Sunday 25 October 2026
Data & AI Infra

Nvidia's full-stack AI factory expands even as supply stays tight

The chipmaker now sells compute, networking, racks and software, and helps finance AI campuses while memory shortages limit its growth.

Nvidia's full-stack AI factory expands even as supply stays tight
Photograph: Nvidia (press image)

On Aug. 26, Nvidia executives told investors that fiscal 2028 revenue should grow about 70%, but the outlook is supply-constrained. CEO Jensen Huang said the shortage runs from high-speed memory and silicon vendors to suppliers of data centers under construction, and CFO Colette Kress said memory prices are rising faster than expected. Supply commitments rose to $279 billion by July 26, up from $119 billion in the prior quarter. Nvidia's revenue for the fiscal second quarter ended in July 2026 was $96.221 billion, up 105.85% from a year earlier, and its data center segment brought in $89.023 billion. A Zacks Investment Research article noted the Vera Rubin platform is in full production.

On Aug. 27, AWS and Nvidia said they will add 2 million more Nvidia GPUs to AWS data centers during 2027 and 2028. The capacity will include Blackwell Ultra, Rubin and Rubin Ultra GPUs. AWS will also add Nvidia RTX PRO 4500 Blackwell Server Edition GPUs to its EC2 G7 instances, expanding Blackwell-based capacity. Nvidia CEO Jensen Huang said demand is running ahead of every forecast. The two companies also plan AI factories for U.S. government workloads, including 100,000 GPUs.

Nvidia sells an AI factory, not just chips

Nvidia management said it sells a full-stack AI factory platform that includes the Vera CPU, Rubin GPU, NVLink, InfiniBand, Ethernet, systems, algorithms, and CUDA software, according to 24/7 Wall St. on Sept. 7. Revenue opportunity per gigawatt has expanded from roughly $18 billion with Hopper to $25 billion with Blackwell and $40 billion with Vera Rubin, according to 24/7 Wall St. on Sept. 7. Nvidia has partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR that are meant to raise more than $500 billion in outside capital for AI infrastructure, according to 24/7 Wall St.. A SoftBank Energy campus in Ohio will host Nvidia compute for OpenAI under 20-year leases.

Financing follows the GPUs

The financing layer became more visible in late September. Crusoe said on Sept. 17 it had a first close for a Series F that is expected to reach $3.9 billion, giving the company a $30.9 billion post-money valuation, with Nvidia among investors. Crusoe reported total contracted value above $140 billion and gross contracted capacity above 6 GW, of which roughly 1 GW is delivered and operating. On Sept. 25, Data Center Frontier reported that Nscale's IPO filing showed four GPU services agreements with Anthropic that could generate about $44.6 billion, but Nscale had not yet secured binding financing for the GPU equipment and data center infrastructure. Reuters reported on Aug. 31 that Anthropic entered a cloud deal worth about $35 billion with Lambda, which Nvidia backs. Data Center Frontier noted that Nvidia is an investor in Lambda and holds the lease on the Hut 8 facility supporting the Anthropic deployment. SoftBank sold $11.1 billion of high-yield debt on Sept. 24, a record corporate bond sale according to Reuters, to back OpenAI and other AI ventures.

As of late September 2026, the story is not about a shortage of demand. Data Center Frontier wrote that large contracts alone cannot guarantee the financing, power, equipment, or construction capacity required to turn projects into operating megawatts. Nvidia now acts not only as a hardware vendor but also as a financial and contractual player in the facilities built around its chips.

Arjun Mehta

Arjun Mehta covers data centres, AI infrastructure and networks for Five Nines. He writes from company announcements, filings and published reporting.

Corrections. Report an error to corrections@thefivenines.com. We list substantive corrections on the corrections page.