Supermicro Guides to $65-$72 Billion in Fiscal 2027 Revenue; Shares Rose 13% on Aug. 12 as Dell and HPE Also Gained

Two Numbers From the Same Report Point in Opposite Directions

Super Micro Computer's fourth-quarter results, released Aug. 11, 2026, contained a miss and a beat sitting side by side. The San Jose server maker, known as Supermicro, reported net sales of $11.1 billion for the quarter — below the roughly $11.55 billion analysts had expected^[5]. That's a revenue miss, plain and simple.

But profit blew past estimates. Non-GAAP earnings came in at $1.70 per share, versus a consensus near $0.92^[1][7]. Gross margin — the share of every sales dollar left after building the product — jumped to 17.5%, up from just 9.5% a year earlier^[1]. And the company guided fiscal 2027 revenue to $65 billion to $72 billion, far above the roughly $54.4 billion analysts had modeled^[5].

Investors picked a side. SMCI shares rose about 13% on Aug. 12, closing at $35.81^[4]. Dell Technologies gained about 5% and Hewlett Packard Enterprise about 3% the same day^[4]. Memory chipmakers Micron and SanDisk, already climbing for months on a separate story, kept rising too^[9][10].

The genuine dispute here isn't whether AI spending is real. Bulls and skeptics agree it is^[13]. The dispute is over which of those two numbers — the miss or the beat — tells you more about where Supermicro is headed.

What $60 Billion in Orders Actually Promises

Supermicro's strongest evidence is its order book. The company says it took in more than $60 billion in new orders during the fourth quarter alone, pushing its backlog to record levels^[2]. That number is doing most of the work behind the aggressive fiscal 2027 guidance.

But an order is not a sale. It's a claim on future revenue, not cash in hand. Supermicro's own disclosures warn that some of those backlog orders aren't firm commitments — they can be canceled, delayed, or fall through if conditions aren't met^[15]. That's why the $11.1 billion of realized sales and the $60 billion of orders are fundamentally different kinds of numbers^[1][2].

The margin jump raises a similar question. Supermicro credited the leap to 17.5% gross margin to "customer and product mix" — meaning which customers happened to buy which products that quarter^[2]. That's exactly the kind of thing that can reverse the next quarter if the mix shifts back.

Wolfe Research analyst George Notter raised his price target but drew a clear line on what he was and wasn't endorsing. He said Supermicro's gross margins "have historically been volatile and subject to large swings in customer mix," adding he's "not ready to call the gross margin improvement a trend"^[5]. Raymond James' Simon Leopold flagged a different risk entirely: a possible shift in how Nvidia allocates its chips, which he called "impossible to quantify"^[5].

Why a Server Assembler Can't Just Keep the Upside

That Nvidia point is the load-bearing one. Supermicro can't build an AI server without Nvidia's GPUs, and Nvidia decides who gets them and when. That single dependency caps how much of the AI boom's profit Supermicro can keep, no matter how strong its order book looks.

Nvidia's own strategy is narrowing that margin further. The chipmaker has been shifting toward selling complete, integrated rack-scale systems — like its GB300 platform — rather than just component GPUs for others to assemble^[8]. Asia-facing brokerage coverage has framed this as pressure squeezing Supermicro's role in the chain, with Taiwan-based manufacturing remaining the real chokepoint upstream^[8][16].

This is a structural reality, not a one-quarter problem. Nvidia and TSMC sit at the narrowest, most defensible points in the AI hardware supply chain and capture the most value. Server assembly sits at the widest point, crowded with competitors bidding against each other for the same customers. That structure doesn't change because one quarter's margin looked good.

Underneath Supermicro's aggressive guidance sits a second pressure that has nothing to do with Nvidia: credibility. The company's auditor, Ernst & Young, resigned in October 2024 during an audit, citing concerns about governance, transparency, and internal controls over financial reporting^[12]. Since then, Supermicro has funded its growth partly through convertible debt and stock sales rather than cheaper bank borrowing^[11][12]. A guidance number far above what Wall Street expects is one of the few tools management has to rebuild trust and lower that cost of capital — a real incentive to guide boldly, separate from whether the number ultimately proves accurate.

The Stock That Was Already Losing the Race

Here's the detail that undercuts the "Supermicro sparked a sector rally" framing: Supermicro entered its earnings report as the group's laggard. It was up only about 8% year to date, while Dell had gained roughly 250% and HPE about 126%^[4]. A single day's 13% pop doesn't erase that gap.

Dell and HPE make a different case for why they've outperformed all along. Their edge isn't assembly speed — it's enterprise sales forces, financing arms, service contracts, and long-standing relationships with big corporate and government buyers that a price-competitive assembler can't easily replicate^[8]. Those relationships are the evidence they point to for why their stocks trade so differently even on days when the same AI news hits all three companies.

Micron and SanDisk's gains have an even more separate cause: physical scarcity. DRAM and NAND chip prices have risen roughly 300% and 200% over the past year, driven by a supply shortage that predates this earnings report entirely^[9][10]. Micron has said meaningful new industry manufacturing capacity won't arrive until after 2028^[9]. Bullish analyst notes on memory stocks were already lifting sentiment across the whole server-hardware group the same week Supermicro reported^[4] — which means causation may run backward from the popular narrative, with memory optimism spilling into Supermicro rather than the other way around.

What the Buyers Are Betting On, and What Could Go Wrong

Behind all of this sits the hyperscalers — Amazon, Microsoft, Google, and others — whose orders are the actual source of Supermicro's backlog. Estimates of their 2026 AI capital spending range from about $660 billion to $800 billion, depending on who's counting^[13]. Their logic is that in an AI capacity land grab, underbuilding is the bigger risk than overbuilding.

A quieter accounting debate hangs over that spending. Hyperscalers currently write off AI servers over five to six years for accounting purposes. Critics argue the real useful life is closer to two or three years, since each new Nvidia chip generation can make the last one obsolete for frontier work^[13]. If the critics are right, today's reported profits at these companies look better than they should, with the true cost showing up later as write-downs. Defenders counter that older GPUs don't get scrapped — they keep working on cheaper, less demanding AI tasks long after they stop being cutting-edge.

Coverage of Supermicro's report split along familiar lines. Trade outlets like SiliconANGLE led with words like "crushed" and "soaring," with the revenue miss largely absent from the top of the story^[7]. Reuters paired the rally with the analyst caution in its own headline, treating them as inseparable^[5]. The Next Platform argued explicitly that the broader AI boom, not Supermicro specifically, is what's lifting the whole sector^[8].

What actually happened is not in dispute: Supermicro's fiscal 2026 revenue nearly doubled, from $22.0 billion to $39.1 billion, and its backlog is real^[1][2]. What remains unresolved is how much of that backlog survives contact with the next few quarters — and how much of it Supermicro gets to keep once Nvidia, its buyers, and its own history all take their share.