2025-01-31
43 分钟When markets opened on Monday,
the US tech industry and its investors were thrown into panic.
One company was hit particularly hard.
NVIDIA has its worst route
In history, in terms of market capitalization, ever on the market,
we have never seen almost half a trillion dollars wiped out from one single name.
Investors were responding to the release of an impressive new large-language model.
But it wasn't one developed in Silicon Valley.
It was built in Hangzhou by the Chinese firm DeepSeek.
And crucially, unlike the models that OpenAI and Meta have been pouring money into,
this one did not rely on Nvidia's extra powerful state-of-the-art chips.
A US export ban on those chips means DeepSeek has done what many thought was impossible,
to build and train a model which is up there with the best,
using comparatively basic chips.
Now the most stunning thing here isn't necessarily that China has developed a pretty good AI app,
it's how cheap it is.
DeepSeek says that their AI model only cost $5.6 million.
That's a fraction of what it takes Meta and OpenAI to train up their models,
throwing the dominance of Western firms and their chip maker of choice into question.
So how could a new generation of cheaper AI models change the world?