Record Revenue, Falling Stock: HPE's Q3 FY26 Shows AI Demand Is Real — and Chained to Memory

Record Revenue, Falling Stock: HPE’s Q3 FY26 Shows AI Demand Is Real — and Chained to Memory

On September 2, 2026, Hewlett Packard Enterprise reported the best quarter in its history — and its stock went down anyway. The disconnect is the story. HPE’s fiscal Q3 2026 results confirm that AI infrastructure demand is accelerating faster than almost anyone modeled. They also confirm that the industry’s ability to actually ship that infrastructure is now gated by something far less glamorous than GPUs: memory chips.

The numbers

HPE’s headline results, announced after the close on September 2:

• Record revenue of $12.2 billion, up 34% year-over-year — smashing the analyst consensus near $11.9 billion.
• Record operating profit, with GAAP operating profit up 464%.
• Non-GAAP EPS of $1.11, up from guidance and well above the $0.93 consensus; gross margin held at 40%.
• AI systems revenue of roughly $1.6 billion, with AI systems orders of $2.4 billion, up more than 30% sequentially — new bookings meaningfully outpacing recognized revenue, which is how backlog gets built.
• Record AI backlog of $3.1 billion, per the earnings presentation.
• Total orders rose 42% year-over-year, with backlog at record levels across segments.

Guidance went up, not down. HPE now expects fiscal 2026 revenue growth of 34% to 37%, raised from the prior 29% to 33% range, and lifted its adjusted EPS outlook to $3.75–$3.85 per share. The company also raised free cash flow guidance and issued fourth-quarter guidance of $13.9–$14.8 billion in revenue with $1.20–$1.30 non-GAAP diluted EPS. As Constellation Research noted, HPE even pulled forward milestones it had previously targeted for 2028 — in June it had already said it expected to hit its 2028 financial targets this year, and this quarter’s momentum pushed the outlook higher again.

Why the stock fell anyway

Shares fell despite the beat-and-raise. Bloomberg’s take: sales are being held back by a continued lack of components. The Reuters framing was blunter — “HPE raises forecasts on AI demand, shares fall on supply concerns.”

The bottleneck, CFO Marie Myers told Reuters in a post-earnings interview, is memory. DRAM first, then NAND. And it’s not a new problem: this is the same memory shortage that Lenovo warned in February would last a year and drag down the whole hardware ecosystem, the same one that pushed HPE to raise server and storage prices multiple times last November and December, and the same one CEO Antonio Neri has been managing through multiyear supply agreements with memory partners since the spring. What changed this quarter is that demand acceleration is now colliding with that ceiling at full speed. On the earnings call, Neri acknowledged supply constraints limited revenue conversion in Q3, while expressing confidence in improvement through Q4 and into FY27.

There’s also a subtler market mechanic at work. HPE shares closed at $50.87 on September 2 — up roughly 113% in 2026 before this report, per IndMoney’s preview. When a stock has more than doubled in eight months, “record everything” is already the price of admission. Anything short of a clean supply narrative gives investors a reason to take profits on the margins.

The Juniper effect

The quarter’s growth is not purely an AI-server story. HPE’s acquisition of Juniper Networks closed in mid-2025, and the networking segment has been compounding ever since: revenue of $2.89 billion in Q3, up 74.9% year-over-year on a reported basis as the Juniper business was absorbed. Back in March, Neri argued networking is “more insulated” from the memory supply squeeze than the server business — a claim this quarter’s results support. Networking gear doesn’t carry the same DRAM-per-rack density as AI servers, and Juniper’s routing and switching portfolio now gives HPE a high-margin, memory-light growth engine exactly when its memory-heavy businesses are supply-capped.

That combination — AI servers capped by memory, networking uncapped — explains HPE’s segment arithmetic this year. It also explains why the Juniper deal, which took years of antitrust wrestling to close, is suddenly looking like the best-timed acquisition in the company’s history.

The Oracle deal

One more detail from the earnings call deserves attention: Neri confirmed a deal involving Oracle, in which HPE’s infrastructure underpins a major AI cloud buildout. Details remain limited, but the mention matters. The AI infrastructure market is splitting between hyperscalers building their own and neoclouds renting by the hour — and HPE, which owns no cloud, is positioned as the arms dealer to both. An Oracle partnership plants a flag with the largest of the AI-cloud builders outside the Big Three.

Why this matters beyond HPE

Three takeaways for anyone watching the AI infrastructure buildout.

Demand is not the constraint — memory is. Every AI infrastructure vendor is now singing from the same songbook: orders accelerating, backlog at records, revenue conversion gated by DRAM and NAND. HPE’s quarter is another data point that the 2026 memory shortage — driven by AI factories absorbing supply and SK Hynix’s new fabs not landing until 2027 — has become the binding constraint on the entire AI hardware economy. Watch memory vendors’ capex and allocation decisions; they are now the true throttle on AI compute deployment.

Backlog is the new revenue. HPE booked $2.4 billion in AI orders against $1.6 billion recognized — a 1.5x book-to-bill on AI systems. When supply normalizes, that backlog converts. The market’s impatience with HPE’s stock is really a bet on when, not if.

The whole supply chain reprices together. HPE has already pushed through multiple server and storage price increases, and rivals face the same math. For AI buyers, the era of falling per-unit infrastructure costs is paused; for memory makers, it’s a seller’s market with multiyear agreements locking in allocation. The margin pool in AI hardware is migrating upstream, from systems integrators to component suppliers — at least until the memory crunch breaks.

Bottom line

HPE’s fiscal Q3 2026 was a record quarter wearing a caution tag. Revenue up 34% to $12.2 billion, EPS nearly 20% above consensus, AI orders up 30% sequentially, guidance raised across the board — and a stock that closed lower because the CFO said the quiet part out loud: memory is the bottleneck, and it will stay that way through Q4. The AI buildout is no longer limited by demand or capital. It’s limited by DRAM.