i
News
News · 2026-10-19

Nvidia’s AI cloud exposure grows as providers face a shakeout

@neuronium_ai @neuronium_ai

A shakeout among AI cloud providers would not be a sideshow for Nvidia. The company’s exposure runs through customers building data centers, buying its chips and, in some cases, relying on its help to secure power, financing and customers. After a week of falling shares and a withdrawn IPO filing, industry executives expect consolidation. Two executives in data centers and cloud computing told The Information that only seven to ten of today’s hundreds of providers may remain in a few years.

Cover: Nvidia’s AI cloud exposure grows as providers face a shakeout

Investors are already repricing the field

Australian provider Firmus abruptly withdrew its IPO filing, while CoreWeave and smaller rivals fell about 20% over the week, The Information reported. Bloomberg reported that CoreWeave’s market value had fallen by a quarter and its credit-risk measure had reached a record high.

Cirrascale CEO David Driggers expects large asset managers, telecom companies and fiber-optic companies to buy individual providers as the market consolidates.

Nscale offers a different warning. Microsoft reportedly exited a deal to lease 23,000 Nvidia chips at a data center in Loughton, England, according to The Telegraph. Energy companies expect electricity to reach the site only in the 2030s. Nscale said its relationship with Microsoft remains strong and long-term.

At the same time, some investors are pressing Nscale to lower its target valuation from about $30 billion to $16–19 billion, Bloomberg reported, citing people familiar with the matter.

Nvidia has more at stake than chip sales

The consequences of a provider failing depend in part on what it owns. CoreWeave mostly rents data centers, which could make those sites easier to lease to new tenants. Nebius, IREN and Nscale build their own facilities; financial trouble could halt construction and slow Nvidia’s sales.

These providers also matter because Nvidia’s largest customers are developing their own chips. In the July quarter, customers outside the largest cloud platforms — including newer providers — generated $40 billion for Nvidia, or 45% of its data-center revenue. That was more than double the share a year earlier.

Nvidia has supported the sector with equity investments, chip allocations, credit guarantees and help finding customers. A team led by Nico Caprez, Jensen Huang’s brother-in-law, now helps providers find sites with available electricity.

One industry executive said Nvidia has also revised its credit-guarantee terms. Large providers rejected an earlier offer that would have given Nvidia a share of future revenue in exchange for a guarantee. The company now offers a shorter guarantee period and a smaller share.

The capacity question is bigger than consolidation

Gartner forecasts that 80% of independent cloud providers will stop operating independently by 2030. Bernstein estimates that North America will build only 35% of its announced data-center capacity, Bloomberg reported.

CoreWeave CEO Michael Intrator has blamed smaller rivals Firmus, Lambda and Nscale for market instability, Bloomberg reported. In a memo to employees, he said they were seeking huge amounts of capital without the execution, revenue, scale and software CoreWeave provides its customers.

Even a successful buildout could create a different problem: supply might eventually exceed demand. Microsoft CEO Satya Nadella made a similar point in a podcast in late February 2025, predicting excess computing capacity from 2027 and arguing that the industry was too focused on supply rather than actual demand. He said data-center construction would bring prices down.

Microsoft has since canceled or delayed data-center projects in the UK and several US states, including Ohio, while remaining one of Nscale’s key customers. More than 40% of Nscale’s $103 billion in contract-backed revenue comes from Microsoft, according to The Telegraph.

My read is that the immediate risk to Nvidia is not simply that weaker providers disappear. It is that projects stop before chips are installed — or that completed capacity arrives into a market where customers have less reason to pay for it. Nvidia is helping providers find power and financing, but that support cannot settle the more basic question: whether demand will keep pace with the facilities being built.

Daily AI news

Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.

Only what matters — every day

Follow on X