The bill for building out artificial intelligence infrastructure is coming due, and the numbers are staggering. University of Pennsylvania finance professor Jessica Wachter cut through the hype by focusing on a single, undeniable fact: hyperscalers are pouring capital into data centers at an unprecedented rate. Her analysis reveals that these companies must see their earnings grow at an extraordinary pace just to break even by 2030. With expenditures projected to hit nearly $1.1 trillion by 2027, the current spending spree is less a calculated investment and more a high-stakes gamble on future productivity.
While financial analysts crunch the numbers, OpenAI is looking elsewhere for value. The OpenAI Foundation announced it will fund Ruxandra Teslo’s proposal to mine data from failed biotech companies. Teslo argues that bankruptcy proceedings hold a “lost archive” of regulatory filings and safety data that could supercharge medical AI. This move highlights a growing bottleneck in the industry: models are hungry for high-quality scientific datasets, and public sources are drying up. Paying to unlock private, dormant data may become the new normal for specialized AI development.
Leadership in Silicon Valley remains defiant despite the mounting costs and risks. Nvidia CEO Jensen Huang and Meta’s Mark Zuckerberg have both rejected calls for a coordinated slowdown in AI development. Huang dismissed the need for new safety laws, while Zuckerberg argued that market competition would naturally drive safety improvements. Their stance puts them at odds with regulators like the FTC chair, who warned against granting antitrust waivers to AI firms, and contrasts sharply with the existential warnings discussed in recent MIT Technology Review roundtables.
This divergence between financial reality and executive optimism creates a fragile ecosystem. If the productivity gains don’t materialize quickly enough to offset the $1.1 trillion buildout, the economic foundation of the current AI boom could crack. Meanwhile, the push for proprietary data sources like OpenAI’s biotech bid suggests that the next competitive edge won’t come from bigger models, but from exclusive access to information others can’t touch.
