Nvidia has reportedly agreed to provide up to $105 billion in credit support for the land, power, and physical buildings supporting OpenAI’s planned computing campus in Ohio. What changes for equity investors when a merchant semiconductor supplier begins financing the physical infrastructure required to deploy its own silicon?
Commercial demand for computing capacity is visibly accelerating. However, major technology platforms are making infrastructure commitments that span multi-decade horizons, precisely when elevated long-term interest rates impose higher hurdle rates on every dollar of invested capital.
Chip Suppliers Expand Into Infrastructure Credit Guarantees
Nvidia’s reported $1.5 billion equity investment in developer SB Energy, combined with a $105 billion credit backstop, represents a significant evolution in business model. The strategic motivation is clear: if electrical utility interconnections and physical facility development constrain compute deployments, the chip supplier can use its financial muscle to remove those bottlenecks.
However, selling high-margin accelerators is a fundamentally different business from underwriting real estate credit and utility generation. The planned Ohio campus is scheduled to begin delivering capacity in 2028, with construction extending through 2032 under a 20-year master lease with OpenAI. By providing balance sheet guarantees on long-duration real estate and power assets, Nvidia is assuming structural project risks that outlast typical semiconductor product cycles.
Because specific lease structures, ownership splits, and loss-sharing covenants were not disclosed, assessing the return on this credit support is impossible today. The $105 billion cap indicates the immense scale of the commitment, but it does not reveal what return Nvidia earns for bearing that risk.
Operating Revenue Validates Customer Demand
These extensive capital commitments would be deeply alarming if underlying software demand remained purely speculative. Fortunately, commercial adoption metrics are demonstrating tangible operational traction across several enterprise leaders.
Anthropic’s reported annualized revenue run rate surged from $9 billion in late 2025 to over $65 billion by July 2026, generating positive adjusted operating income on quarterly revenue exceeding $11.5 billion. Supporting hardware suppliers show parallel operational momentum. Optical supplier Fabrinet reported quarterly revenue of $1.32 billion, up 45% year over year, with adjusted earnings per share climbing 55% to $4.10.
Meanwhile, specialized infrastructure operator CBRS contracted 750 megawatts of capacity through 2028 as the primary compute backbone for OpenAI’s advanced models, while video model developer Higgsfield saw annualized revenue reach $700 million. Nvidia itself guided current-quarter revenue to $91.0 billion, a 96% annual increase, while holding gross margins near 75% without relying on Chinese data center sales. These figures demonstrate that real utilization is expanding alongside physical infrastructure.
Real Demand Does Not Validate Every Mega-Project
While operational demand is expanding rapidly, commercial traction does not automatically validate every multi-billion-dollar infrastructure development. Amazon expanded its planned Louisiana data center budget from $12 billion to $18 billion, adding a third campus to guarantee long-term power allocations.
When capital outlays incorporate construction timelines and lease commitments extending over 10 to 20 years, projects become exceptionally vulnerable to shifts in financing costs. The 30-year U.S. Treasury yield advanced to 5.3%, touching its highest level in nineteen years, while the 10-year Treasury yield rose above 4.7%. Higher benchmark yields raise the hurdle rate for capital-intensive infrastructure, data center construction, and long-duration cash flows.
Capital Allocation Is the Defining Risk
The core investment question has evolved beyond measuring how many tokens are processed or how many accelerators are shipped. Investors must now analyze who finances the underlying facilities, who owns the physical power assets, and who bears the loss if customer utilization decelerates.
AI infrastructure is no longer a simple software or hardware story. As merchant suppliers step in to guarantee land, power, and debt for their largest customers, balance sheet discipline and counterparty underwriting become as vital to investment success as chip architecture.