Anthropic's $35 Billion Deal | What Is the Three-Tier Structure of NVIDIA → Lambda?
機械翻訳 / Machine-translated
A
AI Friends
3日前・12分で読める
機械翻訳 / Machine-translated
@aifriends
AI Friends（https://aifriends.jp）のクロスポスト公式アカウント。AIツールの紹介・使い方・できることを、中学生でもわかるやさしい日本語で届けます。
What you'll learn from this article
On August 31, 2026, it was revealed that Anthropic, an AI development company, had signed a six-year, $35 billion cloud computing contract with a company called Lambda.
The contract involves a data center currently under construction near Corpus Christi, Texas. The facility is called "Beacon Point" and spans 525 acres (roughly equivalent to about 45 Tokyo Domes).
Its power capacity is approximately 350 megawatts — enough to supply around 300,000 average households — making it an enormous facility. It is scheduled to begin operations in the first quarter of 2027.
What has drawn attention to this deal is just how complex the flow of money is. Ordinarily, it would be as simple as "a data center company sells computing power," but in this case, three companies are involved.
First, the data center is being built by a company called Hut 8. Originally a cryptocurrency mining company, it has since pivoted to the data center business.
Next, NVIDIA will lease this data center for 15 years. The lease is reported to be worth approximately $20 billion. NVIDIA is well known as the company that makes chips (GPUs) for AI.
NVIDIA then supplies chips to Lambda. Lambda installs the chips in the data center and sells that computing capacity to Anthropic.
In other words, the structure looks like this:
Even more intriguing is the fact that after selling its products to Lambda, NVIDIA is then leasing them back.
According to reports, NVIDIA has also signed a contract to lease GPU servers from Lambda for $1.5 billion. Specifically, it will lease 10,000 servers over four years for $1.3 billion, plus an additional 8,000 servers for $200 million.
In other words, NVIDIA sells the chips it manufactures to Lambda, and then turns around and borrows them back from Lambda.
Why would it do such a thing? The reason is a severe shortage of AI chips. As the number of companies developing AI has surged, demand for chips has far outpaced supply. NVIDIA itself has constructed this complex capital circulation arrangement in order to ensure that chips reliably reach its customers.
NVIDIA is both an investor in Lambda and has simultaneously become its largest customer. This represents an extraordinary situation unfolding in the AI infrastructure industry.
The backdrop to Anthropic signing such an enormous contract is the race to secure computing capacity.
Anthropic develops an AI called Claude. Highly popular alongside ChatGPT, its user base grew rapidly entering 2026. As a result, computing capacity became insufficient, leading to a supply shortage.
In response, Anthropic signed a series of massive contracts throughout 2026:
The total comes to an astronomical $159 billion (approximately 23 trillion yen).
Each of these contracts involves securing facilities set to become operational years in the future. In other words, companies are competing to claim computing capacity that does not yet exist.
In June 2026, Anthropic submitted a confidential IPO filing to the SEC (U.S. Securities and Exchange Commission). An IPO is the process of going public to raise capital by listing shares.
As of May 2026, Anthropic's valuation stood at $965 billion — more than three times the market capitalization of Toyota Motor Corporation in Japan (approximately 40 trillion yen).
To achieve a successful IPO, Anthropic needs to demonstrate to investors that it can continue to grow. By securing massive amounts of computing capacity, Anthropic is making the case that it can further enhance Claude's capabilities.
Meanwhile, contract partner Nscale is also planning an IPO in September 2026, targeting $3 billion in funding at a $50 billion valuation, with Goldman Sachs serving as lead underwriter.
In other words, Anthropic and Nscale have a mutually supportive relationship for each other's IPOs. By signing a massive contract with Nscale, Anthropic allows Nscale to point to a major customer when pitching to investors.
These developments could have implications for Japan as well.
First, if the shortage of NVIDIA chips continues, it will become harder for Japanese companies to secure computing capacity when developing AI. Startups and research institutions in Japan are already grappling with soaring cloud service costs.
Additionally, the data center construction boom is driving up electricity demand. Beacon Point in Texas alone requires 350 megawatts, so if similar facilities multiply around the world, electricity prices will likely be affected as well.
On the other hand, this also represents an opportunity for Japanese data center operators. For instance, if companies like NTT Data or Fujitsu strengthen their AI-dedicated data centers, they could potentially win contracts from overseas AI firms.
The AI industry shows no signs of slowing down. Yet beneath the surface, a fierce battle for computing power is raging. This latest contract is a symbolic event that offers a glimpse into that struggle.
This article is a cross-post from AI Friends.
