Vultr's third AMD order adds a 20 MW Chennai cluster before AMD's November 3 report

Vultr's 20 MW Chennai supercluster is the third Instinct deployment from AMD's most repeat named customer, following a $1.2 billion Helios rack order through HPE. Ahead of AMD's November 3 report, the question is whether megawatts become audited revenue.

Vincent JiangVincent Jiang · 2 min read
Share
AMD chief executive Lisa Su wearing the National Medal of Technology and Innovation, standing between Michael Dell and Elon Musk at the White House
1 / 5Slide 1 of 5
AMD chief executive Lisa Su at the White House "New Golden Age" summit on 8 October 2026, where she received the National Medal of Technology and Innovation.

Chennai flips the switch on AMD's next leg

Vultr, the world's largest privately held cloud provider, will deploy AMD+2.53% — AMD, up 2.53 percent today Instinct MI355X GPUs in a new 20 MW supercluster in Chennai, the largest such AMD deployment in India 1. Vultr boss J.J. Kardwell pitches it plainly: enterprises need in-country compute for sovereignty, residency and latency 1. That is a real order form, signed with megawatts.

One customer, three escalating orders

The pattern is what investors should price, not the press release. Vultr now runs two of the world's largest AMD AI deployments, Chennai and its Springfield, Ohio flagship 1. Each MI355X carries 288GB of HBM3E memory and 8TB/s of bandwidth per GPU 14. In September HPE+3.69% — HPE, up 3.69 percent today booked a $1.2 billion Vultr order, its first, for AMD Helios racks: 72 MI455X GPUs, EPYC "Venice" CPUs and Pensando networking per rack 25. In July Vultr opened Helios pre-orders spanning 2027 and 2028 3.

AMD's revenue is accelerating, not decelerating, into the Helios ramp

  • Revenue ($B)
  • YoY growth (%)
0204060Q3 '24Q4 '24Q1 '25Q2 '25Q3 '25Q4 '25Q1 '26Q2 '2650.1
Data
Revenue ($B)YoY growth (%)
Q3 '246.8217.6
Q4 '247.6624.2
Q1 '257.4435.9
Q2 '257.6931.7
Q3 '259.2535.6
Q4 '2510.2734.1
Q1 '2610.2537.8
Q2 '2611.5450.1
Total quarterly revenue in billions of US dollars, alongside year-over-year growth in percent. Source: AMD quarterly figures from SEC filings.7

The money behind it is AMD's fastest-growing line. Second-quarter revenue hit $11.54 billion, up 50%, with Data Center revenue up 107% to $6.72 billion and third-quarter guidance near $13 billion 6. Vultr is not OpenAI$500B — OpenAI, private, latest valuation $500B scale, but it is the repeat buyer AMD can name, one who has bought MI300X, MI325X and now MI355X through the roadmap 1.

The receipts are measured in megawatts, not dollars

Here is the complication: Chennai carries no disclosed dollar value, and the announcement is AMD's and Vultr's own words echoed through trade press 14. Only the HPE order carries a price 2.

Meanwhile the stock already trades at 141 times trailing earnings after shedding nearly 19% in a month, a price that assumes promises ship 6. The counterpoint cuts the other way too: bulls note OpenAI, Meta+0.08% — Meta, up 0.08 percent today and Anthropic$350B — Anthropic, private, latest valuation $350B commitments of up to 14 GW, and management expects Data Center revenue to more than double in 2027 6. Vultr's string of orders suggests breadth beyond three whale names.

Watch the November 3 print, not the megawatt headline

AMD reports November 3 4. The tell is not Chennai's ceremony; it is whether Helios shipments appear as revenue and whether the fourth-quarter outlook meets the $16.11 billion consensus 6. Named customers buying three generations running is demand arriving. Conversion to audited revenue is the only scoreboard that pays shareholders.

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.