2 ms·
The issue is capital expense. Buying a bunch of expensive equipment and operating it 10 hours a day vs. 24 hours a day is usually not economical. The uses whe
by phil21 1y ago
The issue is capital expense.
Buying a bunch of expensive equipment and operating it 10 hours a day vs. 24 hours a day is usually not economical. The uses where this make sense are quite rare.
For example training AI models is often cited - but when you're buying $1B of GPUs to stuff in a datacenter that have a 3 year useful lifespan - you are effectively cutting your amortization schedule in half. It would require some really expensive energy to make that pencil out.
Energy storage of various types are probably the currently best bet, but those also have the same problem. Vehicle charging is a clear win, but also a low hanging fruit that is already well in play.
- rz2k 1y agoIn such a case, isn't the "3 year useful lifespan" almost entirely related to energy cost? Why not put machines on 24 hour schedules for their first couple years, then cheap intermittent power for the next five?