The infrastructure chase is accelerating while the value proposition remains opaque. OpenAI's Jalapeño chip outperforms current state-of-the-art on inference benchmarks, Generalist hit a $3 billion valuation on the back of a $200 million extension, and Stability AI raised $76 million for its image generator, yet the same week, enterprises report struggling to measure whether any of this actually delivers return on investment. Radiologists still exist. Entry-level workers aged 22 to 25 face 19% lower employment in AI-exposed roles compared to protected occupations. The pattern is clear: capital flows toward capability and scale while the market for actual utility remains fragmented and hard to price.
Builders are consolidating the infrastructure layer while friction points multiply at the edges. Apple's new desktop refresh targets local AI development, Perplexity launched on-device inference to keep data local and reduce token costs, and Keenable exited stealth with $26 million to build web search indices for agents. Anthropic gave Claude shared memory across sessions to reduce repeated context setup. These are not flashy announcements; they are solutions to friction that emerges only after deployment. Meanwhile, OpenAI lost a top data center executive amid reorganization, and the grid constraint is shifting from chips and cooling to raw power availability. The infrastructure is being built. The operational problems are being solved. But the question of what these systems should actually do remains unsettled.
The labor market is bifurcating faster than policy can respond. Bill Gates published a 6,000-word essay warning that AI will usher in "one of the most turbulent times in human history" and called for "human reserved" jobs. Spirit Airlines is selling flight attendant data to Google for AI training. CSU students and faculty remain divided on whether AI actually helps learning. Durham County, North Carolina voted for a moratorium on data centers; Huntington, West Virginia residents launched a ballot initiative to ban mass surveillance tech. The infrastructure is global and capital-intensive. The governance is local, reactive, and fragmented. Power is concentrating in the hands of those who control compute and data pipelines. Resistance is organizing at the municipal level. Neither side is moving fast enough to catch the other.
Sloane Duvall