EXPLORINEER
en

US — TechCrunch

Nvidia clarifies when it will — and won’t —keep financing neoclouds

Chips1 min read

Nvidia's interim CEO, Colette Kress, said the company will be more careful about helping its customers pay for new AI data centers. The company also explained when its loan guarantees start to apply.

What happened

Nvidia has been helping AI companies build data centers. It did this by guaranteeing loans, which means it promises to pay if the borrower cannot. After its founder and CEO Jensen Huang left the job, Nvidia's finance chief, Colette Kress, explained how these guarantees work. She said a guarantee only starts after the customer has the money it needs to build the site and install Nvidia's chips. She also said Nvidia will be much more selective about which companies it helps in the future.

Key numbers

  • Nvidia's net income was $64.9 billion in the third quarter, up from $59.7 billion in the prior quarter.
  • Nvidia guaranteed a $105 billion loan for OpenAI's PORTS-Pike data center campus in Ohio. This was announced in August.
  • There were reports that the guarantee had first been $250 billion before Kress reduced it to $105 billion.
  • A petition signed by 197,000 people asked Nvidia to bring Huang back as CEO.

Why Nvidia was doing this

Nvidia makes the special computer chips that AI systems need. Under Huang, the company invested in AI startups and smaller cloud companies called neoclouds. Some critics said these deals were circular. Nvidia invested in customers, and those customers then used the money to buy Nvidia's products.

Why it matters

Nvidia's profits depend on customers building more data centers. If Nvidia now helps fewer customers, the AI boom could slow down. At the same time, Nvidia is trying to avoid risky deals that could hurt the company if projects fail.

What to watch

  • Which AI companies Nvidia chooses to help with financing in the future.
  • Whether the OpenAI Ohio data center project meets the conditions for the guarantee.
  • How much Nvidia keeps investing in AI startups and neoclouds.
  • Whether Nvidia names a permanent CEO and how that changes its deals.

Glossary

Nvidia
A company that makes computer chips used to run AI systems.
neoclouds
Newer cloud companies that rent out computer power, mostly for AI work.
data center
A large building full of computers that store data and run online services and AI.
OpenAI
An AI company that makes ChatGPT and other AI tools.

Same story, different eyes

Ad

Add 150 words. Learn 10 a day.

Dicter word books send only the words you learn to your review notes.

ChipsThe Register

Chinese web giant Tencent says new US GPU ban ‘sets everyone back in China’

Tencent president Martin Lau said the new US ban on Chinese companies remotely accessing GPUs will hurt in the short term but will ultimately push China to speed up domestic alternatives. He said the ban will affect Tencent's model training and GPU rentals, while made-in-China chips should be able to handle inferencing workloads.

ChipsThe Register

Business as usual, but Huang ain't coming back, says Nvidia

Nvidia's board says it has no plans to bring back Jensen Huang as CEO after he resigned last week, and it is strengthening oversight and searching for a new leader. Interim CEO Colette Kress reported quarterly revenue up 103 percent to $115.8 billion and raised the fiscal 2028 growth outlook to about 75 percent, though analysts questioned the company's future direction.

ChipsZDNet Korea

"No Return for Jensen Huang"…Why Nvidia Turned Its Back on Its Founder

Speculation persists that Jensen Huang could return as Nvidia CEO after he stepped down over an extramarital affair controversy, but Nvidia board chair Harvey Jones said the board unanimously accepted his resignation and has no intention of reinstating him. Analysts point to the company's CEO-centric governance, the unresolved national security concern over the alleged Chinese intelligence link of one of the affair partners, and the ethics violation tied to the other affair involving a junior employee as reasons the board is unlikely to reverse course.

ChipsThe Register

GPU giants throw HPC users some tasty crumbs

At Supercomputing 2026, Nvidia unveiled Rubin HPC, a variant of its Rubin GPU aimed at high performance computing, promising more than 50,000 CUDA cores and arrival sometime in 2028. Nvidia has not disclosed peak performance, memory capacity, or bandwidth, but says it is using FP64 emulation for dense matrix workloads, with up to 200 emulated FP64 matrix teraFLOPS. The article also notes AMD is developing an HPC-tuned version of its MI400-series GPUs.