Super Micro Computer (SMCI.US) Q4 Earnings Call: Gross Margin Surges to 17.6% Surpassing Guidance, AI Infrastructure Leader Enters Harvest Season

According to Smart Finance APP, Super Micro Computer (SMCI.US) announced its Q4 FY2026 results after market close on August 11, causing the stock to surge over 10% at one point. The data shows revenue reached $11.1 billion, a staggering 93% year-over-year increase, while non-GAAP gross margin hit 17.6%, far exceeding the guidance range of 8.2%–8.4%. Non-GAAP EPS came in at $1.70, also significantly beating expectations. The company disclosed over $60 billion in new orders for the quarter, driving the order backlog to a record high, and provided robust FY2027 revenue guidance of $65–72 billion, while Q1 revenue is expected between $14.5–15.5 billion. Management stated they would leverage DCBBS (Data Center Building Block Solutions) to deepen enterprise customer engagement, optimize product mix and operational efficiency, further enhancing profitability amid rapid growth, and consolidating their leading position as an all-in-one AI infrastructure provider.

The following is the English translation of the Super Micro Computer Q4 FY2026 earnings call transcript:

Executive Remarks

Michael Steig

SVP, Corporate Development

Michael Steig: Good afternoon, and thank you all for joining Super Micro Computer's conference call to discuss financial results for the fourth quarter of FY2026 ended June 30, 2026. As you know, joining me today are company founder, Chairman of the Board, and CEO Mr. Charles Liang, and CFO Mr. David Weigand. By now, you should have received a copy of the press release the company issued at the close of regular trading, which is also available on our company website.

Please note, during today’s call, the company will reference a presentation, which can be found under the “Events & Presentations” tab in the Investors section of the company’s website. We have also posted pre-recorded management remarks on the website.

Please note, some of the information discussed today will contain forward-looking statements, including but not limited to statements regarding revenue, gross margin, operating expenses, other income and expenses, taxes, capital allocation, and future business outlook (including guidance for Q1 and all of FY2027). These statements and other comments are based on management’s current expectations and assumptions, involve significant risks and uncertainties, and actual results or events may differ materially from expectations. Please do not place undue reliance on such forward-looking statements. More information about these risks and uncertainties is available in our press release earlier today, our recently filed FY2025 Form 10-K as well as other SEC filings. All of these are available in the Investors section of the Super Micro website. We undertake no obligation to update any forward-looking statement. Much of today’s presentation will include non-GAAP financial results and outlook. For explanation of our non-GAAP measures, please refer to the accompanying presentation or today’s press release.

We provide non-GAAP measures because we believe they provide investors useful ways to evaluate and understand how management measures our operating performance. These non-GAAP measures should not be viewed in isolation or as substitutes for, or superior to, financial measures prepared under US GAAP. Furthermore, reconciliations between GAAP and non-GAAP results are included in today's press release and supplemental information attached to today’s presentation. After the prepared remarks, we will host a Q&A session for sell-side analysts. Our FY2027—apologies, FY2027 quiet period will begin at market close on Friday, September 11, 2026.

I will now turn the call over to Mr. Liang.

Charles Liang

Founder, Chairman of the Board, President, and CEO

Charles Liang: Thank you, Michael, and thank you all for joining today’s conference call. FY2026 was a historic milestone for Super Micro Computer as we nearly doubled our revenue year-over-year, growing from $22 billion last year to $39 billion in FY2026. The world is being transformed by artificial intelligence (AI), and Super Micro is transitioning from a US-based server manufacturer to a leading AI/IT data center solutions company.

We design and manufacture our complete modular Data Center Building Block Solutions (DCBBS) in the US, with our primary facilities located in the US, Taiwan, Malaysia, and the Netherlands. Demand for our AI/IT solutions is stronger than ever as we transform into a comprehensive DCBBS company, providing one-stop services for customers wanting faster and better data center or AI factory construction.

In our outlook, we disclosed more than $60 billion in new orders, pushing our order book and backlog to a new record as we head into FY2027. While fourth-quarter revenue was $11.1 billion due to some short-term customer delays related to power availability, cooling and networking, we know this is purely a timing issue.

The good news is that our customers can now easily leverage our unique DCBBS solutions and our upcoming new technologies and product lines to accelerate their TTD (time to deploy) and TTO (time to operate), ensuring strong growth and long-term value for Super Micro in the years ahead. Most importantly, our focus on profitability is generating clear results.

For the fourth quarter, we delivered outstanding results, achieving a non-GAAP gross margin of 17.6% and a non-GAAP diluted EPS of $1.70. This margin expansion was primarily driven by our strategic focus on balancing our customer and product mix, although we also recognized a few positive one-time contributions this quarter. Since the beginning of FY2026, we have added dedicated departments and resources focused on developing our enterprise customer base and have expanded our enterprise-class CPU server, storage, and IoT product lines.

Our fast-growing inference and agent-AI-centric products are also paving the way to healthier profit margins for the company’s future. Another key to this margin expansion is our DCBBS, which delivers complete solution value through seamless integration of GPU and CPU servers, enterprise storage, direct liquid cooling solutions, CDUs, chillers, cooling towers, high-speed switches and networking, data center management software, and full lifecycle services.

This turnkey ecosystem allows customers to build and scale AI data centers in quarters versus years, significantly lowering total cost of ownership (TCO), and accelerating customer time-to-deployment and revenue generation. We are further enhancing this value proposition with new proactive service models—our data center management software and on-site teams will automatically alert and immediately prepare to repair or service failed units, preventing declines in customer data center compute capacity. With new management-rich and automated software now embedded into our hardware deployments, this deepens customer trust and drives long-term value. Our DCBBS is becoming tremendously strong and will soon add significant net income to our business. More of these software features and service offerings will be available early next week.

Operationally, we have complemented this high-value strategy by improving manufacturing yield through factory automation, design optimization, and highly universal modular architecture. Meanwhile, we remain highly focused on logistics and inventory management, with significant reductions in inventory reserves and expedite fees. In sum, this operational discipline will help smooth out quarter-to-quarter gross margin fluctuations caused by customer and product mix, and support our long-term gross margin growth goals.

Turning to our key product roadmap: our modular system designs enable us to rapidly optimize for every major silicon platform. Through our deep partnership with Nvidia (NVDA.US), we are shipping volume SKUs in the GB300 NVL72, HGX B300, B200 NVL4, and RTX 6000 Pro lines, while preparing to lead with Vera Rubin VRNVL72, Rubin HGX, Vera C1, and other high-density Vera systems. With AMD, we have launched the new Helios family and MI450 complete solutions, along with strong momentum for EPYC CPUs, MI350, and MI355X.

With Intel, we are taking Panther Lake Edge AI systems to market and shipping in volume for the Xeon 6+ platform. We are also developing products targeting robust demand for Arm AGI processors (code-named Phoenix architecture), which are optimized for high-efficiency inference workloads, reflecting our silicon partners’ deep confidence in our engineering capabilities.

To support massive demand, we continue to expand our physical footprint. In Silicon Valley, we have just announced a new 32-acre DCBBS campus featuring advanced photonics labs and data center-scale manufacturing, bringing our US footprint to nearly 4 million square feet. Globally, facilities in Taiwan, Malaysia, and the Netherlands are rapidly ramping up production to meet demand, giving us total capacity targeting over 6,000 racks per month, including more than 3,000 direct liquid cooling (DLC) racks per month. Notably, most of our DLC lines support the latest densest 250kW rack platforms.

Before I close, a quick update on our capital structure: following the $5.6 billion financing completed in June, our balance sheet is fully supportive of our component supply and business needs. Thanks to our strong cash position and improved customer and product mix, we do not currently plan to utilize the ATM program initiated several months ago.

At the same time, we remain focused on building financial efficiency. With all these operational and product advances, I want to emphasize our growth momentum is accelerating in the most critical areas. By adding hundreds of new enterprise and other customers and leading the transformation toward agent-based and specialized AI workloads, Super Micro has become the architect of today’s AI infrastructure.

Our DCBBS end-to-end solutions—including CPU and GPU compute, storage, 800G and 1.6T high-speed switches, imminent photonic networking, and our suite of management software (including SCM, Super Micro Cloud Composer; SDM, Super Micro Data Center Manager; and SOM, Super Micro Orchestration Manager)—provide a one-stop, comprehensive experience needed by modern enterprises, Neocloud, or any data center customer.

Looking to FY2027, our momentum gives us confidence in achieving our $65–$72 billion revenue target, as we are in the midst of a historic infrastructure buildout. By focusing on expanding our enterprise customer base, customer mix, DCBBS solutions, and operational discipline, we seek to balance revenue growth with profitability. We are shaping the future of AI technology while offering our customers real technology value. I am confident that FY2027 will be another strong and fast-growth year for us.

Thank you. I will now turn the call over to David.

David Weigand

CFO, Corporate Secretary & SVP

David Weigand: Thank you, Mr. Liang. In this quarterly report, the company hit a record FY2026 revenue of $39.1 billion, up 78% from $22.0 billion in FY2025; non-GAAP fully diluted EPS was $3.63, up 76% from $2.06 in FY2025.

Our order backlog stood at record levels at the end of FY2026, with more than $60 billion in new orders in the fourth quarter. We expect to fulfill these orders over the coming quarters. FY2026 non-GAAP gross margin was 10.9%, versus 11.2% in FY2025. FY2026 non-GAAP operating profit margin expanded to 8.1% from 7.1% in FY2025. Our customer base is diversifying, with 9 customers exceeding $1 billion in revenue each in FY2026, compared to 4 customers in FY2025.

Turning to Q4 FY2026 results, we achieved $11.1 billion in revenue, up 93% year-over-year and 9% quarter-over-quarter. Revenue was near the low end of our $11–12.5 billion guidance range, due to customer readiness delays, which we expect to recognize in subsequent quarters.

AI solutions accounted for about 60% of total revenue in Q4, versus over 80% in Q3, due to timing of large AI project launches. Based on our order backlog, we believe over 80% of future revenue will come from AI-related solutions. In Q4, enterprise and channel revenue was $5.6 billion, representing 50% of total revenue, up from 28% last quarter. This segment grew 172% year-over-year and 98% quarter-over-quarter.

In Q4, we saw a recovery in demand from enterprise and channel customers, who are upgrading their compute, storage, and networking infrastructure with more efficient CPU platforms. OEM device and large data center revenue was $5.5 billion, also making up 50% of total revenue, down from 72% last quarter. This segment grew 50% year-over-year but declined 26% quarter-over-quarter.

For full FY2026, enterprise and channel revenue grew 39% year-over-year and represented 31% of total revenue. OEM device and large data center revenue grew 104% year-over-year and accounted for 69% of total revenue. In FY2026, we had one large data center/CSP customer accounting for 28% of revenue.

Geographically, the US accounted for 71% of Q4 revenue, Asia 11%, Europe 8%, and the rest of the world 10%. Year-over-year, US revenue grew 259%, Asia declined 50%, Europe grew 4%, and the rest of the world grew 296%. Quarter-over-quarter, US revenue grew 12%, Asia declined 13%, Europe grew 25%, and the rest of the world grew 1%.

Q4 non-GAAP gross margin was 17.6%, with our guidance at 8.2%–8.4%, higher than the 10.1% in Q3. Gross margin rose 750bps quarter-over-quarter, mainly due to better than expected customer and product mix, including pushing several contracts from Q4 FY2026 to Q1 FY2027 and beyond. This favorable mix accounted for about 75% of gross margin improvement. Lower tariff costs and reduced inventory reserves contributed the other 25%.

Q4 GAAP operating expenses were $455 million, up 44% non-GAAP year-over-year, and up 16% quarter-over-quarter. On a non-GAAP basis, operating expenses were $357 million, up 49% year-over-year, and 28% quarter-over-quarter. The sequential increases reflect higher employee-related costs as well as increased sales and marketing expenses. Q4 non-GAAP operating margin was 14.3%, versus 7.2% in Q3.

Q4 other income and expense amounted to a net expense of $19 million, including $61 million of interest and other income, offset by $80 million of interest expense on convertible notes and credit facilities. Q4 GAAP tax provision was $290 million, non-GAAP tax provision was $316 million. Q4 GAAP tax rate was 19.7%, non-GAAP tax rate 20.1%. Full-year FY2026 GAAP tax rate was 19.9%, versus 12.9% in FY2025. Non-GAAP tax rate was 20.4%, versus 15.4% in FY2025.

Q4 GAAP diluted EPS was $1.62, above our guidance of $0.53–$0.67. Non-GAAP diluted EPS was $1.70, vs. our guidance of $0.65–$0.79. The beat mainly resulted from higher gross margins. For FY2026, GAAP diluted EPS was $3.26, compared with $1.68 in FY2025. Non-GAAP diluted EPS was $3.63, versus $2.06 in FY2025.

GAAP diluted shares increased from 692 million in Q3 to 705 million in Q4. Non-GAAP diluted shares increased from 709 million to 721 million. Cash provided by operating activities was $747 million in Q4, compared with $6.6 billion used in operations in Q3. For all of FY2026, cash used in operations was $6.8 billion, compared to $1.66 billion provided by operations in FY2025.

Ending inventory for Q4 was $12.9 billion, up from $11.1 billion at the end of Q3. Capital expenditures were $28 million; free cash flow was $722 million. Full-year FY2026 capital expenditures were $162 million, compared to $127 million in FY2025, primarily reflecting investment in global capacity expansion. This quarter, we completed a public equity offering, raising $5.6 billion after issuance costs, including $1.4 billion in common stock and $4.2 billion in mandatory convertible preferred stock. These proceeds are primarily for working capital to support new orders. Year-end cash and equivalents totaled $7.5 billion. Bank borrowings and convertible note debt totaled $8.7 billion; net debt was $1.2 billion, versus $7.5 billion of net debt at the end of last quarter.

Turning to the balance sheet and working capital metrics, cash conversion cycle increased by 43 days sequentially, from 106 days in Q3 to 149 days in Q4. Inventory days increased 13 days to 119 days as we built inventory ahead of anticipated FY2027 sales growth. Receivable days decreased 26 days to 59 days from 85 days in Q3 as we collected receivables from some large customers. Payable days decreased 56 days from 85 days in Q3 to 29 days, mainly due to major AI GPU projects closing between Q3 and Q4 and the timing of payments to vendors. Looking ahead, we expect the cash conversion cycle to normalize according to terms in our current backlog.

Now for our outlook for Q1 FY2027, we expect net sales of $14.5–$15.5 billion, GAAP diluted net EPS of $0.89–$0.98, and non-GAAP diluted net EPS of $1.01–$1.10. Based on our expected customer and product mix, we expect gross margin of 10.4%–10.8%.

In Q4, we successfully completed a $4.2 billion mandatory convertible preferred offering. Due to this, our GAAP and non-GAAP EPS are calculated on a two-class method, with part of net income allocated to participating convertible preferred shares. This affects our Q1 FY2027 EPS guidance and should be considered in all future EPS calculations.

GAAP operating expenses are expected to be around $453 million, including stock-based compensation of about $127 million. These SBC expenses are excluded from non-GAAP operating expenses. Q1 FY2027 GAAP diluted EPS outlook includes estimated $106 million of SBC (net of $32 million tax effect), which is excluded from non-GAAP diluted net EPS.

We expect other income and expense (including interest) to be a net expense of about $45 million. Our Q1 FY2027 GAAP and non-GAAP diluted EPS guidance assumes a GAAP tax rate of 20.1%, non-GAAP tax rate of 20.5%, GAAP diluted shares of 745 million, and non-GAAP of 761 million. Q1 capital expenditure is expected to be $50–60 million. For all of FY2027, we anticipate net sales of $65–$72 billion.

Q&A Session

Ananda Baruah, Loop Capital

Ananda Baruah: I have two questions. Congratulations on the strong results and sustained profitability improvement. Let me start here: Charles, David, what's the appropriate way to think about gross margin for FY2027? You benefited from product mix in June. It sounds like you're absorbing some mix effects from transaction delays, yet September’s guidance still calls for decent year-on-year improvement. Can you walk us through the gross margin drivers and factors, including mix, CPU, etc.? How should we think about a useful estimate for gross margin in FY2027 and its possible evolution?

Charles Liang: Okay, thank you for your question. Yes, we very carefully balance growth and profitability. As you know, high-volume GPU gross margin is much lower, while CPU, storage, IoT, and enterprise applications carry higher margins. So we strive for balance between these two verticals. In particular, in the past 12 months, we've continuously increased our sales force in enterprise and application, server application, and storage. Going forward, we will continue improving overall gross margin. Although we're still growing aggressively, at a pace that may even exceed GPUs, we’re paying even closer attention to enterprise and CPU storage. At the same time, the DCBBS line is maturing, and we’re shipping more and more DCBBS hardware, as well as software services and some switches. DCBBS will become our long-term, higher-margin product line.

Ananda Baruah: Before my second question—just to clarify, given what you mentioned, do you expect gross margin to improve from the September quarter level? Did I hear that right?

David Weigand: Yes. That’s why we gave a September guide of 10.4%–10.8%. We are doing everything possible, as Charles mentioned, to achieve the best margin possible.

Ananda Baruah: Got it. Sounds like Q3 AI was 80% of revenue; Q4 was 60%, and you expect it to be 80% in Q1. Charles mentioned taking on more CPU server, storage, and networking. How should we think about the 20% non-AI part? Any updates on the board inquiry? I assume it's nearly done, but any updates would help.

Charles Liang: Yes, it depends on customer mix. When large data centers are buying heavily, the AI share is certainly higher. But when David says 80% will be AI, I believe that covers two segments: traditional AI, and application or agent AI, or edge AI. So pure AI is around 60%–70%, another 10%–20% is CPU-based AI or agent/edge AI. The remaining 20% is classic server, storage or IoT—totals to 100%.

David Weigand: On your second question, Ananda, we expect to provide an update soon; that’s all we can share on this call.

Joseph Cardoso, JPMorgan

Manmohanpreet Singh: This is MP for Joseph Cardoso. My first question: I want to dive into the strong orders seen this quarter. You mentioned over $60 billion in orders. Can you provide any additional color on customer concentration within the order growth and what drove such strong order momentum? I have a follow-up.

Charles Liang: Yes. Of the roughly $60 billion, I’d say 70% is pure AI, and the other 30% is CPU or CPU-based AI, like edge AI applications. Overall, I believe our margin mix will keep improving.

Manmohanpreet Singh: Understood. As a follow-up, you mentioned success diversifying customers; 9 customers above $1 billion in FY2026. Any color on their nature—Neocloud, enterprise, or sovereign AI customers?

David Weigand: Yes, we have many emerging Neocloud and CSPs, as well as some enterprise customers in our mix as referenced.

Charles Liang: Yes. The question about CPU-based AI—Nvidia now also has Vera CPU-based AI, plus based on AMD, Arm, Intel. So AI is still mainly GPU, but CPU-based AI is growing fast, especially agent-AI applications.

Asiya Merchant, Citi

Asiya Merchant: Two questions. Any changes in the buying behavior of large data center and CSP customers? I understand there were some shipping delays into Q1. My understanding is guidance includes Q4 shipments being fulfilled in Q1. But are you sensing any difference in buying patterns from these large customers? Because some investors worry clients may shift more directly to ODMs than to Super Micro. Then just a quick follow-up.

Charles Liang: Sure. Large data centers always have power-readiness and data center-readiness concerns, especially for liquid cooling. So our overall customer base has similar concerns. But basically, Q3 and Q4 orders and deliveries remain strong. Also, Super Micro’s business model is unique—we do OEM, but also cover ODM. Now we have many large data center customers, especially Neocloud types, and are also aggressively growing in enterprise server, traditional server, storage. So overall, we continue to grow on both OEM and ODM fronts.

Asiya Merchant: Understood. For my follow-up: For liquid cooling data centers, what percentage of total revenue are they? If possible, please elaborate on the vertical split between enterprise and large data center/CSP customers for liquid cooled data centers.

Charles Liang: As you know, we were an early leader in liquid cooling technology. For example, in 2024, we shipped 80%+ of the liquid cooling products to market. Now, more and more platforms are ready for liquid cooling—including GPU cooling and CPU cooling, like Vera Rubin and even Vera. Vera is CPU-based, and many of our Vera CPU products are adopting liquid cooling, as are some AMD and Intel products. So overall, liquid cooling is growing rapidly and will soon dominate data center business.

Katharine Murphy, Goldman Sachs

Katharine Murphy: Charles, you mentioned investing in the sales team to capture enterprise opportunities. Can you discuss progress there, and what further investment in GTM or Super Micro’s product features/capabilities is needed to better capture this opportunity? Also, is this operating expense run rate the right way to think about the full year?

David Weigand: Of course. On OpEx: some expenses will increase, others may decrease. We feel current levels are appropriate. If you look at our historic OpEx growth rate, it's less than half our revenue growth rate.

Charles Liang: Yes, as a tech company, we've always invested strongly in new technologies—for example, high-speed switching and optical tech—so our DCBBS-centric data center solutions remain robust.

Katharine Murphy: And can you talk about the sales team and how you interact with the expanded enterprise customer base, understanding it’s a broader opportunity than Super Micro targeted in the past?

Michael Steig: Yes, this is Mike Steig. I'll add on the sales team and changes. You may have noticed we promoted several colleagues; Matt Solberg is our Chief Revenue Officer, Vic Malayala is Chief Commercial Officer. We're focused on efficiency and aligning the sales force with solution sales elements, preparing for AI opportunities ahead, helping drive better margin. So, we're making clear improvements and will update you through the year on our expansion and adjustments for market opportunities.

Charles Liang: As a tech leader, previously we mainly focused on engineering, production, and customer service. Now we're paying more attention to enterprise customers and strengthening overall balance, especially with enterprise clients, as you know, margins are better. So we are expanding our sales force aggressively.

Ruplu Bhattacharya, BofA Securities

Ruplu Bhattacharya: David, with the pace of GPU platform transitions, how do you manage inventory risk around each new product generation? What gives you confidence that a record backlog now won't create significant inventory exposure, if customer deployment timelines or platform configs shift? I'm asking because Super Micro has had issues in the past. And a follow-up.

David Weigand: Of course. I believe everyone in our industry needs to be mindful of technology transitions. But based on our experience, because prices rise quickly, some older inventory can sometimes be resold at good prices. All that said, as you point out, you don't want to be caught holding that inventory. It's a real risk. What we do is ensure as much as possible that we have non-cancellable purchase orders. And we match purchases to shipping schedules tightly.

Charles Liang: On top of that, most of our products are designed around modular solutions. So many of our subsystems are compatible and optimized for different product lines and even different generations. That really helps us manage inventory across transitions.

Ruplu Bhattacharya: Got it. As a follow-up, as the business scales towards $70 billion annual revenue, how should we think about working capital intensity and operating cash conversion for 2027? Charles mentioned this in his prepared remarks, but I wasn't clear. David, do you expect growth in 2027 can be self-funded via cash from ops, or does the company need additional external funding beyond the recent issue to support inventory and AR?

David Weigand: Sure, Ruplu. As I mentioned in my prepared comments, we do expect improvement in the cash conversion cycle. Because when we look at the backlog, contract terms have improved, and that should help our cash flow conversion. So we expect this will allow us to support greater scale. We'll use every opportunity to leverage the balance sheet, which is stronger now. Looking at our current assets and liabilities, we're in a better position than most companies. So we expect to use balance sheet strength and a strong customer base to fund growth.

Charles Liang: Yes, and once we stay in the $65B–$72B range, I think cash flow is sufficient. If there's an opportunity for higher revenue, say $80B or above, we may need more cash. That possibility exists, but we will control it closely.

George Notter, Wolfe Research

George Notter: Wondering if you see any easing or change in AI pricing? And regarding your comments on balancing revenue and profit, should we interpret that as you passing on some low-margin deals? Or how do you approach the types of proof-of-concept deals you did in the past on next-gen racks? And a follow-up.

Charles Liang: Yes, that’s why we’re forecasting $65–$72 billion. We want to support as many customers as possible, but the business must be healthy—at least meet minimum financial margin requirements.

George Notter: Understood. Then, on traditional server and storage earnings—how much is pure CPU demand? How are the attach and synergies with AI? And how is that business’s margin moving versus comparable levels?

Charles Liang: Great question. For many years, we focused solely on the GPU and AI market. But as we grow, yes, we are now also focusing on CPU-based enterprise markets—including enterprise, industrial PC, and IoT storage apps. So there'll be a healthier balance between revenue and net income.

David Weigand: I’d add—we did well year-over-year: revenue grew 78%, and net profit nearly kept pace. That shows our aim at the annual level.

Nihal Chokshi, Northland Capital Markets

Nihal Chokshi: Congrats on the phenomenal gross margin result. Charles, do you think Super Micro’s value add to the Nvidia ecosystem is different from that in the x86 ecosystem? This was hinted at in the earnings presentation on the Arm AGI CPU platform.

Charles Liang: Yes, we still have many ways to add value. For example, our DCBBS offers complete data center buildouts—not just GPU, CPU, and storage, but all key data center components. Additionally, with many agent-AI applications, we offer many optimizations—like Vera-based solutions, Rubin HGX, and many workstation platforms. We still see many opportunities to differentiate our products from competitors.

Nihal Chokshi: What I mean is, Nvidia is designing complete systems, and you help customers customize those. But with the Arm AGI CPU platform, maybe Super Micro can offer even more in complete system design versus the Nvidia ecosystem?

Charles Liang: Yes. For example, a faster time-to-market, right? When CPUs or GPUs are available, with our architecture, we offer shorter time-to-market and better quality—not just design, but production and deployment, and cooperation throughout the customer’s data center buildout to ensure high uptime and minimize failures. More and more customers appreciate our partnership. It’s not just “buy and go”; it’s “buy and work together.”

Nihal Chokshi: Understood. You mentioned backlog has improved payment terms in the cash conversion cycle. Is this due to new customers, or because you see a higher proportion of repeat orders in the backlog, which might have better terms?

David Weigand: Yes, it’s both: we have new customers and long-time ones. We tightened contract terms across the board. That’s what gives us clarity on cash conversion cycle visibility.

Brandon Nispel, KeyBanc Capital Markets

Brandon Nispel: About DCBBS. You guided previously that DCBBS would account for about 20% of gross profit this year. Can you update its contribution to this year's revenue and gross profit, and your outlook for 2027? And a follow-up.

Charles Liang: Yes, thank you. DCBBS is a major project. We provide all data center hardware plus management software and network deployment to ensure customers get top availability and efficient maintenance. It’s a total package. For example, our management software, early next week we’ll launch proactive service packages—a feature to maintain max customer uptime, so every server investment is working, and not idle or down. There’s tremendous growth potential, including switches, high-bandwidth networking, and management tools. So we see huge growth opportunities. 20% is within reach.

Brandon Nispel: Understood. David, can you break down your gross margin commentary, with 75% from mix and 25% from tariffs and inventory write-offs? By my math, mix is about $700 million, the rest about $230 million. Within mix, how were vendor rebates this quarter? And as to tariffs, did you record any tariff refunds?

David Weigand: Yes, great question. Let me say, we didn't record any tariff refunds in our numbers. We're actively pursuing rebates, but until cash is in-hand, we won’t book the gain. Some in the industry expect tariffs to come back, perhaps not the same but maybe partially. So we still see it as a possible one-off, but time will tell. On rebates: we saw slightly higher rebates because our mix was different, with more business qualifying for rebates this quarter.

Brandon Nispel: Thank you. One more: On gross margin guidance—Q1 is 10.6%. If we normalize for tariffs and inventory write-offs, how does this compare year-on-year? I know on a reported basis it's up, but last year you had more tariffs and inventory write-off drag, so I’m hoping you can clarify year-on-year normalization for Q1.

David Weigand: Yes, I’m comparing mainly quarter-on-quarter, with much less in tariffs and E&O inventory write-downs. Year-on-year, it’s similar: the IEEPA tariffs were suspended so we had a big reduction this quarter. And clearly, we did well in E&O, which we consider a one-off.

Victor Chiu, Raymond James

Victor Chiu: I’d like to revisit a comment Mr. Liang made earlier. Can you provide an update on how much shift you’ve seen in backlog and end demand from training workloads to agent and inference workloads? Have you observed this inflection in your results? And how might this shift influence mix between CPU and GPU shipments in the mid and long term?

Charles Liang: Yes, it's really a complex mix. I have to say, it’s still 70%+ GPU, maybe 20%+ CPU. But some GPUs now are agent GPUs or edge GPUs. Profit-wise, edge GPU is in between classic GPU and CPU. So it’s a complex mix, but as DCBBS grows fast, I believe we can maintain our targeted margins.

Victor Chiu: Understood. So are you seeing any changes? I get your commentary on mix, but versus a year ago, are there any factors changing the workload impact?

Charles Liang: Yes, long-term, the GPU share will continue to grow, I believe. But at the same time, many GPUs will be used for applied, agent-AI, or enterprise AI. Overall, the GPU market is not only large, but will penetrate every vertical. That’s how we see it right now.