NextFin News - Hewlett Packard Enterprise lifted its full-year sales forecast on Wednesday after reporting third-quarter revenue and earnings that beat Wall Street expectations, as artificial-intelligence demand for servers and networking gear pulled the company's growth outlook higher. The company now expects revenue to grow 34% to 37% in the fiscal year ending in October, up from 29% to 33% previously, and raised its adjusted earnings-per-share forecast to $3.75 to $3.85 from $3.35 to $3.45. Shares fell 8% in extended trading despite the beat — a reminder that after a roughly 120% year-to-date rally, the market was pricing in more than a clean quarter.
The Quarter: A Beat Built on AI Infrastructure
HPE reported revenue of $12.21 billion for the July quarter, up 33.6% from a year earlier, ahead of the roughly $11.9 billion analysts expected. Adjusted earnings came in at $1.11 per share, well above the 93 cents consensus. The print extends a streak of execution: HPE has beaten estimates in six of its last seven quarters, and the growth is concentrated in exactly the two businesses that define the AI infrastructure trade — servers and networking.
The momentum has been building for three quarters. In the first quarter of fiscal 2026, revenue rose 18% to $9.3 billion with record adjusted earnings of $0.65 a share. In the second quarter, revenue accelerated to $10.7 billion, up 40%, and the stock jumped more than 35% after hours. Now, with the third quarter at $12.21 billion, the trajectory is not a one-off — it is a compounding curve.
For fiscal 2026, the company projects networking revenue to grow 73% to 74%, and it pushed its fiscal 2027 revenue growth outlook to 13% to 17% from 8% to 12%. Fiscal 2027 adjusted earnings are expected to grow 16% to 20%, up from the 12% to 16% forecast before. The raise is not just about this quarter; it is a two-year claim on demand durability.
"It's clear to us that the market, the tailwinds in AI, particularly the adoption of AI in the enterprise, is really starting to settle in," said Marie Myers, HPE's chief financial officer, in an interview.
The phrasing is deliberate. Not "exploding," not "spiking" — settling in. That is the language of a demand curve management believes has moved from early adoption into a durable deployment phase, the kind that supports multi-year guidance rather than a single strong quarter.
Why HPE's AI Trade Is Not Dell's
The comparison writes itself. A day earlier, Dell Technologies reported fiscal second-quarter results showing AI-optimized server revenue up 757% from a year earlier. Both companies are riding the same wave, but with different hulls — and that difference is where the HPE thesis lives.
Dell's AI server business leans heavily toward hyperscale cloud builders, the customers with the deepest pockets and the most aggressive deployment timelines. HPE has tilted its AI server mix toward enterprise and sovereign customers — large corporations and national governments that want AI infrastructure under their own control. An analyst noted that this enterprise and sovereign mix carries a better profit profile than the hyperscale-heavy book. The logic is straightforward: enterprise and sovereign deals tend to carry better margins and longer contractual visibility than volume-driven hyperscaler business, even when the raw order numbers look smaller.
This is a strategic choice, not an accident of sales cycles. HPE is not trying to win the hyperscale server arms race on volume. It is selling the infrastructure that enterprises and governments need when they decide they do not want to rent all of their AI capacity from a handful of cloud providers. Sovereign AI — the ability for a country or large institution to run its own models on its own hardware, for reasons of data control, regulation, or national security — is a structural demand driver, not a cyclical one. Once a government or bank has committed to an on-premises AI stack, it does not unwind it when the next cloud discount arrives.
The scale of the opportunity is visible in the backlog. HPE entered the third quarter with $5.9 billion in backlog, composed primarily of enterprise and sovereign orders, and cumulative AI systems bookings had reached $16.4 billion by the end of the second quarter. The company has repeatedly said its AI server pipeline remains a multiple of its backlog — meaning the constraint is not customers; it is supply.
The Real Engine: Networking, Not Just Servers
The market's instinct is to treat HPE as an AI server story. That is a misread. The more important number is networking.
In the second quarter, networking revenue grew 148% year over year to $2.69 billion, driven by the July 2025 acquisition of Juniper Networks and by demand for the data-center networking fabric that AI clusters require. Servers are the commodity-adjacent layer of the AI stack — competitive, exposed to component costs, and margin-thin. HPE's Cloud & AI segment, which contains the server business, operated at a 12.4% margin in the second quarter, up from 6.6% a year earlier — improved, but still well below the networking unit's profitability. Networking is the toll road. Every GPU cluster needs switches, routers, and fabric software; once a customer standardizes on a network architecture, switching costs are high and pricing power is higher.
The Juniper integration is the mechanism that converts this from a hardware resale story into a platform story. HPE's guidance assumes the combined networking portfolio — Aruba for campus and edge, Juniper for data center and routing — continues to capture the spend that sits adjacent to every AI deployment. That is why the company can raise its full-year outlook while flagging supply constraints in the server business. It is also why the market should watch networking growth, not server revenue, as the tell on whether HPE is building a durable franchise or simply reselling GPUs.
The company's gross margin tells the same story. Non-GAAP gross margin reached 36.9% in the second quarter, up 750 basis points year over year — a jump that no pure-play server reseller delivers. The mix shift toward networking and software is doing the work that volume alone cannot.
Concrete evidence of networking's reach came Wednesday as well: HPE announced an expansion of its collaboration with Oracle to deploy HPE Juniper Networking equipment across Oracle's AI data centers. That is a design win at hyperscale scale, in the networking layer, even as the server business tilts toward enterprise customers. It is the best of both positions — enterprise relationships for servers, cloud-scale credibility for networking.
Supply Is the Constraint, Not Demand
Myers was explicit about the bottleneck: memory is the main constraint, followed by NAND, CPUs, and drives.
"Demand is far outstripping supply," she said.
That sentence inverts the usual risk frame for a hardware company. The question for HPE is not whether customers will buy; it is whether the company can source enough components to fulfill orders it has already won. HPE has responded by signing longer-term supply agreements to lock in components. That is a rational move in a supply-constrained market, but it carries its own symmetry: long-term component commitments are a lever that amplifies in both directions. If AI demand holds, they secure supply and protect margin. If demand proves cyclical and mean-reverts, they become inventory liabilities.
The broader spending backdrop supports the demand side. Big Tech's AI infrastructure spending is set to exceed $730 billion this year, according to industry analysis. Enterprise IT infrastructure spending is strengthening on AI investment, server refresh cycles, and rising demand for data-center capacity, analysts at Citi have said. In that environment, a company that can actually ship hardware has pricing power — and HPE's supply agreements are the tool that converts scarcity into margin.
The Counter-Thesis: What If the Market Is Right to Sell?
There is a strong case for the 8% after-hours decline, and it deserves more than a dismissive paragraph. The most obvious explanation: HPE's stock had rallied roughly 120% year to date into the print. A 120% advance prices in a great deal of good news, and a beat that is merely "good" rather than "spectacular" can trigger profit-taking even when the fundamentals are intact. The market may have wanted a larger raise, or clearer visibility into server margins, and instead got a raise that confirms what was already expected. After a run of that size, confirmation is often sold.
The deeper bear case is about the cyclical nature of infrastructure spending. Server demand is notoriously lumpy — enterprises pull orders forward when they fear component shortages or tariff-driven cost increases, then work through inventory. If a meaningful portion of HPE's current bookings represents pull-in rather than genuine end demand, fiscal 2027's 13% to 17% growth could prove optimistic. The evidence floor for a cyclical read is real: HPE itself has warned that AI demand and revenue will be "uneven" across quarters, and sovereign customers place large orders with extended lead times that can defer shipments into later periods. The company's own guidance for unevenness is an admission that the order stream is not a smooth line.
There is also the valuation question. At a much higher multiple than it commanded a year ago, HPE is no longer a deep-value turnaround; it is a growth stock, and growth stocks are punished for any deceleration. The strongest version of the bear thesis is not that AI demand is fake — it is that the durability of enterprise and sovereign AI spend has been extrapolated too far, too fast, into a stock price that now requires perfection. A single quarter of order lumpiness, or a guide that merely meets rather than beats, could compress the multiple faster than earnings can grow into it.
The answer to that bear case rests on the mix shift. If HPE's AI order book is genuinely weighted toward enterprise and sovereign customers with multi-year deployment cycles, the lumpiness of hyperscaler capex matters less. The falsifying signal is specific and observable: if HPE's backlog conversion slows materially over the next two quarters while management keeps the fiscal 2027 growth guide unchanged, or if the enterprise-and-sovereign share of cumulative AI orders stops rising, the durability thesis breaks. Watch the backlog-to-revenue conversion ratio and the disclosed mix of AI systems orders. If enterprise and sovereign stop climbing as a share of cumulative bookings, the structural claim is in trouble — and the stock's multiple has nowhere to hide.
What Comes Next: Three Horizons
Short term (sentiment and liquidity): The stock's reaction is a reminder that after a 120% run, the bar is high. Expect volatility around component-supply updates and any sign of order pull-in. The next catalyst is the fiscal fourth-quarter report and whether HPE can sustain the 34% to 37% growth trajectory into the year ending in October. A guide that holds or raises from here would signal that the pull-in fear is overdone; a guide that merely meets could extend the multiple compression.
Medium term (fundamentals): The key metrics are networking growth holding in the 70% range and server margins stabilizing as the longer-term supply agreements bite. Base case: HPE delivers the raised fiscal 2026 guide and grows 13% to 17% in fiscal 2027 as enterprise AI deployments continue to convert from backlog. Upside case: sovereign AI mandates accelerate and networking design wins at cloud builders expand the addressable market beyond the enterprise base, lifting networking growth above the guided range. Downside case: component constraints tighten further, or enterprise capital spending softens, and fiscal 2027 growth lands at the low end of the range or below it.
Long term (structural): The structural question is whether enterprise and sovereign AI infrastructure becomes a permanent layer of corporate IT, or a transitional phase before workloads migrate back to hyperscale clouds. HPE's thesis is that governance, data sovereignty, and the need for on-premises control create a durable hybrid model. If that is right, HPE's networking-centric positioning — not its server volume — is what compounds over the cycle. If it is wrong, HPE is left selling low-margin servers into a hyperscaler-dominated market, and the Juniper bet becomes an expensive detour.
The signals to watch across all three horizons: the quarterly mix of AI systems orders between enterprise/sovereign and hyperscale customers; the pace of backlog conversion into revenue; memory and component pricing; and whether networking revenue growth sustains above 50% year over year as the Juniper integration matures. Each of these is a measurable data point, not a sentiment read.
HPE's raise is a validation of the AI infrastructure trade, but Wednesday's decline is a warning that validation alone no longer moves the price. The company has successfully repositioned itself at the intersection of networking, cloud, and AI. Now it has to prove two things: that enterprise and sovereign demand is structural rather than cyclical, and that its stock can grow into a multiple that already assumes the answer.
Explore more exclusive insights at nextfin.ai.
