SiTime (SITM) 2026年第三季度业绩电话会：TPD与AI数据中心驱动增长，营收激增

核心要点

• SiTime公布2026年第三季度营收为2.987亿美元，同比增长257%，环比增长90%。本季度是该公司收购瑞萨时钟产品部门（Timing Products Division，简称TPD）后的首个完整并表季度。
• Non-GAAP毛利率达到70.7%，营业利润率扩大至43%。Non-GAAP每股收益（EPS）为4.01美元，同比增长360%。
• SiTime核心业务营收增长154%至2.124亿美元。通信、企业与数据中心（CED）业务领跑增长，营收达1.273亿美元，同比增长202%。
• TPD实现营收8630万美元，较其尚属瑞萨旗下的2025年第三季度增长67%。管理层表示，两大产品组合之间的交叉销售已经展开。
• 针对2026年第四季度，管理层指引营收在3.31亿美元至3.41亿美元之间，毛利率约为70%（上下浮动1个百分点），在营收中点对应的Non-GAAP每股收益约为4.76美元。
• SiTime的潜在商机管线达到100亿美元，较2025年增至四倍。管理层表示，其中半导体业务部分增至八倍，超过40亿美元，但大部分相关收入预计要到2028年和2029年才开始放量。

核心财务数据

除非另有说明，盈利指标均为Non-GAAP口径。

应收账款周转天数（DSO）环比增加10天至61天，主要受TPD回款节奏影响。库存增至1.12亿美元以支撑预期的第四季度需求，资本支出为1120万美元。

业务与运营表现

CED业务营收同比增长202%，环比增长26%至1.273亿美元。管理层表示，互连与网络业务（包括光模块、有源电缆及相关产品）约占该业务的一半，其余部分与机架、交换机、加速卡、CPU、服务器和GPU挂钩。

SiTime认为，随着数据中心光学器件向计算端靠近，时钟产品的价值量将进一步提升。管理层表示，XPO可在振荡器的基础上增加时钟芯片，NPO可带动振荡器用量提升至最高两倍，而CPO可通过更高价值的时钟产品带来高达三倍的时钟器件价值量。相干光模块亦可采用平均售价（ASP）最高达五倍的时钟产品。公司公布在XPO、相干光学、NPO和CPO领域均斩获了设计定点。

汽车、工业以及航空航天与防务业务营收为3550万美元，同比增长76%，环比增长43%。SiTime的潜在商机管线在航空航天与防务领域达到9亿美元，在汽车领域接近10亿美元，在工业领域达到14亿美元。上述数据代表潜在商机，而非已确认收入。

移动、物联网与消费电子业务营收同比增长133%，环比增长58%至4950万美元。受某款调制解调器产品向更多平台拓展的推动，SiTime最大的消费电子客户贡献了4270万美元。公司还提到，采用Titan谐振器和SiTime振荡器的AI智能眼镜及挂坠设计定点预计将于2027年放量。

TPD的营收结构中，80%以上来自CED，约15%至20%来自工业和汽车领域。管理层表示，时钟和振荡器整合后的产品组合正在双方客户群体中创造交叉销售机会。在SiTime的大客户群体中，业务整合后公司对2027年的营收预测提高了两倍至五倍，主要反映了设计定点的增加及产品组合的互补性。

SiTime还宣布了“1拆3”拆股计划。截至2026年12月3日登记在册的股东，每持有1股将额外获配2股，拆股调整后的交易定于2026年12月7日开始。

管理层业绩指引

随着回款节奏恢复正常且业务整合推进，管理层预计第四季度营运资金和现金流状况将有所改善。

展望2027年，公司预计随着围绕TPD搭建财务、人力资源、法务和销售基础设施，并持续投入产品研发，营业费用将逐季递增。管理层还预计，智能手机设计定点的放量将平抑2026年至2027年过渡期内第四季度到第一季度的典型季节性波动。

风险与关注重点

• TPD仍受到晶圆及更广泛供应链限制的影响。管理层预计这种制约将持续至2027年，不过随着SiTime在2027年初更深入介入供应链，最严峻的问题应会有所缓解。
• 半导体潜在商机管线超过40亿美元，但管理层预计2027年仅有相对较小的一部分转化为业绩贡献。大部分放量预计将在2028年和2029年出现。
• CED业务的季度增长可能并不均衡，因为大客户的发货节点可能导致收入在不同季度之间发生转移。管理层表示，2027年的季节性规律难以预测。
• 最大消费电子客户占第三季度营收的4270万美元，这使得其产品的放量成为影响移动、物联网与消费电子业务业绩的重要因素。
• 受TPD回款节奏和库存投资影响，第三季度经营现金流为负，但管理层预计第四季度将出现改善。

分析师问答要点

管理层表示，从800G向1.6T光模块产品的过渡有望支撑出货量、市场份额和平均售价（ASP）提升。SiTime在400G产品上的业务敞口相对有限，而其800G业务增长了四倍以上。管理层预计1.6T产品将实现进一步增长，并看好3.2T带来的更多机遇。

在半导体合作方面，SiTime已组建专门的全球团队对接头部芯片企业。收入可来自直接产品销售、参考设计以及未来的共封装解决方案。公司表示，这些合作关系使其能够更早参与架构决策，但大部分相关的商机管线收入预计不会在2028年和2029年之前兑现。

在高精度同步技术方面，管理层表示，误差从微秒级向数十纳秒级的迈进目前在很大程度上仍属于方向性探索，不太可能在2026年或2027年产生实质性贡献。近期应用预计主要集中在网卡（NIC）以及用于加速卡平台的高精度温补晶振（TCXO）。

管理层表示TPD的业务整合进展顺利。近期供应受限在TPD业务上更为突出，因为SiTime仍需依赖瑞萨提供的过渡期服务；而SiTime原有业务面临的供应制约则被形容为处于可控水平。

财报电话会完整实录

完整财报电话会议逐字稿

管理层陈述

Operator

Good afternoon, and welcome to SiTime's Third Quarter 2026 Financial Results Conference Call. [Operator Instructions] As a reminder, this conference call is being recorded today, November 4, 2026. I would now like to turn the call over to Brett Perry of Shelton Group Investor Relations. Brett, please go ahead.

Brett Perry

Thank you, Kelly, and good afternoon to those on the call. Welcome to SiTime's Third Quarter 2026 Financial Results Conference Call. Joining us on today's call are SiTime's Chief Executive Officer, Rajesh Vashist; and Chief Financial Officer, Beth Howe. Before we begin, I'd like to remind everyone that during today's call, management will make forward-looking statements regarding our business, financial performance, market opportunities, growth strategy and future operations.

These statements are subject to risks and uncertainties that could cause actual results to differ materially from those discussed today. For a discussion of these risks and uncertainties, please refer to today's earnings press release and the company's filings with the SEC, including our most recent Form 10-K and subsequent Form 10-Q filings. SiTime undertakes no obligation to update any forward-looking statements, except as required by law.

During this call, we will also discuss certain non-GAAP financial measures. These measures should be considered as a supplement to and not a substitute for measures of financial performance prepared in accordance with U.S. GAAP. Reconciliations between GAAP and non-GAAP financial measures are included in today's earnings press release, which is available on the Investor Relations section of SiTime's website at www.sitime.com. Lastly, unless otherwise noted, all financial comparisons discussed on today's call are on a year-over-year basis. With that, it's now my pleasure to turn the call over to SiTime's CEO. Rajesh, please go ahead.

Rajesh Vashist

Thank you, Brett. Good afternoon, everyone. Thank you for joining us today. I'm delighted to say that on July 1, we completed the acquisition of Renesas' Timing Products Division, or TPD. Q3 was our full -- first full quarter of consolidated results. I'll break out TPD results today to give greater visibility into the performance of both businesses. SiTime had an outstanding third quarter. Combined revenue was $298.7 million, up 257% year-over-year. Gross margin was 70.7%, up 11.9 percentage points, and we expect it to remain in this range. Operating margin was 43%, up 25 percentage points, and EPS was $4.01, up 360%.

When we announced the acquisition, we said it would accelerate our path to $1 billion in annual revenue, and our Q3 results put us above that level on an annualized run rate basis. Our core SiTime business is strong and continues to deliver remarkable growth. Q3 revenue was $212.4 million, up 154% year-over-year. Our core CED business, or Communications, Enterprise and Datacenter, again led our growth with revenues up 202% year-over-year. Our other BUs also grew more than 75%, grew at least 55%, and 2 regions grew at triple-digit rates.

Demand is strong. Book-to-bill increased from Q2, average order size grew, and total distributor inventory remained below our target. Our visibility into 2027 and 2028 continues to improve with significant growth in backlog, average design-in value, and the size of our funnel. Our total funnel now stands at $10 billion and has quadrupled since 2025. TPD also delivered exceptional growth, exceeding expectations. Revenue was $86.3 million, up 67% from Q3 2025 when the business was part of Renesas. The integration has gone exceptionally well across the teams, customers, and road map.

The TPD team is happy to be part of a company that shares their passion and focus on timing. We see positive responses in engagement and retention across TPD locations. With the addition of TPD, we can provide more complete timing solutions, increasing our content in customer systems and making SiTime a more strategic partner. At our largest accounts, our 2027 revenue forecast has increased by 2 to 5x. Moreover, our greater focus is already creating new opportunities, and the pull-through that we envisioned is happening in both directions.

In CED, rapidly changing data center architectures are creating new opportunities for SiTime. Optics is moving closer to compute, stitching together tens of thousands of GPUs with the bandwidth and latency needed to operate as one system. For SiTime, this creates 2 growth vectors: more complex optical modules and optics coming directly into the compute fabric. First, the new optical module generations are ramping rapidly, and each one raises the bar with higher frequencies, lower jitter, and greater resilience, all areas where SiTime excels.

Importantly, these generations overlap shipping at the same time, and new module types create additional timing opportunities with XPO, or extra dense pluggable optics, adding clocks along oscillators, doubling our content. Between data centers, coherent optics provides the bandwidth to connect large AI workloads. These modules require extremely low jitter timing in the gigahertz range, where our ASP can increase by up to 5x. Second, when optics comes closer to the compute fabric, there is increasing bandwidth with lower power and latency. NPO, or near-packaged optics, drives up to 2x more oscillators content, and CPO, or co-packaged optics, drives up to 3x more timing content with high-value clocks.

SiTime today has design wins with leading customers in XPO, coherent optics, NPO, and CPO. The trend is clear, more bandwidth requires more precision timing. In our other businesses, Mobile, IoT, and Consumer grew 133% year-over-year, driven by our largest consumer customer expanding its modem product across more platforms. At the same time, AI is moving from the cloud into personal devices. That requires more sensing, connectivity, localization, and always-on processing. Together, these trends increase the need for precise low-power timing, and we are seeing new designs in AI smart glasses and pendants that will ramp in 2027 using both Titan resonators and our oscillators.

In our Aerospace, Defense, Automotive, and Industrial BU, our opportunities are expanding as systems become more autonomous, connected, and mission-critical. In aerospace and defense, our funnel has now reached $900 million, driven by radar, counter-drone systems, and resilient alternatives to GPS. Automotive is approaching $1 billion with increasing timing content in zonal architectures and autonomous commercial vehicles. And finally, industrial has reached $1.4 billion, led by physical AI applications such as robotics and smart grid. As these systems continue to become more intelligent and autonomous, real-time decision-making and synchronization become critical, significantly increasing SiTime's timing content per system.

A few quarters ago, we talked about the significant opportunities that come with deepening our relationships with leading semiconductor companies. We have now built a dedicated global sales team to engage customers when critical architecture and timing decisions are being made years ahead of production. This gives us early insight into future requirements and helps us develop differentiated solutions across AI training, inference, mobile, automotive, and other markets. As a consequence, our semiconductor funnel has grown 8x to more than $4 billion, creating a substantial future revenue opportunity.

Our rapid organic growth and the TPD acquisition have structurally changed SiTime. Here are 3 changes that define the new SiTime. First, our financial model is stronger with higher gross margins, sustainable operating margins of 40%, and broad-based growth across every BU. Our seasonality is less pronounced as CED becomes a larger portion of our overall business. Second, our relationships with customers are deeper and more strategic. Importantly, this early engagement allows us to influence decisions around new concepts like chiplets, synchronization, and trusted time, where better precision timing can improve performance and availability.

We expect that these applications will create a new SAM of $2.5 billion by 2030. And third, we now have the scale to innovate faster. With the industry's largest team dedicated solely to timing, we can develop more products in parallel and extend the precision timing leadership. For all the progress we've made, we believe we're still in the early stages of what is possible. Our ambition remains clear. We want SiTime to be the trusted timing adviser to our customers and the default choice for precision timing in the world's most important electronic systems. I'll now turn the call over to Beth, our CFO, to review the financial details. Beth?

Beth Howe

Thanks, Rajesh. Today, I'll walk through our third quarter 2026 results and provide our outlook for the fourth quarter. As a reminder, my remarks focus on non-GAAP financial results, which are reconciled to GAAP in our press release. This quarter marks an inflection point in SiTime's evolution. As our business continues to scale, we now have a broader revenue base, a richer mix of higher-value products, and greater operating leverage. And with the addition of TPD, we are further solidifying our leadership in precision timing.

This progress is reflected in our third quarter revenue growth, gross margin expansion, and meaningful improvement in net income. At the same time, we are continuing to invest in R&D and go-to-market to support the growth opportunities ahead. With that context, total revenue for the quarter was $298.7 million, up 257% year-over-year and up 90% sequentially. The newly acquired TPD business delivered revenue of $86.3 million, an increase of 67% compared with a year ago when the business was part of Renesas.

Excluding TPD, revenue was $212.4 million, up 154% year-over-year, of which Communications, Enterprise and Datacenter delivered revenue of $127.3 million, up 202% year-over-year and up 26% sequentially. Automotive, Industrial and Aerospace Defense generated revenue of $35.5 million, up 76% year-over-year and up 43% sequentially. Mobile, IoT and Consumer revenue was $49.5 million, up 133% year-over-year and up 58% sequentially, reflecting strong sequential growth in our large consumer customer that contributed $42.7 million in the quarter.

Third quarter gross margin was 70.7%, up 11.9 percentage points year-over-year and 3.6 percentage points sequentially. The primary driver was favorable product mix with a greater contribution from higher-value, higher-margin products in our core SiTime portfolio as well as the addition of TPD. Importantly, this improvement reflects the products we are selling rather than price increases. Better manufacturing absorption from higher revenue also contributed. Operating expenses were $81.6 million, consisting of $48.6 million in R&D and $33 million in SG&A.

Operating expenses increased $29.5 million sequentially, in line with the outlook I provided in August. We continue to invest in the engineering capacity, product development, and go-to-market needed to support our growth while maintaining discipline on our overall cost structure. Operating income increased to $129.5 million or 43% of revenue compared with 34% of revenue in Q2 and 18% a year ago. Other income was $0.8 million. As a result, non-GAAP net income increased over 5x to $130.3 million, and non-GAAP earnings per share were $4.01.

These results reflect strong revenue growth, favorable product mix, and meaningful operating leverage even as we continue to fund our product road map and invest for growth. Turning to the balance sheet and cash flow. DSO was 61 days in Q3, up 10 days from Q2, primarily reflecting the timing of collections related to TPD with those collections received in early Q4. Inventory increased to $112 million to support expected Q4 demand. Together, these working capital movements contributed to the $6.6 million use of operating cash in Q3. We expect working capital and cash flow to improve in Q4 as collection timing normalizes and integration progresses.

Capital expenditures for the quarter were $11.2 million. We ended the quarter with $324 million in cash and short-term investments, providing significant liquidity and flexibility to continue investing in the business. Now looking ahead to Q4, the TPD integration is progressing extremely well. And going forward, we will provide our outlook and report our results on a combined basis. For Q4, we expect revenue of $331 million to $341 million, with both the core SiTime and TPD growing sequentially Q3 to Q4.

We expect gross margin to remain around 70%, plus or minus 1 point, reflecting the expected mix of the business. Operating expenses are expected to remain roughly flat with Q3. We expect interest income of $2 million to $2.5 million and a diluted share count of approximately 32.8 million shares. At the midpoint of the range, we would expect Q4 non-GAAP EPS to be approximately $4.76 per share.

And finally, today, we announced a 3-for-1 stock split of SiTime common stock to make our stock more accessible to a broader base of investors. The shareholders of record at the close of business on December 3, 2026, will receive 2 additional shares for every outstanding share held on the record date, and trading will begin on a split-adjusted basis on Monday, December 7, 2026. With that, I'll hand the call back to the operator to open the line for questions.

Operator

[Operator Instructions]

Our first question comes from the line of Tore Svanberg of Stifel.

分析师问答

Tore Svanberg

Rajesh, you gave us some great detail there on the exposure to AI data center within the various networks, especially on optical. I'm just curious from where you sit, where do you expect the content to increase the most for your products, whether it's clocks or oscillators? I mean it sounds like there's a step function in NPO. My understanding, there's a further step function then as we start to see CPO get deployed?

Rajesh Vashist

Yes. I think there's -- I think to answer your question, Tore, it's across the oscillators as well as the clocks. So one of the easiest ways to say it is my example of the low-jitter gigahertz products, they typically are going to be at a higher price, maybe 3, maybe 4x. So a gigahertz oscillator may cost $6 to $8, whereas a megahertz oscillator might be around $1 to $2. So right there, there's an up level in pricing.

The other thing that happens is that in NPO, for example, the number of DSPs are much greater. And wherever there are more DSPs, there are more oscillators. And there certainly is more clock being used in the NPOs as well, where, as you know, the electrical part of the switch is -- the optics are closer, and we have some design wins around that as well.

So overall, this is a question that's been asked a few times about what is the impact when there is less copper whenever that happens and more optics. And we have always maintained that SiTime benefits to the upside in this trend. And so we're confirming that with some more detail here.

Tore Svanberg

As my follow-up, I'm very intrigued with the work that you've been doing with semiconductor companies, and I think you mentioned the funnel growing 8x to about $4 billion. Could you maybe unpack a little bit how that works? I mean, do you get paid by the semiconductor companies for using your chips? Is this perhaps more of like a reference design? And when do we start to see that becoming or that $4 billion funnel materializing into more revenues?

Rajesh Vashist

Yes. So as I've mentioned before, I've talked in previous quarters and said that that's one area where we wanted to invest more money. And my point that I was making was that we now have a fully funded team of several people that are calling on anywhere from the top 10 to top 20 semiconductor companies and going even beyond that. So the answer is all of the above to how do we make revenue. First of all, we make revenue directly with some of the customers. In some of these cases, we're directly twinned with the product and sold to the semiconductor company itself and then they sell it.

The second part is in the reference design, of course, we get to play in that. And third, there is this co-packaging piece going on, which is going on at a relatively little pace or low pace, which we expect to ramp up in '28 and '29 in particular. So given the time deadlines here, none of this or relatively smaller piece of this $4 billion funnel happens in '27.

Most of it starts to ramp in '28 and '29. And we just wanted to say that there is another aspect, and that is when we engage with semiconductor companies, the conversation is extremely detailed and collaborative in changing architectures to recognize unique products from SiTime. And this gets us embedded further, but it also makes us more consequential to our customers. So I think it's just an amazing win for us, and I'm really glad that we have the team up and running.

Operator

Our next question comes from the line of Quinn Bolton of Needham & Company.

Quinn Bolton

Rajesh, I wanted to start with the TPD business. I know that segment has been fairly wafer constrained. I wondering if you could give us an update on how you see wafer availability for that business? Do you still expect constraints to last into perhaps as far as middle of next year? Or have you been successful with Renesas in increasing wafer support for the TPD business?

Rajesh Vashist

Yes. So I think in general, constraints will continue through the coming year through 2027. It's the severity of the constraints that we have been very concerned about in this handoff not wanting to constrain our customers, frankly, our mutual customers because Renesas continues to sell to them and wanting to make sure that there are no severe consequences, lines down and such. So with that, we've been quite successful both with our team as well as the support from Renesas.

But in general, as I said once before, we expect that we will get closer to the supply chain ourselves by Q1, and we'll start to resolve them. The more severe issues exiting this quarter and getting into Q1, Q2, I think we should be in significantly better shape. However, I've said that is that there is constraints in the supply chain, particularly on the wafers because of the increased demand. Frankly, we -- neither we nor TPD had expected this huge surge in demand, and we're seeing significant growth, all to the good, but really something for us to manage, which I'm pretty confident we will by early part of next year, 2027.

Quinn Bolton

And then a follow-up maybe for Beth. Just wondering, it sounds like the book-to-bill continues to strengthen. Wondering if you could give us a sense how far out do you have visibility or backlog? And what are lead times doing on the products? Are they increasing? Are they stable? Just any sort of comments on lead times and visibility?

Beth Howe

Yes. So we have exceptional visibility heading into 2027. I think we've got a strong design win funnel, and we've got a strong backlog going into the year. So we feel very well positioned as we kind of finish up '26 and going into '27. When you think about lead time, we're managing the lead times. As Rajesh was just talking about in the supply-constrained environment, it's something that the ops team works really hard with our suppliers on in order to ensure that we can get products, and we're managing through it, I think, is the best way to frame it.

Operator

Our next question comes from the line of Suji Desilva of ROTH Capital.

Sujeeva De Silva

I think I heard you say the Renesas product business, you gave a year-over-year number, even though I know it's included for the first time this quarter. Did you say 67% year-over-year? And if you did, what end markets are the key drivers there? And how is that exposure different from SiTime's core end market exposure? Just to understand the sustainability of Renesas.

Beth Howe

Thanks, Suji. So you're correct. The TPD business grew 67% year-over-year, similar to comparable to when they were part of Renesas. The mix of that business is overwhelmingly probably 80-plus percent in what we would call our Comms, Enterprise, Datacenter and the other roughly 15% to 20% is in Industrial and Automotive for them. So that's the kind of mix of the businesses. And like our business, over the last year plus, it has moved a bit more. They previously were more like 75/25. So it's moved a bit more in toward the CED, not unexpectedly, given the growth in the market.

Sujeeva De Silva

Okay. That's helpful color there. And then I know when you guys did the acquisition merger, you talked about Renesas' customer base as being an exciting part of the opportunity here and cross-sell synergies as they had established Fortune 100 customers. Any near-term kind of early updates there, Rajesh, on how those discussions are going, how the cross-sell opportunity looks to you? Is it going to take time? Or is that moving along at a nice pace?

Rajesh Vashist

Yes. No, that's what I sort of hinted at in my prepared remarks, Suji, that the focus has created new opportunities and the pull-through or the drag along, if you will, is happening in both directions. I think it's only been a quarter, and I think we've already seen the benefit of selling into the TPD customers that some of them they were very strong at. And I think the reverse where we have taken them in with the clocking products into customers that we were stronger at, I think, is already happening. I think clocking is clearly more architectural, more "strategic" to customers -- so that probably helps us, the old SiTime more, but it's also true that we are able to take them significantly into some of the new architectures, which will pay off in the future. So it's really working exactly the way I thought, including with the integration.

Operator

Our next question comes from the line of Gary Mobley.

Gaurav Saini

Gaurav Saini from StoneX on behalf of Gary Mobley. I just wanted to ask specifically about the design win pipeline. We're starting to hear more customers shift towards precision timing and away from quartz. So I'd love to get a sense at, a, how has your guys' opportunity pipeline changed over, call it, the last 12 to 24 months? And b, how are you sort of handling that on the resource side, both your own sales team and then also your distribution network? I'd love to get some color on that.

Rajesh Vashist

Yes. So the -- as I said, we've seen quite a surge in demand. I mean, especially when you widen the aperture from 12 to 24 months, there's no doubt that the increase in demand has been amazing. We can now see not just into -- and right, sitting here in '26, not only do we see into '27, but we can see into '28 as well. And there are new architectures that I talked about, whether it's with semiconductor companies or hyperscalers that we can see. That is also true, by the way, in the non-CED business.

I think we have opportunities in military, Aerospace, Defense, and Industrial. We talked about robotics. We talked about the smart grid. I think, as you know, in Automotive, it's a little bit of a longer cycle, and we're starting to see some of those longer cycles play out now in our favor. So I think we have significant breadth coming in that. And so -- our funnel is larger. Our number of salespeople and FAEs is significantly larger. So we probably have the largest sales force now calling on it worldwide with special emphasis in China and in EMEA. So that's helping us.

As far as distribution goes, as you know, we are fulfilling through distribution. We are a design win company that chooses to basically sell through. We don't have a distribution business that's very large, that's selling into stocking and then selling out of stocking programs. It's very much made for the customer, program to the customer, design for the customer thing. So our distributor inventory is at very healthy levels, and we are very happy with that.

Gaurav Saini

Got it. And then just my follow-up. I know you guys mentioned core CED grew over 200% year-over-year, which is great to see. So I just wanted to double-click on that. How has the mix shifted, call it, again, over the past few years between CPUs, GPUs, custom xPUs? And I know you guys also mentioned optical modules. Just trying to get a better understanding of whether this has moved away from being GPU-centric and become much more diversified.

Rajesh Vashist

Yes. I don't think we see a big change in that. We still see the connectivity part, the networking part as a very strong piece, call it, around 50% of our business is around modules, AECs, connectivity. But at the same time, on the rack side of the business, our business continues to grow with the switches, with the accelerator cards, with the level of inference where we get 2x to 3x more in the CPU and the servers and continued strong business in the GPU. So it's kind of like a 50-50 thing. And both of those markets continue to grow solidly.

Operator

Our next question comes from the line of Sajal Dogra of Rosenblatt Securities.

Sajal Dogra

Congrats on the great results. So you said on your '27 revenue forecast that your largest customer accounts have increased 2x to 5x. So what does that compare against? And how does that increase? Like does that refer to TPD? Does it core SiTime? Or is it the combined portfolio? And can you give us a sense of the incremental dollars? And how much of this is already backed by firm customer orders?

Rajesh Vashist

Yes. Thank you, Sajal. Welcome to SiTime. Basically, what we are seeing -- yes, we're basically seeing a significant increase in design wins. So most of that refers back to the design win side of the business. And what we wanted to do was give our investors a flavor of how strong our pipeline for design wins is and how strong our funnel is. And so all of that refers to that, and it's all over in all areas really across the business. And maybe Beth wanted to say -- to add something to that.

Beth Howe

Yes. And in addition, it also points to the complementary nature of our TPD plus SiTime. As we talked about when we did the deal, there were a number of customers where they had a large installation and maybe were less of a customer for SiTime and some of SiTime's larger customers where TPD had less of a presence.

And so one of the things that we're talking about here is this combination of now having TPD and SiTime. You're seeing that complementary nature in that increase in our forecast as we bring these 2 companies together and some of those customers where maybe SiTime hadn't had much of a presence previously, we now have a much larger presence.

Sajal Dogra

Got it. And so as my follow-up, I guess at OCP, we heard architects discussing a move from basically microsecond of synchronization error towards tens of nanoseconds. So in the design wins for '27, where does that first translate into a SiTime socket? Is it on the NIC side? Is it on the compute tray? Or is that still like on the switch? And how much of that stuff is shipping in current design wins versus the next generation?

Rajesh Vashist

Yes. Very little of that will be shipping in the '26 or even in the '27 time frame, Sajal. I think that's all directional. And I was at some part of it, and it's a very kind of a thoughtful conversation on where the system is headed. But I think directionally, you're correct. When it gets to that level of precision, I think there's almost nobody else who can supply what we are talking about with the kinds of latencies that we're talking about.

Clearly, optics, the greater move into optics is going to drive more of that. And I think there's some innovative architectures that are happening now, but I wouldn't expect much beyond NICs and TCXOs from SiTime connecting, right? So the NIC cards connect with the high-precision TCXOs and there's just more use case of them on the accelerator parts.

Operator

Our next question comes from the line of Nicole Kozhukhov of Morgan Stanley. I'm not hearing a response at the moment.

Beth Howe

Let's move on and we can come back to Nicole.

Operator

Our next question comes from the line of Jim Schneider of Goldman Sachs.

James Schneider

I know you don't want to guide more than 1 quarter in advance, but I was wondering if you could help us understand at least the specific design win in the smartphone segment at one of your large customers, what sort of seasonality will that have, if any, heading into the March quarter? Or would we think that some of the model ramps that are happening at the beginning of the year would more than offset that typical seasonality? Just wanted to make sure we're level set on that.

Beth Howe

Jim, this is Beth. Thanks for the question. So as we ramp that design win, I do think it will modulate the normal typical Q4 to Q1 seasonality a bit for -- at least for this '26 to '27 year.

James Schneider

And then as we -- I mean, I guess, as we think about the CED business over the course of '27 and start to model that, how would you encourage us to think about the linearity of that revenue? Because I know there's many moving parts between your own business ramping with the design wins and also the Renesas business, presumably a little bit less supply constrained. And maybe give us -- if you can address anything about the supply constraint end of it, that would be great.

Beth Howe

So as we think about seasonality in CED, I think given all of the dynamics of that market and the way AI is evolving over the last many quarters, I think it's hard to necessarily talk about seasonality per se, but probably '27 looks a bit like '26, but I think it's really hard to predict the seasonality because it can come in the form of with these big customers, a little bit of revenue shifting from one quarter to another depending on when exactly we ship can move kind of those -- the growth percentages there. But '27 probably looks comparable to '26 from that perspective as we continue to see sequential growth in CED. So that's kind of as we see it today.

Rajesh Vashist

And to add to that, in terms of tightness of supply, I think that in the short term, in Q3 and Q4, we've struggled more with the supply chain from the acquisition of TPD because we've had to rely on TSAs from Renesas. And so as we have gotten closer to the business, trying to understand the specific needs has taken a little bit of time given the supply chain that we have had to deal with.

As to the old core business, if you will, of SiTime, the old business of SiTime, sure, we have some supply constraints, but nothing that we can't manage. In general, we think we'll have managed all of this by the start of Q1. And certainly, the more stark parts of the shortages for the TPD business will probably be handled through this quarter, and we'll be in positive space in Q1.

Operator

Our next question comes from the line of Chris Caso of Wolfe Research.

Christopher Caso

Question is about the diversification of the CED business. And I know that your share and your content isn't equal across all the platforms, but I know you're making some efforts in order to do so. Can you comment on that? And I mean, considering that some of the custom xPU solutions, we're expecting to probably grow faster than some of the merchant solutions over the next year. What does that mean for the CED business as you go into '27?

Rajesh Vashist

Yes. I think -- thanks, Chris. I think we expect that to have a pretty strong impact on growth because we are seeing multiple design wins in that space. So we expect that to do very well in CED in general. What we wanted to give was a sense in the -- when we talked about the funnel was a sense of how big it is and how pervasive across all applications it is, whether it is in the GPU, the CPU part, whether it is in NICs and switches, but also in the connectivity pieces.

And the 1 plus 1 equals 3 part, but getting the oscillators and the clocking solutions together of old SiTime and TPD for the new SiTime, I think it makes our portfolio complete and allows us to win designs against people who have been in the clocking business and gives us significant conversation around architectures that we didn't have. So it's really coming through for us the way we expected it to.

Christopher Caso

Got it. As a follow-up, Beth, I know you've spoken about the necessity now that you brought the TPD business on to kind of fill in some of the things that didn't come along with TPD such as sales. What should we expect for OpEx as we go through the next year as you make some of those hires and you integrate that TPD business and fill in some of those blanks?

Beth Howe

You're absolutely right, Chris. When we acquired TPD, we primarily acquired engineers, a few FAEs and a few folks in operations, and we're building out the infrastructure to support that business, whether it be finance, HR, legal, even building up a sales team as well to support that.

So that -- growing that infrastructure combined with supporting the growth of our own business, I would expect that we continue to see kind of sequential increases in OpEx from Q4 to Q1, Q1 to Q2. In general, that's what I do expect is those sequential increases. Now from time to time, as we're seeing Q3 to Q4, the timing of program expenses or some of these items can be a little uneven quarter-to-quarter. But overall, I would expect sequential growth as we go through '27.

Operator

Our next question comes from the line of Nicole Kozhukhov of Morgan Stanley.

Joseph Moore

We figured this out. It's Joe Moore from Morgan Stanley. Somehow I got registered under the wrong name. I wonder if you could talk about the market share on 1.6 terabit. You've talked about being much higher market share than you were at 400 gigabit. Where are you now from a mix standpoint, 400, 800, 1.6? And should we think about that as a growth driver for you as 1.6 becomes a bigger part of the mix?

Rajesh Vashist

Yes, absolutely. We should as it transitions towards from 800 to 1.6 and beyond that for that matter to 3.2. We didn't have a very large portion of our business in the 400. So while we are shipping in the 400 and we expect that to ship for a couple more years, we have a relatively small market share.

In the 800, it grew over 4x. And I would say that in the 1.6, it grows again. So there's a unit uplift, market share uplift, but there's also an ASP uplift going from 800 to 1.6 terabit. And we have indicated that as being a significant value of growth in revenue because of performance and other attributes that we have.

Operator

This concludes the question-and-answer session. I would now like to turn it back to Rajesh for closing remarks.

Rajesh Vashist

Yes. Thank you for that. We really see ourselves as the company that after the acquisition has really transformed itself into this place where there's fundamental changes between our scale. I think the scale of engineers, the scale of sales, the scale of the business itself is very high, and it helps us to grow the financial transition over to high gross margin, high net operating margin company, I think, is already happening.

And then the closer move to customers across the market, I think, has really changed SiTime. So I think our customers recognize it. Our partners recognize it. And I think we wanted to make sure that our investors get to see that we see ourselves as having fundamentally transformed from this $1 billion tipping point revenue onwards to greater growth and greater impact in the market. So with all that, we thank all our investors, and we invite them to join the ride for future and even better times. Thank you very much.

Operator

Thank you for your participation in today's conference. This does conclude the program. You may now disconnect.