Hewlett Packard Enterprise shares were on track to open at a fresh record high on Friday, September 4, 2026, after the company put a new Nvidia-powered server at the center of its pitch to enterprise customers racing to deploy autonomous AI agents. HPE stock rose more than 7% in premarket trading, according to Stocktwits, as investors digested both a strong fiscal third-quarter earnings report and a fresh reminder that Nvidia’s newest CPU architecture, not just its GPUs, is now central to HPE’s growth story.
The catalyst combines two threads that had already been building for months: HPE’s September 2 earnings beat, which showed AI systems orders climbing faster than revenue, and a server platform built around Nvidia’s Vera CPU that HPE and Nvidia have been promoting since it was first shown on stage in Taipei in June. Nvidia founder and CEO Jensen Huang summed up the pitch in a line that outlets including Stocktwits have now repeated across market coverage: “Agentic AI has arrived, and it needs a new CPU.” For HPE, a company that spent much of the last decade being written off as a legacy enterprise hardware vendor, the moment amounts to a bet that agentic AI, not generative AI chatbots, is the next multi-year hardware upgrade cycle in the broader AI chip market.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
Key Facts at a Glance
The numbers below summarize the announcement and the earnings report behind it; sourcing detail for each figure is in the sections that follow.
| Metric | Figure |
|---|---|
| HPE premarket stock move | More than 7%, September 4, 2026 |
| New server | HPE ProLiant Compute DL394 Gen12 |
| New CPU | Nvidia Vera |
| General availability | Fall 2026 |
| Fiscal Q3 2026 earnings date | September 2, 2026 |
| Q3 FY26 revenue | ~$12.2 billion (up ~33-34% YoY) |
| Q3 FY26 non-GAAP EPS | ~$1.11 (up ~152% YoY) |
| AI systems backlog | ~$6.3 billion to ~$7.6 billion (varies by source) |
HPE Stock Heads Toward a Record High
The HPE stock move did not come out of nowhere. Shares had already climbed sharply through the summer, touching a 52-week high near $58.79 in mid-August before pulling back ahead of the company’s fiscal Q3 2026 earnings report. HPE stock had gained roughly 159% year-to-date as of that mid-August high, with a 12-month gain of about 189%, according to a market recap covering the period. Morgan Stanley had assigned the stock an Overweight rating with a $69 price target, implying analysts still saw room to run even after the run-up.
The path since then has been choppier than a straight line up. HPE reported quarterly results on September 2, 2026, and the stock initially slipped, trading near $51.96 that day according to a MarketBeat earnings summary, before firming to roughly $52.83 the next session. It was two days later, on September 4, that the combination of digested earnings detail and renewed attention to the Nvidia Vera CPU server pushed HPE stock back toward record territory, with the premarket move of more than 7% reported by Stocktwits. The whipsaw is a reminder that even a well-received earnings beat can take a news cycle or two to translate into a sustained stock reaction, especially when the more durable story, in this case a multi-year AI infrastructure buildout, needs time to sink in with investors, a dynamic also visible in how chip announcements have moved rival stocks in recent months.
What HPE and Nvidia Actually Announced
The specific product driving the headlines is the HPE ProLiant Compute DL394 Gen12, described by HPE as a compute-optimized foundation for agentic AI and high-performance data processing. The server was first unveiled on a Taipei stage on June 1, 2026, positioned as part of the broader “Nvidia AI Computing by HPE” portfolio and folded into HPE’s existing Private Cloud AI and AI Factory offerings. HPE has said the server line is slated for general availability in fall 2026.
What makes the DL394 Gen12 notable is not another GPU refresh. It is built around Nvidia’s Vera CPU, a chip Nvidia has framed as purpose-built for what it calls the “agent loop”: the constant cycle of tool calls, orchestration decisions, and real-time data handling that autonomous AI agents run through thousands of times a minute. That framing is a meaningful shift in how Nvidia talks about its own hardware roadmap. For years, the company’s data center pitch centered almost entirely on GPU throughput for training and inference. Positioning a CPU as the bottleneck-breaker for agentic workloads signals that Nvidia sees the control plane around AI models, not just the models themselves, as the next place customers will need to spend, a shift that has already shown up in other hyperscaler orders tied to the same Vera architecture.
Inside the ProLiant Compute DL394 Gen12
HPE has not published a full traditional spec sheet for the DL394 Gen12 covering socket count, memory ceilings, and I/O lanes in the way it typically does for established ProLiant lines. What the company and Nvidia have emphasized instead is the workload the server targets: deterministic, low-latency handling of massive numbers of concurrent AI agents, rather than raw floating-point throughput. That is a deliberate contrast with HPE’s GPU-heavy Blackwell and Rubin-based AI Factory systems, which remain the company’s answer for large-scale model training and inference.
In practice, HPE is now selling two distinct flavors of AI infrastructure under one roof: GPU-dense systems for building and running large models, and Vera CPU-based systems like the DL394 Gen12 for orchestrating what those models actually do once they are deployed as autonomous agents inside a business. Enterprise buyers evaluating AI systems purchases in late 2026 will increasingly need to think about both halves of that equation, since a company that trains a capable model but cannot orchestrate thousands of simultaneous agent tool calls efficiently will bottleneck on the CPU side long before it runs out of GPU capacity.
Why an “Agent Loop” Needs Different Silicon
To understand why Nvidia is pitching a CPU rather than a GPU as the answer to agentic AI, it helps to look at what an autonomous agent actually spends its time doing. A single customer-support agent, coding assistant, or supply-chain agent does not run one big inference pass and stop. It loops: it calls a tool, waits on a response, updates its own state, decides on the next action, and calls another tool, often dozens or hundreds of times to complete one task. Multiply that by thousands of agents running concurrently across an enterprise, and the bottleneck shifts away from raw model math and toward orchestration, scheduling, and low-latency state management, which is classic CPU territory rather than GPU territory.
A simplified version of that loop looks roughly like this, and illustrates why HPE and Nvidia argue the orchestration layer needs its own dedicated hardware rather than borrowing cycles from a general-purpose server CPU designed for other jobs:
while task_incomplete:
decision = agent.plan_next_action(context)
result = tool_registry.call(decision.tool, decision.args)
context = agent.update_state(context, result)
if agent.is_done(context):
breakEvery iteration of that loop involves scheduling, memory access, and coordination overhead that adds up fast at enterprise scale. HPE and Nvidia’s argument is that a CPU tuned specifically for this pattern, rather than a repurposed general-purpose chip, cuts the latency and cost of running that loop across thousands of agents at once.
Jensen Huang’s Quote and the Market’s Reaction
The line that has traveled furthest through market coverage is Huang’s: “Agentic AI has arrived, and it needs a new CPU.” It is a short, quotable summary of Nvidia’s broader strategic pivot, and it has been repeated across financial outlets including Stocktwits as shorthand for why HPE’s stock reaction has outsized meaning for the wider AI infrastructure trade. Coming from the head of the company that has defined the AI GPU market for most of the current decade, the statement functions as validation that CPU architecture, an area many investors assumed was largely settled, is still being actively reinvented for AI-specific workloads.
For HPE specifically, the quote does double duty. It reinforces Nvidia’s endorsement of HPE as a launch partner for a genuinely new chip category, at a moment when HPE is trying to prove to Wall Street that its AI hardware business is more than just reselling GPUs bought from Nvidia. Being first to market with a Vera CPU-based server gives HPE a differentiation story against rivals selling largely similar GPU-based systems.
HPE’s Q3 Fiscal 2026 Earnings, By the Numbers
The earnings backdrop matters as much as the server announcement itself. HPE reported results for its fiscal third quarter, ended July 31, 2026, on September 2, 2026, detailed in materials posted to its investor relations site. Total revenue came in around $12.2 billion, up roughly 33% to 34% year over year, with non-GAAP earnings per share of about $1.11, up 152% year over year and ahead of consensus estimates, according to earnings summaries from outlets including MarketBeat. Free cash flow for the quarter was approximately $958 million.
The segment breakdown shows where the growth is concentrated. HPE’s Cloud & AI segment, which includes both servers and storage, generated about $9.0 billion in revenue, up roughly 25% year over year, with an operating margin around 17%. Within that segment, server revenue reached about $6.8 billion, up roughly 35% year over year, while storage revenue was approximately $1.3 billion, up about 10%. Networking, the segment most reshaped by HPE’s acquisition of Juniper Networks, posted revenue of about $2.9 billion, up roughly 75% year over year, with an operating margin near 22%.
| HPE Segment (Q3 FY26, ended July 31, 2026) | Revenue | YoY Growth | Operating Margin |
|---|---|---|---|
| Total company revenue | ~$12.2 billion | ~33-34% | — |
| Cloud & AI segment | ~$9.0 billion | ~25% | ~17% |
| Server revenue (within Cloud & AI) | ~$6.8 billion | ~35% | — |
| Storage revenue (within Cloud & AI) | ~$1.3 billion | ~10% | — |
| Networking segment | ~$2.9 billion | ~75% | ~22% |
| AI systems revenue | ~$1.6 billion | — | — |
The AI Backlog Figure Everyone Is Citing
Beyond the headline revenue figures, the number getting the most attention from analysts is HPE’s AI systems order backlog. AI systems orders came in at approximately $2.4 billion for the quarter, up more than 30% sequentially, while AI systems backlog itself grew roughly 14% sequentially. Reported totals for the overall AI backlog vary somewhat depending on the source and exactly what is included: one breakdown put total AI backlog at more than $6.3 billion, including $5.9 billion specifically in AI systems backlog, while another report cited a total AI backlog figure closer to $7.6 billion. Both figures describe the same underlying trend, a record backlog that is growing faster than reported revenue, which is generally read as a sign of durable rather than one-off demand.
HPE CEO Antonio Neri has framed that gap between orders and revenue as a function of supply constraints rather than softening demand, pointing to bottlenecks in areas like DDR5 memory and wafer capacity as reasons the company cannot ship its full order book immediately. That framing matters for how investors read the AI backlog number: a backlog that is growing because HPE cannot build fast enough is a different story, and arguably a more bullish one, than a backlog growing because customers are stacking speculative orders they may later cancel.
HPE vs. Dell, Supermicro, Lenovo and Nvidia
HPE’s AI systems revenue of roughly $1.6 billion for the quarter is real growth, but it remains modest next to the scale some rivals are now reporting. Dell Technologies disclosed AI-optimized server revenue of $16.40 billion in its most recently reported quarter, up 100% year over year, according to CNBC and a MarketBeat earnings summary. Super Micro Computer reported total revenue of $11.12 billion for its fiscal Q4 2026, up 93% year over year, with AI solutions representing roughly 60% of that total according to a Yahoo Finance recap of the earnings call. Lenovo’s Infrastructure Solutions Group posted $8.51 billion in revenue for its most recently reported quarter, up 98% year over year, well ahead of analyst estimates, while the company separately disclosed an AI server pipeline of $54 billion, up 157% sequentially, according to Investing.com.
Nvidia itself, sitting above all of these hardware vendors as the chip supplier, reported Data Center revenue of $89.0 billion for its most recently reported quarter, up 117% year over year, according to Nvidia’s own results and CNBC’s coverage of the report. That scale gap explains why HPE’s move to secure early access to a genuinely new Nvidia chip category, rather than simply reselling the same GPUs every competitor can also buy, matters disproportionately to HPE’s investment case relative to its current AI revenue base, even as Nvidia keeps widening its silicon ecosystem through moves like its MediaTek investment tied to NVLink Fusion.
| Company | Most Recently Reported Quarter | AI-Related Revenue / Metric | YoY Growth |
|---|---|---|---|
| HPE | Q3 FY26 (ended July 31, 2026) | ~$1.6B AI systems revenue | Backlog +14% seq. |
| Dell Technologies | Most recent reported quarter | $16.40B AI-optimized server revenue | +100% |
| Super Micro Computer | Fiscal Q4 2026 | $11.12B total revenue (~60% AI) | +93% |
| Lenovo | Most recent reported quarter | $8.51B Infrastructure Solutions Group revenue | +98% |
| Nvidia | Most recent reported quarter | $89.0B Data Center revenue | +117% |
The Juniper Networks Payoff
HPE’s networking growth is inseparable from its 2025 acquisition of Juniper Networks, a deal announced in January 2024 at a value of $14 billion and completed on July 2, 2025. HPE has positioned Juniper’s AI-native, cloud-native networking portfolio as the connective tissue for its AI Factory systems, arguing that the network fabric linking thousands of GPUs and CPUs together is as important to AI performance as the compute itself. The 75% year-over-year growth in HPE’s networking segment during Q3 FY26, alongside orders growing 36%, roughly 3.5 times the pace of networking revenue itself, suggests that integration is translating into real order flow rather than remaining a paper synergy.
That matters for the HPE stock story because it diversifies the AI narrative beyond a single product line. A company whose AI growth depends entirely on one server family is more exposed to a single bad product cycle. HPE’s pitch now spans GPU-dense AI Factory systems, the new Vera CPU-based DL394 Gen12, and Juniper-powered networking gear, giving the company three separate levers tied to the same underlying AI infrastructure buildout.
From Printers and PCs to AI Factories
It is worth remembering how recently HPE was seen as a legacy business. The company split from HP Inc. in 2015, keeping the enterprise servers, storage, and networking business while HP Inc. retained PCs and printers. For years afterward, HPE traded as a slow-growth value stock, associated more with steady dividends than with any AI upside. The stock’s roughly 159% year-to-date gain through mid-August 2026 represents a dramatic re-rating from that identity, driven almost entirely by the market deciding HPE has a credible seat at the AI infrastructure table alongside Dell, Supermicro, and Lenovo.
The GreenLake platform has been a quieter part of that transition. Originally built as HPE’s consumption-based, as-a-service alternative to buying hardware outright, GreenLake now serves as the delivery layer for much of HPE’s cloud, compute, storage, and AI workloads, including on-premises and edge deployments. That matters for enterprise customers who want AI infrastructure without the capital outlay of buying servers outright, and it gives HPE a recurring-revenue wrapper around what would otherwise be a purely transactional hardware sale. Market research firms such as IDC have tracked enterprise server spending shifting toward AI-optimized systems for several consecutive quarters, a broader trend HPE’s results are riding rather than creating on its own.
What This Means for Enterprise IT Buyers
For IT decision-makers evaluating infrastructure purchases heading into 2027 budget cycles, the HPE-Nvidia announcement adds a new variable to procurement conversations that used to be almost entirely about GPU allocation. Buyers deploying agentic AI systems, customer service agents, coding assistants, autonomous research tools, will now need to factor in CPU-side orchestration capacity alongside GPU-side training and inference capacity when sizing deployments. A company that over-invests in GPU capacity while under-provisioning for agent orchestration risks bottlenecking on the exact problem Nvidia’s Vera CPU is meant to solve.
It also raises the stakes for infrastructure vendor selection. HPE’s early access to Vera CPU-based systems gives it a temporary window to differentiate from Dell, Supermicro, and Lenovo, all of which are expected to eventually offer competing Vera-based platforms once Nvidia’s chip becomes more broadly available, much as traditional server CPU lines from Intel have found themselves competing for attention against Nvidia’s AI-first silicon. Enterprise buyers with long procurement cycles will need to weigh the benefit of adopting early against the near-certainty that pricing and configuration options will broaden as more vendors bring Vera-based servers to market through fall 2026 and into 2027.
The Case for Caution
Not every part of the HPE stock story is unambiguously bullish. The gap between AI systems orders and AI systems revenue, while framed by CEO Antonio Neri as a supply-side constraint rather than a demand problem, is also the kind of gap that widens the room for disappointment if supply catches up faster than expected and backlog growth decelerates. A backlog that has been treated by the market as evidence of durable demand can just as easily be read, if growth slows even slightly, as evidence that the AI infrastructure buildout is closer to fully priced in than investors assumed.
There is also the question of how much of HPE’s stock reaction is genuinely about the Vera CPU server versus a broader momentum trade in AI infrastructure names generally. HPE stock’s initial dip immediately after the September 2 earnings release, before rebounding into the September 4 rally, suggests the market needed a second look, and a fresh news hook around the Nvidia partnership, before fully re-rating the stock. That pattern, a beat that does not immediately move the stock followed by a delayed rally on secondary news, is not unusual, but it does suggest some of the move is sentiment-driven rather than purely mechanical repricing off the earnings numbers themselves.
Five Predictions for HPE and the Agentic AI Server Market
- Vera CPU competition arrives fast. Expect Dell, Supermicro, and Lenovo to announce their own Vera CPU-based server lines within two to three quarters of HPE’s fall 2026 general availability window, narrowing HPE’s first-mover advantage.
- AI backlog stays the headline metric. HPE’s AI systems backlog, not quarterly revenue alone, will remain the number analysts scrutinize most closely each quarter as a proxy for whether the agentic AI buildout has staying power.
- Supply constraints persist into 2027. DDR5 memory and wafer capacity bottlenecks cited by CEO Antonio Neri are unlikely to fully clear before mid-2027, keeping the gap between AI orders and AI revenue elevated for several more quarters.
- Networking becomes a bigger swing factor. As Juniper Networks integration matures, expect HPE’s networking segment growth rate to increasingly influence overall HPE stock sentiment alongside the AI systems numbers.
- Volatility around earnings continues. Given the pattern of an initial post-earnings dip followed by a delayed rally, expect HPE stock to remain more volatile around quarterly reports than its historical profile as a steady, low-beta enterprise hardware stock.
Frequently Asked Questions
What server did HPE unveil with Nvidia?
HPE unveiled the ProLiant Compute DL394 Gen12, a server built around Nvidia’s Vera CPU and positioned as a compute-optimized foundation for agentic AI and high-performance data processing, with general availability planned for fall 2026.
Why did HPE stock jump toward a record high?
HPE stock rose more than 7% in premarket trading on September 4, 2026, according to Stocktwits, as investors combined a strong fiscal Q3 2026 earnings report released September 2 with renewed attention to the new Nvidia Vera CPU server, reinforcing the view that HPE’s AI infrastructure business has durable momentum.
What is Nvidia’s Vera CPU?
Vera is Nvidia’s CPU architecture described by the company as purpose-built for agentic AI workloads, specifically the repeated cycle of tool calls, orchestration, and real-time data handling that autonomous AI agents run through as they complete tasks, rather than the general-purpose server workloads traditional CPUs are designed to handle.
Who said “agentic AI has arrived, and it needs a new CPU”?
The quote is attributed to Nvidia founder and CEO Jensen Huang, and has been cited across market news coverage, including by Stocktwits, in connection with the HPE ProLiant Compute DL394 Gen12 announcement.
How did HPE perform in its most recent earnings report?
For its fiscal third quarter of 2026, ended July 31, 2026 and reported September 2, 2026, HPE posted revenue of approximately $12.2 billion, up roughly 33% to 34% year over year, with non-GAAP earnings per share of about $1.11, up 152% year over year, according to earnings summaries reported by outlets including MarketBeat.
How big is HPE’s AI systems backlog?
Reports on HPE’s total AI backlog vary by source, with figures ranging from roughly $6.3 billion to $7.6 billion depending on what is included, but all point to a record backlog that grew faster than quarterly revenue, which analysts generally view as a sign of durable demand.
How does HPE compare to Dell and Supermicro in AI servers?
HPE’s AI systems revenue of roughly $1.6 billion for the quarter remains smaller than Dell’s $16.40 billion in AI-optimized server revenue and Super Micro Computer’s $11.12 billion in total quarterly revenue, of which about 60% was AI-related, according to figures reported by CNBC and Yahoo Finance.
What role does the Juniper Networks acquisition play in HPE’s AI strategy?
HPE completed its $14 billion acquisition of Juniper Networks on July 2, 2025, and now positions Juniper’s AI-native networking portfolio as the connective infrastructure linking HPE’s AI Factory and Vera CPU-based systems, a factor behind the networking segment’s roughly 75% year-over-year revenue growth in Q3 FY26.
Related Coverage
- HUMAIN Horizon Ultra AI PC Debuts With 80 TOPS Chip [2026]
- NVIDIA's Local AI Play Undercuts $20 Cloud AI Plans [2026]
- NVIDIA Local AI Push: 24GB VRAM GPUs Get 1.9x Boost [2026]
- Nvidia RTX Spark N1X Launches October: 128GB RAM [2026]
- Acer Predator Atlas 7 Debuts: Intel's 42% AMD Claim [2026]
