@FinanceYF5: Model companies want to slow down? First, take a look at the computing power bill. Oguz Erkan's statistics show that An…

X AI KOLs Following News

Summary

This article discusses the high computing power costs of AI companies like Anthropic and OpenAI, with estimates indicating that their annual expenses could reach 115–120 billion USD and long-term spending over a trillion USD.

Model companies want to slow down? First, take a look at the computing power bill. Oguz Erkan's statistics show that Anthropic's announced computing power capacity exceeds 11.5GW, but the current actual operational estimate is only 1–2GW. According to his calculations, once fully built, it could cost 115–120 billion USD annually. If OpenAI also expands on a similar scale, the two companies' computing power spending over the next five years could reach 1 trillion USD. Slowing down—how are they going to pay this bill?
Original Article

Similar Articles

When AI Costs More Than the Engineer

Hacker News Top

Analysis of AI compute spending vs engineer salaries, showing Anthropic spends 2.3x payroll on compute while most companies spend far less, with scenarios projecting costs through 2029.

@FinanceYF5: How Anthropic used less than $9 billion in annualized revenue to secure nearly $50 billion in AI infrastructure. 1/ When Anthropic announced its $50 billion US AI infrastructure plan, its annualized revenue was still under $9 billion. Then, it secured nearly $50 billion in debt financing for over 1GW of TPU and 5 data centers. This…

X AI KOLs Timeline

Anthropic announces a $50 billion AI infrastructure plan, leveraging under $9 billion in annualized revenue to obtain nearly $50 billion in debt financing, highlighting that long-term contracts and credit support are critical bottlenecks in AI expansion.

Can tech companies learn to love cheaper AI models? 

TechCrunch AI

TechCrunch reports on a potential industry shift as companies consider switching to cheaper, smaller AI models instead of always using the most powerful ones, driven by escalating costs. Predictions like Brian Armstrong's suggest 80% of workloads could run on 99% cheaper models within 12-18 months, which would significantly impact major AI labs like OpenAI and Anthropic.