HPE's $1.2 billion Vultr order proves AI demand, and shows how thin it arrives

The first AMD Helios order proves AI demand is real, but HPE's own forecast says the AI mix that order represents will squeeze fiscal-Q4 margins. The December print decides whether the stock's 139% run was paying for demand or for cash.

Richard TangRichard Tang · 3 min read
Share
Antonio Neri, chief executive of Hewlett Packard Enterprise
1 / 6Slide 1 of 6
HPE chief executive Antonio Neri, whose company will build AMD's Helios racks for Vultr under a $1.2 billion order.

Hewlett Packard Enterprise+3.86% — Hewlett Packard Enterprise, up 3.86 percent today just became the launch builder for AMD's+2.66% — AMD, up 2.66 percent today answer to Nvidia+2.42% — Nvidia, up 2.42 percent today, and its own management says that business will compress its margins. On September 30 HPE announced a $1.2 billion order from Vultr, the largest privately held cloud infrastructure company, for its new AMD Helios AI racks, the first order anyone has placed for the system, to be deployed in Vultr's US data centers 1. The deal is a measure of how real AI demand is, and, read against HPE's own guidance, of how little of it HPE keeps.

The demand is big, and the margin on it is thin

Each Helios rack integrates 72 AMD Instinct MI455X GPUs, EPYC "Venice" CPUs, Pensando networking and ROCm software, built for trillion-parameter training and high-volume inference, with six HPE Juniper QFX5252 scale-up Ethernet switch trays connecting all 72 GPUs 1. HPE sells the complete rack, not parts 1. The order extends a Vultr and Juniper relationship that predates HPE's acquisition of Juniper 1, and Vultr's CEO said demand for AI infrastructure "continues to outpace available capacity" 1.

The scale matters to the ledger: $1.2 billion is equivalent to nearly a tenth of the $12.21 billion HPE booked in its July quarter 2. But management's September 2 call is the constraint 3. It expects gross margin to move back toward historical levels as AI systems grow, and operating margin to fall in fiscal Q4 on a higher AI-systems mix and on pricing; the Vultr build is that mix 3. The growth arrives thin, by the seller's own telling.

The cash engine behind the thin margin is real

That is why the tape has paid: HPE is up nearly 139% this year, trailing revenue reached $41.9 billion, and trailing free cash flow hit $4.16 billion against $225 million the year before 3. Management raised its fiscal 2027 free cash flow target to at least $5 billion 3. Barclays+2.81% — Barclays, up 2.81 percent today answered the run on October 14 by cutting its target to $77 from $83 and its rating to Equal Weight, a call on valuation rather than on estimates 2.

Free cash flow jumped eighteenfold, and the FY27 target raises it again

  • Estimate
012345TTM Jul 20250.23TTM Jul 20264.16FY27 target5
Data
Value
TTM Jul 20250.23
TTM Jul 20264.16
FY27 target (estimate)5
Free cash flow, trailing twelve months through July of each year, in billions of dollars; fiscal 2027 figure is management's raised target, marked as guidance. Source: Trefis, October 23, 2026.3

December decides what the run was paying for

So watch December. The fiscal Q4 print, expected that month, is the first full reading on margin at this AI mix and on the record fiscal Q3 order book, when HPE booked more orders than in any prior quarter 3. AMD reports on November 3 first, and that print is the demand test for the Helios ecosystem HPE assembles for 4.

That overlap is the trade's unresolved question: management's guidance says the AI mix that Vultr's order represents compresses margins, so the December numbers decide whether the market has been paying for demand or for cash. Both readings are live until then. Vultr's order confirms the demand; the margin line confirms who bears the cost of making it real.

Get Data Center by email, free

Who is building AI's data centers, and who is paying for them. In our daily email, every weekday.

Loading AMD

Reader comments

0 comments

    More in Data Center

    All in this topic →

    Data Center

    Murati's Thinking Machines just took on Firmus's delivery risk

    Mira Murati's lab signed its models to Firmus's Asian AI factories mid-fundraise. The contract is real, but so is the US$3 billion Firmus must raise first, and the OpenAI termination clause shows what a delay would cost.

    Data Center

    Nebius's NEOM demand receipt is a preliminary agreement sitting under a one-year loan

    The Saudi framework Nebius signed at GTC Berlin is billed as proof of international demand, but a preliminary agreement for capacity that arrives in 2027 does not collateralize a loan due in 364 days.

    Data Center

    Jane Street paid $109 a share for CoreWeave now near $54, with $6 billion of rented compute on the line

    The quant trading firm paid $109 a CoreWeave share in April and now marks it at about $54. Its $6 billion of rented AI compute depends on a landlord whose interest bill is growing faster than its revenue.

    Data Center

    Comfort Systems beat everything, refused the AI labs, and the stock fell 9%

    Comfort Systems beat every estimate and booked a record $17.8 billion backlog, yet CFO Bill George told investors the company has not reached terms it would accept from frontier AI labs and colocation providers. The stock fell 9 percent anyway, and investors must judge discipline against lost growth.

    Data Center

    A $90 million Korean filing says Sanmina's AI racks are still selling. The tape says otherwise.

    A Korean supplier's mandatory filing shows a third straight year of data-center orders flowing to Sanmina's India operations, worth about $90 million, even as the stock shed 30.4% in four weeks. Both camps get tested at the November 9 fiscal-year report.

    Data Center

    CoreWeave's $104.2 billion backlog depends on 2.7 gigawatts it has not switched on

    CoreWeave's $104.2 billion revenue backlog rests on 4.2 gigawatts of contracted power, of which only 1.5 gigawatts is switched on. The November earnings report will show whether the backlog is a contract or a construction schedule.

    Get the newsletter

    The day's AI and market news, in your inbox every weekday.