Apollo President Jim Zelter on leading $35 billion financing on Broadcom

Key Concepts

• Compute Expansion: The massive scaling of AI processing power required to meet surging demand.
• Structured Finance: A multi-layered approach to funding large-scale capital expenditures (CapEx) involving banks, private capital, and equity.
• Hyperscalers: Large-scale cloud providers (e.g., Google, Microsoft) driving massive infrastructure investment.
• TPU (Tensor Processing Unit): Specialized AI accelerator chips developed by Google and Broadcom.
• Integrated Financing: The modern strategy of combining public equity, bank debt, and private capital to fund massive corporate projects.

---

1. The Anthropic $35 Billion Financing Deal

Apollo Global Management, led by President Jim Zelter, has spearheaded a $35 billion financing deal for Anthropic. The primary objective is to support Anthropic’s "compute expansion," which is scheduled to begin in mid-2026.

• Mechanism: The deal facilitates the acquisition of TPU chips developed by Broadcom and Google. Anthropic will receive these chips over the next two and a half years.
• Structure: The financing is divided into multiple tranches:
  • Banks: Act as senior lenders, taking an initial portion of the risk.
  • Apollo: Provides financing over a 4-to-5-year horizon.
  • Residual Risk: Anthropic retains the remaining risk.
• Analogy: Zelter compares this to automotive financing (e.g., Ford or GM Finance), where the lender facilitates the purchase of high-value assets that are essential to the borrower's operations.

2. Market Trends and Future Outlook

Zelter argues that this deal is a blueprint for future infrastructure financing.

• Replicability: The model is expected to be replicated across the AI industry and other sectors. As hyperscalers commit roughly $700 billion in CapEx, the demand for financing the underlying hardware (chips and data centers) will grow exponentially.
• Competitive Advantage: Access to capital has become a critical competitive advantage. Companies like Anthropic and OpenAI, which may not yet have a traditional "debt following" or public market history, are increasingly relying on private capital partners to bridge the gap.
• Shift in Financing: The traditional "or" (debt or equity or banks) has been replaced by "and." CFOs are now coordinating public capital, private capital, and bank capital simultaneously to manage massive funding requirements.

3. The Role of Private Capital and Insurance Balance Sheets

A significant portion of the discussion focused on where the massive global capital requirements—estimated at $90 trillion for next-generation infrastructure and $30–50 trillion for energy transition—will originate.

• Integrated Toolboxes: Firms like Apollo utilize "long-dated liabilities" from insurance balance sheets (specifically Athene) to provide stable, long-term funding.
• Investment Grade Focus: Unlike the previous 20–25 years, which often focused on high-yield, speculative disruptors, the current market is seeing "disruptors" (AI firms) that are being evaluated at the highest levels of investment grade, allowing them to tap into more conservative, large-scale capital pools.

4. Notable Quotes

• "The 'or' is now replaced with 'and.' ... You're talking about public capital, private capital, and bank capital all together." — Jim Zelter, on the evolution of corporate financing.
• "The ability to have access to financing is a competitive advantage." — Jim Zelter, regarding the necessity of sophisticated capital structures for AI firms.

5. Synthesis and Conclusion

The $35 billion Anthropic deal represents a paradigm shift in how capital-intensive AI infrastructure is funded. By moving away from traditional, siloed financing methods, companies are now leveraging integrated, multi-layered structures that combine the stability of insurance-backed private capital with traditional bank lending. As the demand for compute capacity continues to surge, this model of "asset-backed" AI financing is likely to become the standard for hyperscalers and AI developers alike, marking a transition toward treating AI hardware as essential, financeable infrastructure.