A year ago Intel and Nvidia became partners. This week at Computex they were rivals again.

Nvidia invested in Intel and runs on its chips in the data center, then launched a PC chip aimed at Intel's home turf. Intel is pushing its server chips into the AI work Nvidia's GPUs used to own. Each is walking into the other's house.

Globes asked me what to make of it. The contradiction is the point. Two companies this size do not fight on every front unless they have stopped agreeing on where value will land. And the reason they disagree is that the bottleneck keeps moving. For decades it was the general-purpose processor, and Intel owned it. AI moved it to parallel compute, and Nvidia took it. Now it is splitting again: memory and cost in the data center, speed and proximity to the user on the device, and it weighs on training the models and running them alike.

Once the bottleneck no longer sits in one place, value no longer concentrates in one company. It spreads to both ends, and I think some of the largest companies of the coming years will be built there.

Grove Ventures

https://lnkd.in/dfBDqijZ