Key takeaways
- Goldman Sachs sees potential AI infrastructure investment exceeding $10 trillion
- Rising borrowing costs could threaten data center project profitability
- Wall Street banks expect major opportunities financing AI expansion
Goldman Sachs CEO David Solomon expects the massive expansion of artificial intelligence infrastructure to continue, even as growing borrowing requirements and higher financing costs raise questions about whether every planned data center will generate adequate returns.
During the bank's October earnings call, Solomon cited studies estimating that US data center construction could require more than $10 trillion through 2032, equivalent to approximately 3.6% of annual US GDP.
The outlook reflects a widening gap between demand for AI computing and available infrastructure. However, it also comes as investors increasingly scrutinize the debt financing supporting expensive projects across the technology sector.
Goldman Sachs Sees Trillions in AI Financing Opportunities
Solomon said demand for computing capacity continues to exceed supply, creating substantial opportunities for banks to arrange financing for data centers, chips and supporting infrastructure.
Goldman Sachs' October 28 research estimated that AI infrastructure investment could reach $8 trillion over four to five years, while other forecasts suggest spending could exceed $10 trillion.
The projections cover overlapping investment categories and different periods, meaning they should not be added together.
For Goldman Sachs, the opportunity extends beyond traditional corporate lending. Investment banks can arrange debt issuance, structure project financing, connect developers with institutional investors and manage capital for infrastructure funds.
Goldman is also involved in Nvidia's initiative to mobilize more than $500 billion in third-party capital for AI infrastructure, although the plan depends on attracting investors rather than representing completed financing.
That initiative reflects the broader AI financing model, where expensive computing equipment increasingly supports loans, leases and long-term investment structures.
AI Debt Risks Grow as Borrowing Costs Rise
Despite strong demand, Solomon warned that growing capital requirements could place upward pressure on financing rates, reducing the profitability of certain developments.
The concern is already visible among specialized cloud providers. CoreWeave reported approximately $104 billion in revenue backlog but also recorded $640 million in quarterly net interest expense during its second quarter.
Its expanding financing costs illustrate why substantial customer commitments do not automatically translate into positive earnings or cash flow.
Credit markets have also shown signs of strain, with approximately $18 billion in Oracle-linked loans reportedly trading below face value as investors reassessed data center financing exposure.
Electricity Constraints Threaten the AI Buildout
Financing is only one obstacle. Goldman Sachs Research forecasts global data center electricity demand will increase 170% between 2025 and 2030, with more than 60% of that growth coming from the United States.
Grid connection delays, equipment shortages and local opposition could postpone projects, increasing borrowing expenses before facilities begin generating revenue.
These challenges have already complicated America's planned $10 trillion infrastructure expansion, with developers confronting uncertain timelines for construction and power availability.
