

Large cloud vendors face a $3.5 trillion external financing gap over the next five years as AI infrastructure spending pushes their capital expenditure-to-revenue ratio above 40% by 2027, according to a CICC research report published Thursday. "AI applications need to generate roughly $1 trillion in annual revenue by 2030 to service the debt costs from this buildout," the CICC analysts wrote. "At a 50% EBITDA margin and five-year depreciation schedule, the math works — but margins below 20% would make it impossible to cover capital costs." The report estimates that cloud vendors' capex-to-revenue ratio has already climbed from about 12% in 2023 to roughly 23% in 2025, with consensus projections pointing to 40% or higher by 2027 — levels approaching or exceeding historical peaks in both the internet and energy sectors. Alphabet's second-quarter results illustrate the trend: the company generated $39.1 billion in operating cash flow but reported negative free cash flow of $5.9 billion, its first since becoming a public company, as it raised its 2026 capital expenditure guidance to between $195 billion and $205 billion. **Where the $3.5 trillion comes from** CICC's base case allocates the external financing across five channels: $1.5 trillion in investment-grade bonds, $1.1 trillion in private capital, $400 billion in public equity, $300 billion in leveraged finance, and $300 billion in asset-backed securitization. Investment-grade bonds and private capital serve as the foundation, with the former relying on cloud vendors' balance sheet strength and the latter offering flexible off-balance-sheet structures for higher-risk projects. Alphabet's estimated $250 billion in off-balance-sheet obligations — including GPU commitments, leases, and joint ventures — reflects this trend. A Nikkei investigation found the broader technology industry carries roughly $1.65 trillion in such commitments, compared with about $1.35 trillion in reported debt. The financing wave is already generating revenue for banks. The six largest US banks reported non-interest income up 32% year over year in the second quarter of 2026, while loans to non-bank financial institutions rose about 25% as of June, with AI financing a significant contributor. **The trillion-dollar question: can AI revenue keep pace?** Bond maturities for large cloud vendors will peak between 2027 and 2032 at roughly $28 billion annually, 60% higher than the 2024-2026 period, CICC estimates. Refinancing can buy time but cannot substitute for cash flow: if project utilization, margins, and asset life fall short, rolling over debt would push leverage and financing costs higher. The report draws a distinction between risk to individual companies and systemic risk. Because the primary investors remain cash-rich cloud vendors with strong balance sheets — Alphabet, Microsoft, and Amazon collectively generated more than $150 billion in annual operating cash flow — even underperforming projects would likely trigger valuation adjustments and capex slowdowns rather than a financial crisis. Alphabet's $514 billion cloud backlog, up from roughly $460 billion three months ago, suggests contracted demand already exists to support much of the spending. Still, the rapid growth of external financing — particularly private capital and leveraged loans — creates financial spillover risk. "The scale of financing itself is not the problem," the CICC analysts said. "The real test is whether cash flow can materialize before financing costs rise and assets depreciate." For investors, the calculus comes down to timing. Alphabet trades at roughly 22 times forward earnings, a discount to Microsoft at 30 times, as the market prices in near-term cash flow pressure. If AI revenue growth sustains the roughly 100% annual trajectory implied by the $1 trillion target, today's spending spree could underwrite a decade of returns. If it falls short, the refinancing wall that arrives in 2027 will separate the winners from the over-leveraged. This article is for informational purposes only and does not constitute investment advice.

PG&E Corp. reported second-quarter GAAP earnings of 33 cents a share, up from 24 cents a year earlier, as cost savings and customer capital investment offset a lower authorized return on equity. "Our PG&E team continues to make meaningful progress delivering safe, affordable, reliable, and clean energy to our customers," Chief Executive Officer Patti Poppe said. "We're encouraged by the engagement around California wildfire liability reform and are advocating for a durable solution that delivers for customers and investors." Non-GAAP core earnings, which exclude wildfire fund amortization and other non-core items, rose to 40 cents a share from 31 cents, beating the company's implied quarterly run rate of 41 cents based on full-year guidance of $1.64 to $1.66. Revenue was essentially flat at $5.9 billion. Operating and maintenance costs fell 11 percent to $2.54 billion, reflecting the company's push to cut non-fuel O&M by 2 percent to 4 percent. The Oakland-based utility is betting on data center load growth as a key earnings driver, with a pipeline exceeding 12 gigawatts. Every 1 GW of new data center demand could reduce customer bills by 1 percent or more, the company said. PG&E also completed $2.2 billion in utility bond issuance in June, bringing year-to-date debt financings to $4.4 billion. The guidance reaffirmation signals management expects steady earnings growth from rate base expansion and cost controls. Investors will watch California's wildfire liability reform proceedings, which Poppe called critical to PG&E's capital plans and path to investment-grade credit ratings. This article is for informational purposes only and does not constitute investment advice.

**Upstart's conditional OCC approval marks the first national bank charter built entirely on AI-powered underwriting.** Upstart Holdings Inc. received conditional approval from the Office of the Comptroller of the Currency to establish Upstart Bank, N.A., becoming the first nationally chartered bank built from the ground up with AI underwriting. "Conditional approval from the OCC is an important milestone for Upstart Bank and we will continue to work with the OCC, the FDIC, and the Federal Reserve on the remaining steps," said Paul Gu, Upstart's co-founder and chief executive officer. The charter, applied for in March 2026, would allow Upstart to originate loans nationwide and accept FDIC-insured deposits from its Delaware base without physical branches. More than 100 banks and credit unions already use Upstart's AI models to approve borrowers, with over 90 percent of loans processed without human intervention. The company's platform covers personal loans, automotive loans, home equity lines of credit, and its Cash Line revolving credit product. The approval signals regulatory validation of AI-driven credit scoring, a model that has faced scrutiny over algorithmic fairness. Upstart's applications to the FDIC for deposit insurance and to the Federal Reserve for bank holding company status remain pending, with operations not expected to begin until all conditions are satisfied, including capitalization, governance, and operational readiness requirements customary for de novo national bank charters. Upstart's model differs from traditional lenders such as JPMorgan Chase and Wells Fargo, which rely on FICO scores as their primary underwriting input. Upstart uses machine learning to evaluate thousands of data points — including education, employment history, and loan purpose — to predict default risk. The company has said its model approves 27 percent more borrowers than traditional credit scoring at the same loss rate, though critics have raised concerns about potential bias in algorithmic lending. The charter is designed to reduce operational, regulatory, and financial complexity for Upstart and its third-party capital partners. Banks, credit unions, and institutional credit funds are expected to continue purchasing the vast majority of loans originated on the platform, with Upstart Bank intended to complement rather than replace those funding partnerships. The company was advised by Klaros Group on its de novo charter application. For investors, the OCC approval removes a key regulatory overhang that has weighed on Upstart shares since the company went public in 2020. A national bank charter gives Upstart direct access to deposit funding, potentially lowering its cost of capital and improving margin structure over time. The company's ability to hold loans on its own balance sheet also reduces dependence on third-party bank partners, whose willingness to fund loans can shift with the economic cycle. Upstart trades as a high-growth fintech rather than a traditional bank multiple, and the charter could narrow that valuation gap if it delivers on promised cost savings. This article is for informational purposes only and does not constitute investment advice.