AI's Most Under-Appreciated Giant Just Had Its Best Quarter Ever — So Why Did Shares Slip?

The AI story's most under-appreciated giant just posted the best quarter in its history — and the stock slipped anyway. That gap is the story, so let's sit with it. Hewlett Packard Enterprise delivered its first quarter ever with adjusted EPS above $1, on record revenue, record orders, and a raised outlook for this year and next. Wall Street's takeaway? Focus on the supply squeeze and sell — which is exactly what makes HPE interesting right now.

Quick primer, because HPE is easy to overlook and hard to overstate: it builds the servers and network gear that run the AI boom. Think the Nvidia-powered AI systems humming inside data centers, plus — since its roughly $14 billion Juniper Networks acquisition closed last year — the routers, switches, and security gear that stitch data centers together. Its buyers run from enterprises modernizing data centers to cloud providers, sovereign AI programs, and banks.

The numbers, fast, because they're seasoning and not the meal: revenue of $12.2 billion, up 34%, a record. Non-GAAP operating profit of $2.0 billion — about two and a half times a year ago — on a record gross margin of 40.4%. And orders up 42%, more than any quarter in the company's history. That last bit matters most: orders are tomorrow's revenue, and HPE has never booked more of it.

Here's the signal the market is underweighting: AI demand is broadening out. Phase one of this boom was a handful of hyperscalers buying GPUs by the warehouse to train models. Phase two — visible in HPE's order book right now — is everyone else buying AI-ready machines to run agents and inference workloads on their own turf. Traditional server orders jumped 75% year over year as enterprises stocked up on gear to run models where their data lives. That's enterprise AI graduating from pilot projects into a real, recurring business.

Then there's Juniper, which is paying off the way management drew it up. Networking revenue hit $2.9 billion, up 75% as reported, with routing up roughly 270% and data-center networking more than doubled — Juniper has become HPE's growth engine, not a merger integration project. Even on a like-for-like basis the segment grew 10%, while networking orders rose 36% — three and a half times the pace of revenue. Demand keeps outrunning supply. The AI-systems backlog reached $6.8 billion exiting the quarter, a record, up about 14% sequentially; fold in AI networking and the total is $7.6 billion. Cumulative AI-networking orders have already blown past the company's original full-year target, so HPE raised it to $2.5–3.0 billion by year-end.

Two deals show where the demand comes from. Alongside the quarter, HPE expanded its Oracle collaboration into a gigawatt-scale buildout — Oracle is deploying HPE Juniper routers and switches across one of the largest AI cloud infrastructure projects on the planet, multi-gigawatt and multi-year. And after the quarter closed, a hyperscaler handed HPE a deal worth about $3.5 billion to build AI servers for its own internal inferencing. Read that twice: a hyperscaler — the most self-sufficient buyer in tech — decided it was cheaper and faster to buy inference machines from HPE than to build them itself.

So why did shares slip? Supply. HPE can't get parts fast enough. DDR5, DDR4, and NAND memory, plus wafer and clean-room capacity, are the binding constraint, and management is locking in multi-year supply agreements to fix it — purchase commitments for networking components more than doubled quarter over quarter. They also warned the record gross margin will moderate as the crunch eases. Read that again: demand is not the problem here. The problem is the opposite of the problem.

The guidance says management sees the same thing. HPE raised its fiscal 2026 outlook to revenue growth of 34–37% and hiked EPS guidance by 40 cents, then did something rare: it pre-announced a higher fiscal 2027 framework — revenue growth of 13–17%, up from 8–12%, with EPS growth of 16–20% — a year early. None of it stopped the post-earnings dip of roughly 5% so far, as the market fixated on margin moderation instead. Worth remembering while that plays out: this is an AI-infrastructure compounder valued at a fraction of Dell — the market is treating a supply problem like a demand problem.

Bottom line: HPE is the answer to the question of whether the AI trade has room beyond hyperscalers and chipmakers. It does — and the company that builds the plumbing collects tolls on all of it. The bottlenecks here are the good kind: too many orders, not enough wafers. When the market sells that story, patient investors get the durable part of the AI buildout at a discount.

Disclosure: The Signal holds no position in HPE. Positions may change. This is not financial advice.