In recent days, a single remark by Sarah Friar, CFO of OpenAI, triggered a wave of commentary across the technology and finance worlds. At a Wall Street Journal panel, she suggested that the U.S. government could “backstop the guarantee that allows the financing to happen” for large-scale AI infrastructure, referring to the vast data centers and chip foundries required to train next-generation models.

The reaction was swift and polarized. Critics accused OpenAI of asking for a taxpayer-funded safety net. Supporters argued she was merely describing a capital-structure problem: how to finance national-scale infrastructure in a sector where equipment obsolescence is measured in months rather than decades.

Soon after, CEO Sam Altman clarified that the company was not seeking federal guarantees for its build-out, but rather acknowledging that government participation is essential for maintaining U.S. competitiveness. The exchange nonetheless spotlighted a critical issue: how should the United States finance strategic infrastructure in an era defined by exponential technological decay?

1. The Public Reaction

Much of the debate focused on semantics rather than substance. The term “backstop” sounds like “bailout,” and that is where the discussion became stuck. Yet the underlying question is valid: how can private markets efficiently finance a technology layer that becomes obsolete every two to three years?

• Investors worry about stranded assets, meaning hardware that loses value faster than loan amortization schedules.
• Taxpayers worry about moral hazard, meaning government underwriting private risk.
• Policy thinkers worry about national security, because AI infrastructure is now as strategically important as energy or defense manufacturing.

2. Why the CFO Was Right

Friar’s core point was not about socializing losses; it was about stabilizing innovation. The financing model for AI compute is fundamentally broken. Traditional equity and debt structures assume linear depreciation. In AI, the decay curve is exponential.

That is why the right question is not whether the government should backstop. The real question is how the government can strategically de-risk national technology platforms in the same way it has for every other transformative industry.

3. The Precedent: DOD and DARPA

The United States already knows how to do this. For decades, federal investment has catalyzed private innovation without nationalizing it.

• DARPA funded the foundational work that became the Internet, GPS, and early semiconductors.
• The Department of Defense created procurement frameworks that gave companies predictable revenue streams to build technology first for defense, and later for civilian markets.
• NASA used a similar model with SpaceX, combining government risk absorption with private-sector scalability.

None of these efforts were bailouts. They were structured partnerships that transferred risk in order to encourage investment in technologies critical to U.S. leadership.

This is exactly what a modern AI-infrastructure policy should look like: a public-private financing structure in which the government reduces systemic risk and private investors drive efficiency and scale.

4. China’s Approach: Industrial Policy at Warp Speed

While the United States debates the meaning of “backstop,” China is executing a full-spectrum industrial policy for AI and chip manufacturing.

• The Chinese Communist Party’s 14th Five-Year Plan designates AI, advanced semiconductors, and quantum computing as “core national priorities,” directing tens of billions of dollars in state-backed financing through the National Integrated Circuit Industry Investment Fund, often called the “Big Fund.”
• Local governments supplement this with subsidies for chip fabrication, AI startups, and supercomputing centers.
• State-owned banks provide low-interest, long-term loans that effectively serve as backstops, absorbing risk so private capital does not have to.
• Military-civil fusion policy ensures that every commercial AI advance can serve national-defense objectives.

China has recently gone further. In 2025, Beijing approved up to a 50 percent reduction in electricity costs for data centers that use domestically manufactured AI chips. The policy specifically excludes data centers running on foreign chips such as those from NVIDIA. The goal is to offset the chip-efficiency gap, since China’s domestic accelerators consume roughly 30 to 50 percent more power than NVIDIA’s latest models.

This energy subsidy is a direct industrial countermeasure. By cutting energy costs in half, China is making its domestic AI ecosystem cost-competitive through deliberate state intervention.

The result is that China can deploy at massive scale, absorb inefficiencies, and iterate faster. This is not because its markets are more innovative, but because its capital and policy structures are insulated from short-term risk.

That is the strategic threat. A nation willing to subsidize time and tolerate failure will outpace one that is paralyzed by fear of moral hazard.

5. The U.S. Path Forward

To maintain leadership, the United States does not need to copy China’s model of state capitalism. However, it must acknowledge that national competitiveness now exists in the space between Wall Street and Washington.

Policy recommendations

• Establish AI-infrastructure loan guarantees. These should be modeled after the procurement logic of the Department of Defense. The government would reduce downside risk for core infrastructure such as compute, chips, and power, while avoiding the selection of corporate winners.
• Fund frontier R&D through a modernized DARPA-AI. Expand DARPA’s charter to support long-horizon research into AI, energy-efficient compute, and next-generation fabrication.
• Create transparent and competitive bidding frameworks. Ensure that every firm, from NVIDIA to emerging U.S. startups, competes on innovation and cost rather than political proximity.
• Retain private accountability. Maintain market discipline so that if a firm fails to deliver, it fails. Government guarantees should stabilize the system, not individual companies.

6. Conclusion

The OpenAI CFO was not wrong; she was early. The conversation she started is one America needs to have. The question is not whether the United States should “backstop” AI. The real question is whether we have the courage to modernize our public-private partnership models for an era in which strategic infrastructure no longer means ships, tanks, and planes, but instead data, chips, and power.

AI is not a technology bubble to be managed; it is a national capability to be secured. The right model is not bailout economics but strategic de-risking, the same logic that built the Internet, put humans on the moon, and kept America at the forefront of every major technological revolution in modern history.

Antony Barran
Candidate for U.S. Congress (WA-3)
www.BarranForCongress.com