OpenAI CFO clarifies 'backstop' talk as industry-wide
OpenAI CFO Sarah Friar clarified late Wednesday that her comment about a possible U.S. government "backstop" referred to the AI industry overall, not to OpenAI.
The clarification came hours after her remarks at a business event drew criticism online. In a LinkedIn post around midnight Eastern time, she wrote: "OpenAI is not seeking a government backstop for our infrastructure commitments. I used the word 'backstop' and it muddied the point." In finance, a “backstop” typically refers to a government guarantee or last‑resort support that underwrites lenders and investors and allows financing to proceed if private capital is insufficient or a default occurs.
During the event, Friar described an effort to build an "ecosystem of banks [and] private equity" to fund AI infrastructure. She added that the federal government could potentially "backstop the guarantee that allows the financing to happen," framing a role for public and private sectors in expanding U.S. capacity.
The remarks triggered pushback from investors and politicians. Julian Brigden on X questioned why taxpayers should guarantee the company’s debt. Florida Governor Ron DeSantis posted: "No. Privatized profits and socialized losses." Computer scientist Peter Voss wrote: "Let OpenAI fail if they can't survive by themselves." Analyst Gordon Johnson criticized the idea of federal guarantees while raising concerns about a potential IPO.
OpenAI has outlined plans for extensive spending on data centers, chips and power that industry estimates place at more than $1 trillion over time. The company remains unprofitable and has signed multi-billion-dollar agreements for cloud capacity and semiconductors with Nvidia, Advanced Micro Devices, Amazon and Oracle. Executives have explored a public listing that could rank among the largest initial offerings.
Friar did not offer specifics on potential financing structures, timing or any government programs.
As GNcrypto wrote previously, OpenAI on Nov. 3, 2025 agreed to spend $38 billion over seven years on Amazon Web Services capacity, securing access to large numbers of Nvidia accelerators. Most of the planned capacity is expected online by the end of 2026, with options to expand. The company has discussed long-term targets of roughly 30 gigawatts of compute to train and run its models.
