Humanoid robots deploy to airports while H100 GPU prices climb

A humanoid robot is now greeting travellers at San José airport, answering questions in 50+ languages-a quiet milestone in the shift from laboratory novelty to working public infrastructure. IntBot's José represents the emerging category of what builders call "social intelligence" robots: systems designed for interaction in complex environments rather than manufacturing or logistics. It's the kind of deployment that only happens when companies believe the technology is ready for scrutiny and scale.

The Economics of AI at Scale Are Shifting

Meanwhile, GPU markets are telling a different story. H100 prices have reversed their three-year decline-they're now worth more than they were three years ago, according to rental market data. That's counterintuitive. Cheaper models exist. Newer chips are available. Yet demand is surging because reasoning models and inference software have gotten dramatically better at using older silicon. A 2022-era GPU, properly optimised, is suddenly more valuable than the depreciation curve suggested. For anyone building with AI infrastructure, that changes the unit economics.

The practical takeaway splits two ways. If you're building AI applications for developers, the NexaAPI story matters: image inference at $0.003 per call removes the hardware problem entirely. A startup can run the same neural networks Tesla spent billions building custom silicon for-just via API. If you're operating data centre infrastructure, the climbing H100 prices signal that the market believes in these systems for years to come.

What Builders Should Watch

The robotics pipeline shows three patterns. First, the deployment focus is shifting from controlled environments (factories, warehouses) to public-facing spaces (airports, cities). That demands different engineering: safety margins, social interaction, graceful failure. Second, the engineering is maturing around specialisation-hospitality robots aren't manufacturing robots. Third, and most important: these deployments need reliable satellite connectivity in remote areas. BVLOS (beyond visual line of sight) drone operations can't happen if the communications link fails, and terrestrial networks don't reach where the work happens (offshore wind farms, pipeline monitoring, remote infrastructure).

For builders, the signal is clear. The abstractions are hardening. APIs are handling complexity you used to solve yourself. Deployment is moving from niche to normal. That's when careers get interesting-when the frontier shifts from "can we do this?" to "how do we do this well, at scale, without breaking things?"