Tech news in 3 minutes
The Modulator Is Not the Product: Why AI Photonics Needs an Electro-Optical Realization Corridor
OpenAI and Broadcom have formed an unusual infrastructure financing partnership, with Apollo Global Management acting as investor. Broadcom’s June 9, 2026 10-Q filing discloses a guarantee of up to $29 billion for lease obligations on AI racks built around its custom chips. Apollo purchased the racks and holds the lease; Broadcom guarantees payments if OpenAI defaults. This structure allows OpenAI to access compute without upfront capital, provides Apollo a yield-bearing asset, and secures Broadcom a committed customer. This marks a shift: chip vendors now underwrite deployment rather than just selling silicon. Nvidia is reportedly in talks to guarantee lease obligations for a 10-gigawatt AI campus in Ohio (former Portsmouth uranium enrichment plant), potentially covering energy financing—but no SEC filing confirms this. Nvidia’s financial capacity supports such a move (Q1 FY2026 revenue $81.6B, operating cash flow >$50B), but the $500 billion figure cited in trade press likely exceeds eventual commitments. The semiconductor implications are significant. Broadcom’s guarantee corresponds to thousands of AI racks, boosting volume for its custom AI ASIC designed with OpenAI. This elevates Broadcom from networking and custom chip supplier to a primary compute element provider. However, CFO Kirsten Spears noted that lower-margin custom AI accelerators are diluting consolidated gross margins (from ~78% to 74%), though she called it a mix shift, not structural change. OpenAI generates ~$12 billion annual revenue against projected cumulative cash burn of $115 billion through 2029, and has signed ~$1 trillion in infrastructure agreements. The financing model—where vendors and asset managers carry the burden until inference revenue scales—is unprecedented for the semiconductor industry but may become standard.