The cost of the GPUs massively dwarfs the cost of the electricity. GPUs have a short lifetime of a few years before the next generation makes them mostly obsole

The cost of the GPUs massively dwarfs the cost of the electricity. GPUs have a short lifetime of a few years before the next generation makes them mostly obsolete. Therefore it makes sense to use even very expensive electricity to run GPUs. Thus Musk insanely running generators to keep his new DC running. A little money can solve really really bad planning. And these SMRs, if ever built, will be very expensive electricity, a long long time from today. Investing in nuclear has nothing to do with satisfying current AI power needs. This announcement aims to have 5GW running in 15 years. A small amount of energy too far away. When has any tech company planned for energy 15 years away? Or any other specific R&D need 15 years away? Never. These announcements are about vibes and politics. I hope it means that the nuclear industry, for once in its history, actually delivers on its promises and doesn't fail. But the history of nuclear is to fail fail fail at every opportunity, so I'm not optimistic. Also, note the investment size, $500M. Basically nothing. That would buy a DC a bunch of power right now, in the form of solar and the batteries to convert the solar to constant output. Building the 5GW of SMR nuclear is going to cost a minimum of $50B, probably $100B or $150B. It's not going to be competitive with today's prices on solar and batteries. The nuclear purchase is all virtue signalling, not based in actual needs or solutions.