Jensen Huang Voices Support, Musk Pours Cold Water: The Neocloud Compute Rental Model Faces a Test

Jensen Huang is still trying hard to prop up the Neocloud wall, but Musk has already predicted its collapse.

On October 25, Musk issued a warning on X: in recent weeks, the debt market has almost stopped providing funding to companies that lack substantial revenue outside of AI data centers.

He predicted that more than 200 Neoclouds will face a "guppy-level" death and consolidation, and ultimately only a few companies may survive.

But just three days earlier, Nvidia CEO Jensen Huang offered an almost opposite judgment.

On October 22, Jensen Huang insisted in an interview with CNBC that demand for Neocloud computing power remains extremely strong, that the only thing truly limiting their development is insufficient working capital, and that financing difficulties will eventually be resolved.

Jensen Huang props them up, Musk pushes them down

Neocloud, simply put, refers to a new type of cloud computing company that specializes in renting out AI computing power. Unlike traditional cloud providers with broad businesses such as Amazon AWS and Microsoft Azure, these companies mainly provide services around AI computing demand. They purchase large numbers of GPUs, build data centers, and then rent out computing power to enterprises that need to train and run AI models.

The most representative companies among them are CoreWeave, Nebius, and Nscale.

Since entering the AI computing power rental market, their business has been highly dependent on Nvidia's GPUs. For Nvidia, Neoclouds are naturally very important customers.

Jensen Huang's support for them is not only verbal; he has put in real money and even personally backstopped Neocloud computing power revenue.

In September 2025, Nvidia signed a computing power purchase agreement with CoreWeave with an initial value of $6.3 billion. Under the agreement, if CoreWeave's data center computing power cannot all be sold to other customers, Nvidia will purchase the remaining computing power according to the agreement, with the relevant commitment lasting until 2032.

In January of this year, Nvidia further invested $2 billion in CoreWeave to support it in accelerating data center construction, and the two sides plan to build more than 5 gigawatts of AI computing infrastructure by 2030.

In March, Nvidia also announced a $2 billion investment in another Neocloud company, Nebius, to help it expand its AI cloud computing business.

In addition, Nvidia also designed a more direct support mechanism. On the August 26 earnings call this year, Nvidia management disclosed that the company had begun providing minimum revenue guarantees to some Neoclouds, promising to purchase a portion of the computing power in projects to help them obtain bank financing.

In exchange, when the revenue generated by this computing power exceeds the guaranteed level, Nvidia will share in the profits.

"Under this model, we can make money twice: once from hardware sales, and once from rental revenue sharing," Nvidia Chief Financial Officer Colette Kress said directly.

Some media commentary even simply compared Nvidia to "the central bank of the AI industry," describing how it sells GPUs to Neoclouds while providing funding support to these companies through equity investment, computing power purchase commitments, and other means, helping them continue to obtain loans, expand data centers, and keep the entire AI computing power market running.

Jensen Huang's active support for Neoclouds of course has very real commercial interests. These companies need to continuously buy the latest GPUs and expand data center scale in order to meet customer demand. The continued expansion of Neoclouds means Nvidia can continuously obtain new chip orders.

However, Nvidia's willingness to provide financing support to Neoclouds does not mean the capital market is also willing to continue lending to them.

This is also the reason Musk issued his warning.

It is worth noting that Musk himself is also borrowing heavily to buy GPUs. On October 25, he had just announced that SpaceX's $40 billion debt financing for purchasing Nvidia GPUs had been "closed," mainly corresponding to GPUs delivered in the first and second quarters of 2027.

But unlike Neoclouds, which mainly rely on computing power rental revenue, SpaceX also has mature businesses such as Starlink and rocket launches as support. In Musk's view, Neoclouds lacking other sources of revenue are likely unable to survive the coming industry reshuffle.

When a netizen speculated that Musk was deliberately talking down Neoclouds so as to acquire them cheaply after stock prices fell, Musk's answer was:

"SpaceX will not buy any Neocloud. It makes no sense."

Have the good days for Neoclouds come to an end?

Since 2023, with the explosion of generative AI, Nvidia GPUs have been in short supply. Traditional cloud providers such as Microsoft, Google, and Amazon could not meet market demand in time, and Neoclouds seized the opportunity to rise rapidly.

They do not need to build a complete cloud service system; as long as they can buy GPUs and deploy clusters well, they have a chance to win large orders.

CoreWeave is a typical representative: in 2023, the company's full-year revenue was only $229 million, and by 2025 it had reached $5.13 billion, an increase of about 22 times in two years.

Nebius's AI cloud business has grown equally astonishingly, with revenue of only $9.6 million in 2023 and already $480 million in 2025.

But a surge in revenue does not mean long-term profitability.

Although CoreWeave's revenue reached $5.131 billion in 2025, its net loss was $1.167 billion, and full-year net interest expense was as high as $1.229 billion.

Neoclouds need to invest huge sums upfront to buy GPUs and build data centers, then recoup the costs through future rental income. But as GPUs are upgraded, rental prices are also falling.

A McKinsey study from November 2025 showed that from 2021 to 2025, public rental quotes for Nvidia's A100 fell from $3.10 per hour to $1.50 per hour, a drop of more than 50%.

McKinsey also used the B200 GPU as an example to calculate that, with full equity funding and no debt financing considered, the operating margin of the GPU rental business after deducting depreciation and other costs is only about 8%. If equipment utilization falls below 80%, or rental prices decline slightly, the return on investment could be wiped out.

Neoclouds must bear huge capital expenditures while also facing GPU depreciation and competition from large customers. Once supply and demand change, past advantages may quickly disappear.

And recently, the capital market has begun to re-examine this business.

Since October, CoreWeave, a representative Neocloud company, has come under pressure in both its stock and bonds. As of October 26, its share price had fallen nearly 38% from its October 1 closing price to $55.19. In mid-October, yields on some of its bonds briefly rose to 15% to 16%; by October 26, the price of some CoreWeave bonds maturing in 2033 had fallen to a record low.

On October 22, foreign media reported that investors shorting five AI infrastructure companies including CoreWeave and Nebius had made paper profits of about $6 billion that month. Among them, CoreWeave's share price fell nearly 30% that month.

Interestingly, on October 26, CoreWeave's share price also experienced a rebound. That day, its shares fell about 5% in premarket trading due to Musk's pessimistic remarks, then were lifted by the news that Anthropic had publicly released its IPO prospectus, and finally closed up 1.76%.

However, this rebound did not change CoreWeave's overall decline over the past month.

Beyond the capital market, data center construction has also run into trouble.

On October 19, Microsoft withdrew from a data center project of Neocloud company Nscale in the UK. The project had been scheduled to begin operations in early 2027 and deploy more than 23,000 Nvidia GB300 GPUs, but power supply may not be secured until the 2030s.

For Neoclouds that rely on debt financing to build data centers, project delays mean revenue cannot be realized for the time being, while upfront investment and financing costs continue to accumulate. The longer the delay, the greater the financial pressure.

Research firm Gartner expects that by 2030, more than 80% of Neoclouds will stop operating independently due to cost pressure and lack of differentiation, including being acquired, merging, or exiting the market.

This industry, which rose rapidly on the back of the GPU shortage, has now begun to pay the price for its own fragile business model.

Where is the way forward?

Neoclouds' current troubles may be only the beginning.

On October 25, Silicon Valley investor and Social Capital founder Chamath Palihapitiya, in response to Musk, offered a bold and pessimistic prediction: in 2027, the AI industry may see a large amount of idle computing power.

Chamath himself holds a batch of data center land assets that are already connected to power, so he has direct observation of the construction of AI infrastructure. In his view, Neoclouds are generally underfunded and poorly organized; although some companies perform relatively well, the industry's overall construction capacity remains worrying.

However, traditional cloud providers are not much better—Chamath complained that these large companies are severely bureaucratic, that signing a single non-disclosure agreement can take three months, and that they demand escrow periods of up to 18 months.

The bigger problem is that data center projects already planned are being delayed.

Chamath believes these problems will be concentratedly exposed in 2027. By then, large cloud providers may prioritize their own computing needs, while large numbers of GPUs already purchased may not be fully put into use because data centers are not completed, demand and supply are mismatched, and other reasons.

He therefore expects 2027 to be a year of industry turmoil, with many wrong investment decisions liquidated and some companies disappearing in consolidation.

Moreover, even if these construction and delivery problems are eventually resolved, Neoclouds' situation may not necessarily improve.

Research firm Gartner expects spending in the AI-optimized cloud infrastructure market to grow from $66 billion in 2027 to $174 billion in 2030. Even so, market growth may still not be enough for most AI computing suppliers to recoup their investments before the current generation of chips is replaced.

As tech giants such as Microsoft, Google, and Amazon continue to expand their own AI infrastructure and GPU supply increases, the market environment on which Neoclouds rose may also gradually change.

McKinsey pointed out as early as November 2025 that Neoclouds were originally born to fill the GPU supply gap. If they cannot break away from the business model of simply renting out computing power and establish long-term competitive advantages, they may repeat the fate of early cloud computing companies, ultimately being acquired by large cloud providers, squeezed out of the market, or retreating to niche areas.

Nvidia can provide financing support and guaranteed purchases to help these companies get through their current difficulties, but it cannot guarantee that they will always be able to make money from the GPU shortage.

Jensen Huang can support them for a time, but not forever.

In particular, in the entire AI industry chain, Neoclouds may be the link most easily bypassed and with the weakest risk resistance.

For tech giants that truly need computing power, renting GPUs from Neoclouds is only one way to meet demand. They can build their own data centers, acquire ready-made computing assets, or rent computing power from large cloud providers.

However, although large companies such as Oracle, Microsoft, Amazon, and even SpaceX also bear huge capital expenditures, they at least have other mature businesses providing financial support and can withstand longer market fluctuations.

Neoclouds have no such fallback; as long as revenue growth from computing power leasing slows even slightly, the business model originally built on high-growth expectations may become unsustainable.

The same market changes may be merely a period of turbulence for large tech companies, but could become an existential crisis for Neoclouds.

As the GPU shortage gradually eases, customers have more choices, and Neoclouds cannot prove that they are irreplaceable, this group of companies may become the first casualties in the AI industry's expansion process.