📌 AI Compute Financing Enters a New Stage
The focus of AI capital expenditure is shifting from “who spends more” to “who can turn commitments into cash flow.” Google’s and Meta’s moves around Anthropic show that large technology companies are not only continuing to expand data centers, but are also using guarantees, leases, and compute resale to share the pressure.
Google is supporting Anthropic’s planned 1.6 GW Texas AI campus. The related financing is being sought by Nexus, at a scale of about USD 15 billion, for data centers and on-site power. More importantly, Google will provide guarantees for billions of dollars in lease and power obligations, in exchange for about 20% equity. This structure ties three things together: Anthropic obtains compute, Google locks in TPU demand, and financiers obtain stronger credit support. The cost has not disappeared; it has only partly turned into guarantees and long-term obligations.
Meta’s thread is similar. The company disclosed future AI infrastructure commitments of about USD 700 billion, covering purchase obligations and leases; at the same time, it is reportedly discussing leasing about USD 10 billion of excess compute to Anthropic. If the transaction is completed, external leasing revenue can buffer part of the cash flow pressure. But it still depends on the continuity of customer demand, pricing, and actual delivery capability.
Data center deployment is also affected by non-financial constraints. NBIS used low water usage, waste-heat recovery, and reductions in local heating bills to support facility approval, showing that power, water, and local approvals will affect the pace of construction. Investors cannot look only at chip orders or capex metrics.
The market is currently more willing to reward revenue visibility. Microsoft used Azure’s high growth rate to prove that AI investment has already produced returns; Google Cloud’s growth is also strong, but after it clearly signaled a significant increase in capital expenditure in fiscal 2027, it instead came under pressure. What needs to be verified next is not whether AI investment will continue, but whether these leasing, guarantee, and resale arrangements can steadily translate into cloud revenue, profit, and free cash flow.
▌ Sources