How much electricity and investment does it take to run GPT?
Machine-assisted translation of the archived article. Read the Chinese original.

Gtp is so power-consuming, you know? We all think he's great, but what you don't know is that it's behind the energy mass.
Today we still support the steady operation of seven gtp, behind which there are more than 30,000 top-of-the-art IC chips in Britain.
These 30,000 top-level chips work together to meet the needs of tens of millions of visitors per day, at a cost of about $1.5 million per month, or almost 10 million yuan.
And the more than 30,000 chips are far more expensive than these electricity charges, and each chip is about $10,000, plus all other hardware costs.
So today OpenAI's hardware costs more than $1 billion.
So here's one thing that you can't understand, why the other gtp thing was not born in China, but in America.
It's because it's not so much technical, but it's too high a resource threshold.
Of course, there are a lot of rich companies and institutions in China, but there's too few people who can devote a billion dollars to this research.
So, to put it white is a question of probability: which country pays for this high-end business, the more companies and institutions there are, the more new technologies there are, but the more likely it will be.
