Anthropic buys 1.7 GW of capacity: AI's bottleneck is now measured in energy, not chips

By FourWeekMBA · November 8, 2026.

FourWeekMBA's daily roundup opens with two headlines: Anthropic buys 1.7 gigawatts (GW), and Nvidia's board opens an investigation. From the available material we only know those headlines. We do not have the counterparty to the deal, the timeline, the price or the reason for the Nvidia investigation. So we do not treat them as established and stick to what can actually be said.

A gigawatt is a thousand megawatts. For reference, 1.7 GW is roughly the output of a large power plant. That an AI lab talks about capacity purchases in that unit shows energy has become its strategic input.

Our view is that competition among labs has shifted from algorithms to infrastructure. A 1.7 GW move fits a logic of long-term commitments. The winner is whoever secures electricity and data centers before the others. The loser is whoever arrives late and has to pay more or wait.

There is a short-term risk. These commitments are rigid, and the lab takes them on before knowing how much real demand there will be. There is also growing pressure on power grids and on energy prices for other users. This transition will not be free.

In the long term, that same capacity is what makes possible models that accelerate drug discovery and scientific research, which is where the promise of abundance is at stake. Energy being the limiting factor also pushes toward producing it cheaper and cleaner, and that benefits everyone.

As for the investigation at Nvidia's board, there is not enough information to assess it. We will wait for the details before weighing in. If it proves relevant to chip supply for the labs, we will come back to it.