The real story today isn't the robotaxis in Texas or the app store rebound, though both matter. It's that the infrastructure layer, chips, models, talent pipelines, is consolidating while the surface layer fractures into a thousand startups. Cerebras files for IPO on the back of deals with Amazon and OpenAI worth billions, which means the chip wars are already decided in favor of whoever can lock in compute commitments years in advance. Tesla expands robotaxi to Dallas and Houston, but it's operating in a regulatory vacuum where three cities in Texas represent the entire addressable market for driverless cars without a safety driver; that's not scale, that's a sandbox. Meanwhile Anthropic is simultaneously designated a supply-chain risk by the Pentagon and courting the Trump administration while its Mythos model has sparked fears among security researchers that it could accelerate hacking faster than fixes can ship. The company is also betting on Schematik, a tool for coding physical devices, which suggests the real leverage is moving from software to hardware integration. Northwestern's artificial neurons communicating with living brain tissue and quantum-AI beating classical models at chaos prediction are genuine technical breakthroughs, but they live in a different timeline than the product launches. The tension underneath is this: capital and regulatory attention are flowing to whoever controls the compute layer and the government relationship, while the flood of new apps and consumer-facing tools suggests that the marginal cost of building something AI-powered has collapsed to nearly zero. One layer is consolidating into oligopoly. The other is exploding into commodity. The winners will be whoever can operate at both levels simultaneously.
Sloane Duvall