The real story today isn't what companies are building with AI, but what they're no longer willing to pay for it. Cost discipline is arriving. DoorDash, Siemens, and Airbnb are actively migrating workloads to cheaper Chinese models, a vote of no-confidence in the premium pricing of US-based systems that were supposed to be worth the markup. Simultaneously, employers who pushed staff toward AI adoption are now reversing course as bills hit the ledger, suggesting the productivity gains either never materialized or were never worth the operational expense. The margin between hype and actual ROI has compressed enough that even well-capitalized companies are shopping around. Meanwhile, the compliance and safety work accumulates without generating revenue: auditing techniques for malicious capabilities, red teaming frameworks, federal agencies building monitoring infrastructure. These are real costs with no direct business case, which means they'll either get underfunded or become regulatory requirements that level the playing field. The quantum computing angle for drug discovery is a reminder that legitimate scientific work still happens in the margins, cobbled together from whatever funding researchers can scrap together, while the main stage fills with cost-cutting exercises and defensive infrastructure. When the conversation shifts from capability to price and the conversation shifts from innovation to safety, the market is telling you the first wave of value capture is complete and the second wave is about managing what you've already built.
Sloane Duvall