{"id":359114,"date":"2024-12-27T17:56:39","date_gmt":"2024-12-27T16:56:39","guid":{"rendered":"https:\/\/climatescience.press\/?p=359114"},"modified":"2024-12-27T17:56:41","modified_gmt":"2024-12-27T16:56:41","slug":"2025-the-year-of-the-gigawatt-ai-project","status":"publish","type":"post","link":"https:\/\/climatescience.press\/?p=359114","title":{"rendered":"2025: The Year of the Gigawatt AI Project"},"content":{"rendered":"\n
\"\"
3d rendering humanoid robot with ai text in ciucuit pattern<\/figcaption><\/figure>\n\n\n\n

<\/p>\n\n\n\n

From Watts Up With That?<\/a><\/p>\n\n\n\n

Essay by<\/strong> Eric Worrall<\/a><\/p>\n\n\n\n

\u201cthe entire industry is chasing the AI dragon\u201d \u2013 $27 billion+ for a single project. Bye bye climate action virtue signalling.<\/p>\n\n\n\n

\n

Million GPU clusters, gigawatts of power \u2013 the scale of AI defies logic<\/strong><\/p>\n\n\n\n

It\u2019s not just one hyperbolic billionaire \u2013 the entire industry is chasing the AI dragon<\/p>\n\n\n\n

Tobias Mann <\/a>
Thu 19 Dec 2024  \/\/ 17:30 UTC <\/a><\/a><\/a><\/a><\/a><\/p>\n\n\n\n

COMMENT<\/strong> Next year will see some truly monstrous compute projects get underway as the AI boom enters its third year. Among the largest disclosed so far is xAI\u2019s plan to expand its Colossus AI supercomputer from an already impressive 100,000 GPUs to a cool million.<\/p>\n\n\n\n

Such a figure seemingly defies logic. Even if you could source enough GPUs for this new Colossus<\/a>, the power and cooling \u2013 not to mention capital \u2013 required to support it would be immense.<\/p>\n<\/blockquote>\n\n\n\n

\n

At $30,000 to $40,000 a pop, adding another 900,000 GPUs would set xAI back $27 to $36 billion<\/strong>. Even with a generous bulk discount, it still won\u2019t be cheap regardless of whether they\u2019re deployed over the course of several years. Oh, and that\u2019s not even taking into account the cost of the building, cooling, and the electrical infrastructure to support all those accelerators.<\/p>\n\n\n\n

Speaking of power, depending on what generation of accelerators xAI plans to deploy, the GPU nodes alone would require roughly 1.2 to 1.5 gigawatts of generation. That\u2019s more than the typical nuclear reactor \u2013 and the big ones, no less. And again, that\u2019s just for the compute.<\/p>\n\n\n\n

\u2026<\/p>\n\n\n\n

The AI fever with which the tech giants have collectively come down has served as\u00a0a sort of sea change for the nuclear industry as a whole<\/strong>, with cloud providers fronting the cash to reinstate retired reactors \u2013 and even plop their datacenters behind the meter in the case of AWS\u2019 new\u00a0Cumulus datacenter complex<\/a>.

\u2026Read more:\u00a0<\/p>\n\n\n\n

https:\/\/www.theregister.com\/2024\/12\/19\/scale_ai_defies_logic\/<\/a><\/p>\n<\/blockquote>\n\n\n\n

Think about it: $27 \u2013 $36 billion for just\u00a0one<\/strong>\u00a0of these new AI data centers. The Register lists xAI, Meta (Facebook), AWS (Amazon), Oracle, and of course we already know about Microsoft \u2013 all of which appear to be initiating Gigascale AI projects in 2025 or have already initiated their projects.<\/p>\n\n\n\n

\n
AI Energy Surge – Microsoft to Re-Open Three Mile Island<\/a><\/blockquote>