A filing by TeraWulf with the US SEC reveals an interesting detail about how AI infrastructure gets built. TeraWulf is building a data center campus in Kentucky to be leased to Anthropic, the maker of Claude, with lease commitments of about $19 billion over 20 years. According to the filing, Google signed credit support agreements backing Anthropic's obligations, to help TeraWulf raise the debt it needs for construction.
How does it work? If Anthropic fails to meet its obligations, Google can step into its shoes and take over the lease, or end it and pay a termination fee. The backstop is meant to end if Anthropic achieves and keeps an investment-grade credit rating for at least three months. Put simply: lenders are willing to fund construction because one of the world's strongest companies stands behind the tenant.
Why is Google doing this? Google is both an Anthropic rival, with its Gemini models, and one of its biggest investors and compute suppliers. In the AI race, the lines between competitors, partners and customers blur. Google has an interest in Anthropic continuing to grow and consume cloud services and chips, even if it competes with it on some products.
The deal also shows something broader. Building AI infrastructure takes such huge sums that even fast-growing companies like Anthropic need giants' backing to raise financing. It comes as Anthropic prepares for an IPO, in a week when Wall Street investors questioned the return on enormous AI spending. For lenders, a Google backstop lowers risk. For the market, it is a reminder that the AI industry's big money is ultimately concentrated in very few companies.
