk about 100 gig per lane solutions and 200 gig per lane solutions specifically, we're going to have various products that are going to enable our customers to target specifically what they're most interested in. So our 12-nanometer family of DSPs is by far the lowest cost from standpoint of any DSPs in the market. So super cost-effective in 12 nanometer. The product line we came out with recently in 5 nanometer, this sets new benchmarks for power efficiency. And so that is a trend that we see that, that is something that our customers are pursuing pretty aggressively right now.
As it relates to full DSP versus LRO, we're very agnostic. And certain customers, if they could fit under the power ceiling for the module design with a full DSP, they'll go with a full DSP. In other cases, if power efficiency is really critical and we're seeing this market for 800-gig optical DSPs and optical modules really, really becoming much more prioritized, so sub 10 watts is very much an objective. And in the past, those in the industry have said that the only way to do that is to go to an LPO solution, which drops the DSP and creates a whole host of issues with inter-op and diagnostics. So the LRO is a solution. As we've shown, we can deliver on this requirement of sub-10 watt modules and still maintain the benefits of having the DSP designed into the system.
Vijay Rakesh
Got it. And then last, on the scale-up side, I know you should be able to [ at CES ], but when do you expect scale-up revenue to start to get -- or start to show up and get material for you?
William Brennan
Yes. For us, we've been pretty consistent saying that our design wins will come this calendar year, and our revenue ramp will start in calendar '26. And I feel like we're absolutely right on top of that time frame. When I think it's going to become a very material and significant part of our business, I think over the next 2 to 5 years, this is going to be a part of the network where there's intense demand across the board. And so I see our business growing pretty dramatically beyond, say, our fiscal '27 time frame.
Operator
Your next question comes from the line of Karl Ackerman with BNP Paribas.
Karl Ackerman
Yes. I have 2, please. First, Dan, you spoke about CapEx doubling this year to support your sales outlook for fiscal '26. Given the capital commitment to support these programs, could you discuss whether any of these programs are take or pay that may give you better visibility into the manufacturing ramp requirements for these programs over the next couple of quarters?
Daniel Fleming
No. This is -- the largest thing that drives our CapEx is our production mask set tape-outs. So -- and we've talked now for a while about upcoming 3-nanometer tape-outs in fiscal '26. So that's really what's driving it. So it's, in a sense, kind of disconnected from any customer contracts or customer specifically. It's really fundamental devices that we're taping out that we capitalize.
Karl Ackerman
Yes. I understand that you may not have full visibility into every AEC connection. But do you have a general rule of thumb to think about your AEC sales being used to connect GPU compute racks versus custom compute server racks? I ask because there's a misperception that you were over-indexed to custom ASIC server racks for AECs.
William Brennan
Yes, I think you got it right that our AECs are used to connect any kind of GPU to any kind of switch. And we've seen that across the board with our customers. There's definitely no way to look at our AEC business as a proxy for any kind of custom GPU solutions. But we're connecting broadly at a lot of our customers with both.
Operator
Your next question comes from the line of Suji Desilva with ROTH Capital.
Sujeeva De Silva
Bill, Dan and Dan, congrats on the progress here. Maybe it's been a few quarters since some of your competitors have announced AEC products and talked about them. Can you just update us on the competitive landscape? And obviously, you said it would be a large market where you get some share. But maybe you can just talk about the competitive advantages that Credo continues to bring to the marketplace with more folks talking about AEC offerings.
William Brennan
Sure, yes. From our perspective, the competitive environment has not changed meaningfully in the last 90 or 120 days. We know that our customers want multiple sources, and it's really becoming increasingly clear that the AEC market can support multiple winners. So our goal is to maintain our position as the leader in doing that or continuing that leadership. It's really based on delivering innovative solutions more quickly than our competition.
So we're very, very focused at each one of our customers on delivering the next-generation solution, first sampling and then ultimately going through qualification and being the first to ramp. And that has to do with our ability to act quickly in a sense that the entire responsibility for the system-level product sits within the Credo organization. And that really comes from product definition through development, through delivery of -- for samples all the way through qualification and ultimately in production. We take full ownership.
There's no question -- there's no time loss. All of our engineers that work on the different parts of this from SerDes to the ICs, to the system-level cable system to the firmware team, all of them are sitting right next to each other. And so our ability to iterate to close the design and bring it through qualification and production, I think, is probably best in class. And that's really our focus, is to make sure our team is the one that's delivering for our customers first and satisfying them. And I think we've done a great job with that thus far.
Sujeeva De Silva
Great. Okay. It's very helpful there, Bill. And then my second question is on the rack-to-rack 7-meter solution that you're planning, I guess. Is that going to -- the timing of that dovetailing as you get into scale up more so second half of this year, next year? Or can you just talk about when the demand for rack-to-rack extended cables would come in versus the ones you're selling today?
William Brennan
Yes. So I think that the catalysts are really in the way that next-generation data centers are being built. We've talked a lot about liquid cooling and the nonlinear power sourcing increase that's happening. And so we really see the rack-to-rack opportunity come as those deployments increase. So you can do a lot to get significantly higher density for the compute. And that really opens the door to improving the reliability of the network by replacing optical solutions with AECs, effectively eliminating link flaps.
And so we see the -- there's one customer we've publicly talked about being xAI, and that's gone very, very well. We've accomplished their very high-level objectives in building the most reliable cluster possible. But we've got a second customer that's going to ramp this year that the catalyst there was similar in the sense that their ability to move to these longer-length AECs really opens the door for them to improve the reliability. So I believe we'll see that over time. As it relates to scale-up, that's really going to be PCIe Gen6 to begin with. And we expect that, that will become somewhat popular in the calendar '26 to calendar '27 time frame.
Operator
Your next question comes from the line of Christopher Rolland with Susquehanna.
Christopher Rolland
Congrats on these results. Bill, just a broad question for you. Beyond AECs, like if we look out maybe 3 years, 5 years, you've talked about optical becoming larger, I think 10% of revenue or higher at some point. But how would you view your total product mix, AECs versus optical versus, let's say, retimers for scale-up versus other? How do you see this mix broadly playing out 3 to 5 years from now?
William Brennan
It's a good question in a sense that there's a lot of conversation about the necessity to move to optical over time. And I think the market has spoken very loudly that if you can use copper, you will. Even the folks like NVIDIA have been pretty outspoken in saying that. So we believe that the market for copper is going to be very large in the next 3 to 5 years. But with that said, if we look at the largest investments that we're making right now as a company, they're all in the optical space.
So my belief is that over time, yes, we'll have 10% represented by optical, but I think that number will grow if -- especially if you're talking about 3 to 5 years. We see a tremendous opportunity in the optical space, really even beyond what we would think of as traditional optical ASPs. We're looking at implementing the same kind of system-level innovative solutions that we've done with AEC. So there's a lot of learning there. And so when we talk about going beyond the standard, that's something we're very focused on right now.
I will also say that we do see a large opportunity for ICs. And maybe it's in areas that are related to emerging applications within the AI space, specifically with inference. I'm not going to talk too much about that, but addressing different bottlenecks within the landscape is definitely part of our plan, and there's been -- we're investing in different spaces there. But that will definitely play into our revenue in a significant way in the 5-year kind of time frame. So we see a lot of opportunities across the board.
Christopher Rolland
Excellent. Excited to hear more about those products. My second question is around supply. So you guys are putting up just some really massive growth numbers here. And usually, when we have a product take off like this, you can end up with some bottlenecks. So I guess my question is, are you seeing any bottlenecks? Any constraints in either front end or back end or at your cable suppliers? And what has happened to lead times for your products during this growth? Have they grown? And where are they now? Just -- yes, the question is generally like how far do these guys have to book out to get a product.
William Brennan
Yes. I think we've shown in the past couple of quarters a great ability to ramp in a short period of time. And if we kind of break it down specifically to the AEC, there's really 2 pieces. There's the silicon aspect of it, and then there's the system-level aspect. So we've got 2 different operations teams in the company. We've got a silicon operations team and a systems operations.
And so the longest lead time that we've got is really on the silicon side, and that's with TSMC and our assembly partners. On the systems side with the AEC, it's really important that we stay close with all of the different suppliers in our supply chain but specifically the cable assembly partners. We've shown great ability to increase volumes quickly. And so this isn't the same kind of challenge as it would be to expand fab capacity. Basically, if we install an additional line, if we look at, say, a 24/7 operating kind of situation, every line represents about 1 million units in production capacity annually.
And so the investment that's required is not as significant as taping out, say, a 5-nanometer chip or a 3-nanometer chip. It's substantially less. And so the CapEx required, the lead time to get the equipment and the lead time to bring up the line is actually shorter than what it would take to build a semiconductor from start to finish. So I believe we're in a good position that even if we do see the kind of percentages increase that we've seen over the last 12 months that will be in good shape.
So -- and again, it's a huge benefit that our systems operations team has close relationships with all of the supply chain partners. And so it's important that we have those direct connections so that we don't miss a beat when we're trying to ramp quickly. All of our supply chain partners know that we're -- they know that we're ultimately responsible for making commitments and delivering the solution.
Operator
Your next question comes from the line of Joshua Buchalter with TD Cowen.
Joshua Buchalter
Congrats on the results and guidance. I actually wanted to ask about your PILOT software that you talked about in the prepared remarks. How does the SDK and debug tool differ from what your competitors are bringing to market? And I was also hoping -- are there any synergistic elements of -- that the PILOT software brings across your AEC, DSP and retimer hardware?
William Brennan
Sure. So we recently announced the platform, but this has been -- this debug and development tool has really been something that has been key to our success really over the past decade. And so in the Ethernet space, at the speeds where we've been operating, having the ability to offer a solution like this to customers is really critical for them to be able to develop the platforms that they're trying to develop. And so when we talk about getting into the PCIe space, what we're bringing is years of experience and years of knowledge to the space. We are going beyond what we have traditionally done, especially as it relates to system-level features like telemetry.
We learned a huge amount from our efforts in the AEC space. And so the platform that we refer to as a PILOT, it really touches all levels, all 3 tiers of innovation that we've talked about and even extending into our customer system. So it gives great visibility into the SerDes IP, retimer ICs or even the system-level solution. So additionally, adding diagnostic and analytic capabilities along with this telemetry establishes what we think is a new benchmark for the competitive space and specifically related to reliability and uptime stability.
Joshua Buchalter
For my follow-up, I wanted to ask, as your customer base diversifies through the year, can you maybe compare and contrast what types of infrastructure build-outs that you're servicing with your new customers? Are these primarily accelerated AI builds? Are they general-purpose servers? Are they for internal versus external offerings? Like anything that we could -- that would help us better understand the composition?
William Brennan
I would comment that both of the new customers that we've talked about, the first part of the ramp will definitely be related to AI deployments. And then longer term, we see opportunities related to related to disaggregated switch chassis. But as you would expect, it's really the AI application that's driving the increased need for AECs.
Operator
Your next question comes from the line of Richard Shannon with Craig-Hallum Capital Group.
Richard Shannon
I'll echo congratulations on a couple of great quarters in a row. My first question is on your DSP win with a hyperscaler here. I think you said there's 800 gig, but I'm not sure if you said whether it was a full DSP or LRO. If you can share that one. But I think more importantly here, would love to understand how you can describe the share allocation here. It does seem like it's a much bigger deal than any of your wins in the past year. So maybe we can get a sense of the size of this win versus the ones you've had in the past.
William Brennan
Yes, I think it's hard for me to contrast the size of the opportunity. We can look at volumes, and we can look at revenue, kind of 2 different things. And so I think that if you would break it out from a volume standpoint, I think it would be similar in the sense that we expect it to be high volume, as you've seen in the past with our designs. Given that it's 100 gig per lane from a revenue standpoint, it's probably going to be the largest opportunity that we've had to date. And so I'm not going to give too much color given the fact that this is a super competitive space we're in, but I will say that this implementation is a full DSP implementation.
Richard Shannon
Okay. My second question is, I think, a topic that hasn't come up a lot in recent quarters given how well your AEC business is going. But looking at your IP business, obviously -- I mean it's -- obviously, the revenue is coming down here as you apply seemingly more engineering resources to the product side. It certainly makes sense here. But I'd love to get your sense of how you see this business going over the long term. And in response to one of the last questions, you referred to UALink and NVLink Fusion. It just seems like there's some great opportunities there. So maybe you can help us understand kind of the long term for your IP business.
William Brennan
Sure. So I think going forward, we're not going to be breaking it out since it's not going to be above 5% of our revenues. So what we've seen is we've actually we've seen an acceleration of our product revenue that causes us to -- I mean when we did our IPO 3.5 years ago, we signaled that we thought our IP business would be in the 10% to 15% range on a short-term basis of our total revenue. But we've accelerated so quickly that we now find it being sub-5%. From my perspective, it's really a return on investment scenario. And I've always been very consistent in saying that from a percentage revenue standpoint of our total available market, this is tiny and I don't expect that to change.
I think the way we think about that business internally now is engaging where it makes sense from a strategic perspective. And so we've had lots of opportunities we'll be talking to a customer about an overall system solution, and a key enabler can be us providing the core IP for kind of the main chip that they're developing and then complementary solutions that would connect to that. And so those are very valuable contracts as we engage with those types of customers. And I expect those types of relationships will continue, existing relationships and new relationships. So I think as I think about our IP business, we're going to be somewhat opportunistic, of course, but the strategic aspect of it is really enabling system-level solutions with key customers.
Operator
Your next question comes from the line of Tore Svanberg with Stifel.
Tore Svanberg
Yes. I just want to follow up, Bill, because there's a lot of focus on your AEC business, obviously, copper. You did talk about the optical DSP win. But you've also talked about sort of taking that further and perhaps work on a system-level optical solution. I'm just wondering, would you intersect the 200 gig per lane market with that particular product? Or would you consider doing it 100 gig?
William Brennan
I think yes to both questions. I think yes, for 200 gig for sure, 200 gig per lane solutions. But I expect the 100 gig per lane market to extend for several years. And so we're definitely looking at what kind of value can we add above the standard to make our customers' networks more reliable and more power efficient. And so I would say yes, that we're looking at solutions that don't require a significant shift in speeds.
Operator
There are no further questions at this time. Mr. Brennan, I turn the call back over to you.
William Brennan
Thank you very much. I'd like to thank everybody for the continued strong interest in Credo and for joining the call, and we'll talk shortly. Thank you very much.
Operator
This concludes today's conference call. You may now disconnect.
Your email address will not be published. Required fields are markedmarked