HPE's Record Quarter Proves the AI Turn — the Discount Has Mostly Closed

Hewlett Packard Enterprise reported its third fiscal quarter on September 2 with the kind of numbers that used to be reserved for software companies. Revenue hit a record $12.2 billion, up 34% from a year earlier. Adjusted profit (what the company calls non-GAAP, or earnings stripped of one-time items like stock compensation and deal charges) came in at $1.11 a share — roughly 20% above what analysts expected and well past its own prior forecast of $0.88 to $0.93. It raised its targets for the rest of the year and for next year in one move.

Here is the part that matters for whether you should care: the stock, already up about 145% for the year and sitting near its 52-week high, briefly fell in after-hours trading right after that beat. That wobble is not a signal to chase or to fade. It is a clue to the real question a value investor should ask, and it is not "is the AI story real?" The answer to that is already provable. The question is "what has the price already paid for the answer?"

The beat is real, and it is in the margins

The honest news in this report is not the revenue — big revenue is table stakes now that every hardware maker has an AI order book. The news is the profit it produced. Non-GAAP operating margin, the share of each revenue dollar that falls to the operating-profit line, nearly doubled to 16.2% from 8.5% a year earlier. Put beside the prior year this is not a cyclical blip. Across fiscal 2025, HPE was a low-margin, roughly break-even business on a reported basis; this quarter it is clearly profitable again and growing. That is an earnings-power rewire, not a one-off.

The demand behind it is provable, not narrative, and the company's own operating metrics show it. HPE booked $2.4 billion in AI system orders in the quarter, up more than 30% from the quarter before, and its AI backlog rose 14% to a record. Networking orders grew 36% — about 3.5 times faster than networking revenue — which is the classic signature of demand outstripping what a factory can ship. Management attributed the pull to a shift from one-off AI pilots to what it calls "agentic AI" and inference: workloads that run continuously on-premises, which pull not just GPU boxes but traditional servers, storage, and networking. It also cited a multi-billion-dollar traditional server deal with a hyperscaler for inference, and an expanded gigawatt-scale networking deployment with Oracle. Whether you call it "agentic" or not, the orders and the backlog are on the record.

Why it can't ship the demand — and why that cuts both ways

The single most important sentence in the call is also the most important risk. HPE's constraint is not demand; it is supply. Chief Executive Antonio Neri named DDR5 memory and NAND flash as the tightest points, with the shortage expected to run into 2027. The deeper cause is that high-bandwidth memory for GPUs is consuming the wafer capacity that would otherwise go to the ordinary memory in standard servers.

Read that twice, because it is the whole investment in two directions at once. On one side, a supplier that cannot ship as fast as customers want gets pricing power — traditional server revenue grew 35%, in part because average selling prices rose as HPE pushed up prices to offset memory costs — and it builds a real order book that converts to future revenue. On the other side, HPE does not control that supply. It has doubled its networking purchase commitments and locked in multi-year agreements, and it has let inventory climb to $11.8 billion to feed the backlog, but the bottleneck is a commodity made by someone else. The same shortage that is boosting margins this quarter is the reason the upside is capped until 2027 — and, when the memory market eventually loosens, the pricing benefit fades with it.

The debt you see isn't the debt you should worry about

A number that looks alarming in a screener is not the right number here. Data feeds show HPE carrying tens of billions of dollars of total debt, which reads like a leveraged hardware company in trouble. It is not. A large slice of that debt sits in HPE Financial Services, the captive lending arm that finances customers' equipment purchases. At the end of the second quarter, HPE reported total net debt of about $16 billion, of which roughly $10.5 billion belonged to that finance company — and that borrowing is matched by the equipment loans it owns, not floating against the business. Corporate net debt is a small fraction of the headline.

That matters for a value investor because it removes the main structural objection to the stock: leverage. HPE Financial Services exited the quarter at a net leverage ratio of 1.8x, below its own 2.0x target, and management said it would use proceeds from the H3C sale to retire more notes. The dividend is modest — about $0.1425 a quarter, a roughly 1% yield — but the bigger commitment is that the company plans to return at least 75% of free cash flow to shareholders, and it raised the fiscal 2027 free-cash-flow floor to $5 billion.

The gap has mostly closed

So the business is genuinely better. That is the real story, and it is earned, not hype. But a value investor's job is to measure the gap between price and provable value — and that gap has closed most of the way.

At roughly $59 a share and a market value near $78 billion, HPE trades at about 29 times trailing earnings and about 15 times trailing EV/EBITDA. Set against the forward numbers the company itself just raised, the multiple is doing real work: about 16x the $5 billion fiscal 2027 free-cash-flow floor, and roughly 13x the midpoint of its $4.40 to $4.60 fiscal 2027 earnings forecast. The stock has gone from a depressed hardware name near its $19.84 low — where even mediocre earnings produced a bargain multiple — to a price that assumes the earnings power holds and keeps compounding.

That is not "too expensive" in the way a broken company is too expensive. The AI infrastructure tailwind is real, the order book is real, and the balance sheet at the corporate level is conservative. But the deep-value version of HPE, the one you could buy on asset value with a floor under it, is largely gone. What is left is a quality business that has been re-rated on growth, and it only works at this price if the fiscal 2027 free cash flow actually reaches the $5 billion the company has guided to, and if the AI buildout does not pause. The decisive test is that number — the $5 billion floor — not the direction of the stock. If HPE delivers it, the re-rating is backed by cash the business actually generates, and the price is fair rather than cheap. If the memory shortage eases and the AI order surge normalizes before that cash shows up, a multiple this far above its historical range has little asset value beneath it to fall back on.