You know how every March, San Jose turns into this weird tech pilgrimage site? That’s happening right now. NVIDIA’s GTC is here, and with 30,000 people crammed into convention halls, it’s hard to ignore .
Jensen Huang stood on stage at the SAP Center and basically said: we’re done with the chatbot era. AI is now infrastructure. Not a tool you use, not a feature in your software—the actual foundation that everything else sits on top of .
The announcement that matters most is Feynman. It’s their next architecture after Rubin, and it’s built specifically for inference. That’s the part of AI that actually runs the models after they’re trained. TSMC’s 1.6nm process, optical communication to cut data center power draw, and a 2028 target . The timeline feels distant, but in chip years, that’s practically next week.
Meanwhile, Tesla dropped a grenade over the weekend. Terafab—their in-house chip fab project—launches in seven days . They’re designing AI5, their fifth-gen chip, aiming for 50x the performance of what they have now. Think about that number for a second. If they hit it, everything changes. Cars that drive themselves, robots that actually work, data centers that don’t need NVIDIA.
The contrast is worth sitting with. NVIDIA is building the infrastructure for everyone else. Tesla is building its own parallel universe where they control everything from the silicon up.
Oh, and the energy people are finally waking up. There’s this concept floating around called “Neo Energy”—basically treating power grids like software platforms instead of dumb wires. The idea is that if fintech could reinvent banking without building better bank branches, energy can reinvent itself without just building more power plants . It’s about data, not just electrons.
#GTC2026, #NVIDIA, #AIHardware, #Tesla, #Terafab, #Semiconductors, #TechTrends, #Feynman, #EnergyTech, #AIIndustry
