Super Micro's Q4 FY2026 Shock: Margins Double, Revenue Nearly Hits $40 Billion, FY2027 Guidance Blows Past Wall Street

Super Micro Computer (NASDAQ: SMCI) just delivered the single most surprising earnings print of the AI infrastructure cycle. On August 11, 2026, the San Jose-based server maker reported fiscal Q4 results that shattered expectations on profitability while guiding fiscal 2027 revenue to a range that Wall Street hadn’t even contemplated. The numbers confirm that the AI data center buildout is not slowing down — and that Supermicro, often dismissed as a low-margin box assembler, is quietly transforming into one of the most profitable hardware companies in the industry.

The headline numbers

For the fourth quarter of fiscal year 2026, ended June 30, Supermicro reported:

• Net sales of $11.1 billion, up 93% year-over-year from $5.8 billion and up 9% sequentially from $10.2 billion in Q3.
• GAAP gross margin of 17.5%, more than doubling from 9.5% a year earlier and crushing the company’s own July preliminary guidance of 15–17%, which itself had been double the original 8.2–8.4% forecast.
• Non-GAAP gross margin of 17.6% and non-GAAP diluted EPS of $1.70, versus analyst expectations of $0.65–$0.79. GAAP EPS came in at $1.62.
• Net income of $1.18 billion in a single quarter, more than the company earned in all of fiscal 2025.

For the full fiscal year 2026, Supermicro posted revenue of $39.1 billion, up 78% from $22.0 billion, with net income of $2.2 billion ($3.26 per diluted share). Non-GAAP EPS for the year was $3.63, a 76% increase from $2.06 in fiscal 2025. Nine customers each generated over $1 billion in revenue during the year, up from just four the year before — a powerful signal of demand diversification beyond a handful of hyperscalers.

The margin mystery solved

The margin explosion was the story of the quarter. Supermicro had guided Q4 gross margins to just 8.2–8.4% at the start of the quarter, then shocked the market in late July by raising that to 15–17%. The actual result of 17.6% non-GAAP exceeded even that revised range.

CEO Charles Liang attributed the dramatic improvement to a deliberate rebalancing of customer and product mix. Enterprise and channel revenue doubled sequentially to $5.6 billion, representing 50% of total sales — up from just 28% in the prior quarter. Meanwhile, OEM and large data center revenue declined 26% sequentially to $5.5 billion. CFO David Weigand explained that roughly 75% of the 750-basis-point sequential margin jump came from this favorable mix shift, while the remaining 25% reflected lower tariff costs and reduced inventory reserves.

The key insight: high-volume GPU rack deals carry thin margins, often in the high single digits. CPU-based servers, storage systems, IoT appliances, and enterprise application servers carry significantly higher margins. By leaning into these categories — and by deferring several large but low-margin AI contracts into the September quarter — Supermicro demonstrated that it can choose profitability over raw volume without sacrificing growth.

Revenue: a temporary dip, not a slowdown

Q4 revenue of $11.1 billion came in near the low end of the $11.0–$12.5 billion guidance range, below the $11.55 billion analyst consensus. But the shortfall had nothing to do with demand. Management explained that several large AI deployments were pushed into the September quarter due to customer-side delays in power provisioning, cooling readiness, and network infrastructure. These are the unglamorous realities of building AI factories at gigawatt scale — the servers are ready, but the buildings that house them are not.

The $60 billion in new orders booked during the quarter — a figure that approaches Supermicro’s entire fiscal 2026 revenue — makes the demand picture unambiguous. Backlog stands at record levels. About 70% of the new orders are for GPU-based AI systems, while 30% are for CPU-based AI platforms, including systems built around NVIDIA’s upcoming Arm-based Vera processors and AMD and Intel CPUs designed for inference and agentic workloads at the edge.

FY2027 guidance: $65–72 billion

The most consequential number in the entire report was Supermicro’s fiscal 2027 revenue guidance of $65 billion to $72 billion. Wall Street had been modeling approximately $52.5 billion, according to LSEG data. Even the high end of analyst estimates fell well short of the company’s range.

The guidance implies 66% to 84% revenue growth from fiscal 2026’s $39.1 billion, an extraordinary acceleration for a company already generating nearly $40 billion annually. For the first quarter of fiscal 2027, Supermicro guided revenue to $14.5–$15.5 billion with non-GAAP EPS of $1.01–$1.10 and gross margins of 10.4–10.8%. The margin normalization from Q4’s 17.6% reflects the expected reintegration of deferred GPU-heavy contracts, but even at 10–11%, it remains meaningfully above the company’s historical 8–9% range.

Liquid cooling dominance and manufacturing scale

Liang used the earnings call to reinforce Supermicro’s position as the dominant player in direct liquid cooling (DLC), the technology rapidly becoming mandatory for high-density AI deployments. The company claims to have shipped over 80% of the market’s liquid-cooled systems in 2024, and its next-generation DLC-2 platform is designed for racks drawing up to 250kW — the power levels required by NVIDIA’s Rubin and AMD’s MI450 accelerators.

Manufacturing capacity is scaling toward 6,000 racks per month, with more than half — 3,000 racks — capable of supporting 250kW liquid cooling. This manufacturing depth is Supermicro’s moat: while competitors can buy the same GPUs, few can assemble, integrate, and deploy complete liquid-cooled AI factory systems at this scale and speed.

A fortified balance sheet

Earlier in fiscal 2026, Supermicro raised $5.6 billion through a combination of $1.4 billion in common stock and $4.2 billion in mandatory convertible preferred shares. The result: cash and equivalents of $7.5 billion at quarter-end, with net debt falling sharply to $1.2 billion from $7.5 billion in the prior quarter. The cash conversion cycle extended to 149 days as inventory built ahead of anticipated shipments, but Weigand assured investors that improved contract terms would accelerate cash flow conversion going forward.

The strengthened balance sheet gives Supermicro confidence to self-fund its growth trajectory. Liang noted that current cash flow is “pretty enough,” though he acknowledged that revenue above $80 billion might require additional capital.

The export control overhang

No discussion of Supermicro is complete without addressing the ongoing independent board investigation into export-control matters. In early 2026, the Department of Justice charged three individuals — including a co-founder — in connection with an alleged scheme to divert approximately $2.5 billion worth of NVIDIA AI chips to China using intermediary companies and falsified end-user documentation. The company’s board launched an independent review, which remains ongoing.

Management declined to provide specifics on the earnings call, saying only that an update would be provided “shortly.” The investigation is a real risk factor: if prior period results are affected, the financial figures reported today could be subject to revision. Investors are pricing in this uncertainty, but the fundamental demand picture — $60 billion in orders, record backlog, $72 billion guidance — is overwhelming the headline risk for now.

What this means for the AI infrastructure thesis

Supermicro’s Q4 results arrive at a moment when the AI infrastructure investment thesis is under intense scrutiny. Bank of England warnings about systemic financial risk, circular financing concerns around NVIDIA’s $500 billion Wall Street financing deal, and broader questions about whether AI demand will translate into sustainable profits have all created noise.

Supermicro’s numbers cut through that noise. The company is generating real revenue from real customers, improving margins through genuine product mix optimization, and guiding to growth rates that would be extraordinary at any company size. Nine billion-dollar customers, a $60 billion order book, and a clear technology lead in liquid cooling are not speculative metrics — they are hard data points on a balance sheet.

The AI infrastructure supercycle is not a bubble. It is the largest capital expenditure program in the history of the technology industry, and Supermicro is one of the companies proving it — quarter after quarter, with numbers that keep getting harder to dismiss.