America’s Cosmic Bet on AI

Foreign Policy | Graham Allison

The United States has heavily concentrated its economic growth and national security strategy on artificial intelligence, with AI-related investments driving half of all business investment and 85 percent of S&P 500 gains in 2026. However, this high-stakes bet risks a major financial bubble as hyperscalers outpace commercial demand and face severe commoditization from low-cost foreign alternatives.

Massive capital expenditure on data centers has driven market valuations relative to earnings toward levels unrecorded since 1929. DeepSeek charges just 28 cents per output. This represents a 99 percent discount compared to Anthropic’s Opus 4.8 pricing. Chinese competitors are leveraging lean engineering teams to challenge American technological leads without matching Silicon Valley’s trillion-dollar infrastructure spending. Instead of chasing speculative artificial general intelligence, Beijing is deploying an "AI Plus" policy aimed at industrial productivity and widespread economic integration. Consequently, systemic fiscal vulnerabilities will intensify if projected software revenues fail to justify current U.S. corporate debt.

Comment

The price asymmetry between DeepSeek V4 and Claude Opus 4.8 exposes a structural weakness in Western defence-industrial spending models. Capital-intensive frontier labs rely on continuous high-margin software revenues to service private debt and cloud compute infrastructure. When low-cost algorithms achieve functional parity at a 99 percent discount, market-driven capital allocation models for dual-use technology collapse.

This price collapse directly limits the U.S. Department of Defense's ability to leverage commercial software margins for military AI integrations. Privately financed hyperscalers facing margin compression will increasingly demand direct federal subsidies or long-term defence procurement guarantees to sustain their data center operations. Consequently, the Pentagon risks absorbing the fixed infrastructure liabilities of frontier labs whose commercial monetisation models have been undercut by open-weights models like DeepSeek V4.

Strategic Question for Discussion

If low-cost models like DeepSeek V4 neutralise the commercial pricing power of frontier developers like Anthropic, how will the U.S. Department of Defense restructure its capital-heavy dual-use acquisition model?

The trajectory indicates that the Pentagon will be forced to transition from commercial off-the-shelf software procurement to sovereign funding models similar to traditional defence prime contractors. As venture-backed business models face commoditisation, federal defence budgets will increasingly need to underwrite bare-metal infrastructure and data center construction directly to preserve national compute capacity over Chinese competitors.

Share your assessment in the comments below.

To read in details click here