[–] Ringz 14 hours ago
[–] boringg 15 hours ago
[–] Zigurd 14 hours ago
Bloom fuel cells are a plausible alternative, but they are in the process of building out their manufacturing capacity, and operating experience at data center scale hasn't been tested. Bloom's process also involves extremely high temperatures, which could result in uptime and maintenance issues.
[–] boringg 14 hours ago
I would say you end up with battery+solar and gas turbine/gensets providing power when solar+batteries cant fill the gap. Its a bit expensive to be adding so many different sources on site to shore up power supply.
FWIW obvious maintenance work required to keep turbines going.
[–] Zigurd 10 hours ago
I don't know why you think solar needs two forms of back up. Batteries are very reliable.
[–] Y-bar 14 hours ago
[–] boringg 13 hours ago
And gas/fuel has high functional use as a differentiated energy product - we just don't need to use so much of it -- keep it on the smaller end of the use case spectrum.
[–] lenerdenator 15 hours ago
[–] boringg 14 hours ago
[–] lenerdenator 13 hours ago
[–] senordevnyc 4 hours ago
[–] solyin 4 hours ago
The answer is there's a lot at that price, so then we go upwards and do price discovery to find out what applications there are and how much they're able to pay, but one thing is certain, even without LLMs there's demand for GPUs, the only question is at what price.
I think Google's first offshore purchase was 92 MW from the Norther wind farm in Belgium. TSMC will buy all the power from Ørsted's 920 MW Greater Changhua 2b & 4 offshore wind farm for the next 20 years.
There are examples in Germany but mainly for chemical and steel factories.
The best way to “force them” is a global emissions trading system + carbon levy. But his might be very unlikely in the next 10–15 years and maybe as complex as Peace in the Middle East.