Anthropic has signed another $35 billion contract for compute capacity. This time, the counterparty is Lambda. The WSJ revealed an intriguing transaction chain. At the bottom of the chain is the Beacon Point data center in Texas, built by Hut 8. Previously, Hut 8 only disclosed that a “high-investment-grade client” had signed two 15-year leases totaling 704 MW, with base rent amounting to $19.6 billion. Now, that client has been identified as Nvidia. Nvidia first leases the data center space. Lambda then uses this capacity to install GPUs purchased from Nvidia and provides cloud compute services to Anthropic. Lambda is itself an Nvidia-invested company. The resulting chain is as follows: Hut 8 builds the data center → Nvidia signs a long-term lease → Lambda installs Nvidia GPUs → Anthropic pays $35 billion for compute. Just days ago, Anthropic signed another $45 billion contract—with UK-based Neocloud Nscale—for six years of 460 MW of compute capacity in West Virginia, also running Nvidia’s Vera Rubin chips. Nscale has also received funding from Nvidia: Nvidia was listed as an investor in its 2025 Series B and Pre-Series C rounds, as well as its 2026 Series C round. Thus, within a single month, Anthropic’s newly signed contracts total $80 billion. Both suppliers are backed by Nvidia, and both ultimately deliver Nvidia compute power to Anthropic. The Texas deal goes even further. Nvidia no longer waits for cloud providers to raise capital before purchasing GPUs. Instead, it now invests directly in cloud companies, uses its own credit to secure long-term data center leases, helps finance entire projects, and then ensures these cloud providers fill the infrastructure with Nvidia GPUs before targeting end customers like Anthropic. Hut 8’s Beacon Point is a clear example: backed by this investment-grade long-term lease, Hut 8 raised $4.25 billion in investment-grade secured debt for the project in June this year. A 1 GW data center project that would normally require sequential steps—securing customers, raising capital, obtaining chips—has now been pre-locked by Nvidia’s balance sheet. This is also why analyzing Nvidia’s financials has become increasingly complex. A single GPU sold today may be underpinned by Nvidia’s equity investments in cloud firms, long-term data center leases, and even various minimum revenue guarantees and credit supports. Chip demand, capital investment, and infrastructure financing are being woven into a single interconnected web. Anthropic’s $80 billion in contracts are real agreements. But if one wants to assess how “market-driven” AI compute demand truly is, it’s no longer enough to look only at the final purchase order. One must dig deeper—to see who invested in the seller, who leased the data center, who financed the project, and—who first absorbed the initial risk onto their own balance sheet.