$NVDAB $MSFTB $GOOGLB #AI #微软 #谷歌

Anthropic is in talks with UK AI compute provider Nscale for a second multibillion-dollar computing lease. CNBC reported on the 20th, citing people familiar with the matter, that the deal could be worth around $45 billion. Bloomberg was first to report the news. Anthropic declined to comment, and Nscale has yet to respond.

Second 460 MW data center at the same campus

People familiar with the matter said the new agreement covers a second data center at the same campus in West Virginia, with a power capacity of 460 MW. CNBC reported in August on the companies’ first similar deal at the same campus.

Nscale, which is backed by Nvidia, builds or leases data centers and sells computing capacity to major cloud providers and AI labs. It filed to list on the New York Stock Exchange in September.

Earlier this year, Anthropic said that growing demand for Claude models and products was putting “inevitable pressure” on its infrastructure. Since then, it has signed agreements to expand computing capacity with companies including AMD, SpaceX, Google, and Broadcom.

Peers CoreWeave and Nebius fall nearly 30% in a month

AI computing providers of this kind have come under pressure recently. CNBC reported that the market is concerned about excessive debt, with peer companies CoreWeave and Nebius seeing their stock prices fall 28% and 27%, respectively, over the past month.

Nscale itself also faces uncertainty. The UK’s Daily Telegraph reported Sunday that Microsoft had withdrawn from a deal to be the main tenant at an Nscale data center near London because of delays in connecting it to the UK power grid. An Nscale spokesperson told CNBC that the company remains “committed” to the project and is seeking alternative power sources for the campus. The spokesperson also said its relationship with Microsoft “remains strong and long-term,” and that the London grid issues do not affect other contracts.

This article, “Anthropic reportedly leases computing capacity from Nscale again, with a second deal worth about $45 billion,” first appeared on .

【Further Analysis】

【My Take】

Anthropic’s continued massive investment in infrastructure reflects the generative AI arms race reaching its peak. However, the enormous $45 billion lease highlights the immense capital expenditure pressures facing AI labs. Meanwhile, problems affecting suppliers such as Nscale—including Microsoft withdrawing from a lease and power grid delays—show that the pace of infrastructure expansion is facing serious constraints from physical realities. Energy bottlenecks and financing burdens will be major pain points for the industry in the coming years.

【Market Impact】

For U.S. stocks, the news is moderately positive for AI chipmakers such as Nvidia and partners AMD and Broadcom, signaling that demand for computing capacity upstream remains strong. But for second-tier computing capacity rental providers such as Nscale, which rely on heavy capital spending, and its peer CoreWeave, concerns about high debt and cash flow are a near-term negative and could continue to weigh on the valuations of related stocks. In Taiwan, server manufacturing and cooling demand could benefit, but if U.S. computing-related stocks weaken due to financing concerns, sentiment could spill over to Taiwan’s related supply chain and make trading more volatile.

【What to Watch】

If the $45 billion deal is formally finalized, close attention should be paid to Anthropic’s funding sources and whether its revenue growth can support such a massive fixed cost. In addition, delays in power supply for AI data centers—such as whether the UK grid situation could be repeated in the U.S.—will be key to determining whether computing infrastructure can be delivered on schedule.

【Risk Warning】

The AI infrastructure market currently relies heavily on leverage and external financing. If macroeconomic conditions tighten or interest rates remain high, computing capacity rental providers could face liquidity crises. Investors should be mindful of the risk of valuation corrections in the AI sector if excessive investment fails to generate commensurate returns.