Jensen Huang just published his seventh blog post in a decade, and the timing tells you everything about what he's seeing in the buildout numbers.
The Signal
Nvidia's CEO doesn't write long-form essays when things are going according to plan. He writes them when he needs to shape the narrative, and right now that narrative is "we're still in the first inning." This matters because Huang has access to demand signals no one else sees. Every AI lab, every hyperscaler, every sovereign AI project flows through Nvidia's order books. When he says the buildout is "still in the early stages," he's not cheerleading. He's reading forward indicators.
His framing of AI as "essential infrastructure" rather than software is the tell. Infrastructure buildouts don't follow app adoption curves. They follow decades-long capital deployment cycles. Roads, power grids, telecom networks. Huang is positioning AI capacity the same way, which means he's seeing sustained enterprise commit, not pilot fatigue. The five-layer stack he describes (chips, systems, software, models, applications) maps to a supply chain that's still being assembled, not optimized.
The jobs argument is predictable CEO-speak, but it's worth noting what he didn't say. No mention of displacement timelines. No acknowledgment of the coordination problem between "productivity creates growth" and "your job will be different." He's selling the buildout, not solving for the 18 months between automation and whatever comes next.
The Implication
If Huang is right about early innings, the next 24 months are about infrastructure capacity, not application innovation. Watch datacenter construction, power purchase agreements, and cooling technology deals. That's where the real capital is moving. For workers, "your job will be different" is now the official line from the man enabling the change. Start treating that as operational reality, not a future state.