AI is not just prominent. What’s really worrying is that DeepSeek has made it low, very

Nothing gives a tech journalist more cheer than to see$ 600bn being wiped off the market cap of a software giant in one time. And yet, that’s what happened to Nvidia, the world’s leading manufacturer of digital hammers for the AI golden rush, next Monday. The biggest one-day decline for a company in history occurred in a group of companies that were exposed to the AI-related industries, including those that were power and infrastructure, which collectively lost more than$ 1 billion in value on the same day.

The news that a Chinese tech startup, of which some had previously heard, had released DeepSeek R1, a strong AI assistant significantly less expensive to train and operate than the US tech giants ‘ powerful models and yet was close to OpenAI’s o1 “reasoning” model caused this chaos. To illustrate the difference, R1 was said to have cost simply$ 5.58 million to build, which is a small shift in comparison to the billions that OpenAI and i have spent on their designs, and R1 is about 15 times more resource-efficient than anything else made by Meta.

The DeepSeek app immediately rose to the top of the Apple app store, drawing a sizable audience of users who were clearly unaffected by the fact that the terms and conditions and privacy policies they needed to accept were in Chinese. And it unmistakably energized Silicon Valley business. ” DeepSeek R1″, , one of the loudest mouths in California, “is AI’s Sputnik moment”. Additionally, he described it as “one of the most amazing and impressive breakthroughs I’ve ever seen,” and as a “” as well. Donald Trump, who does not believe in giving gifts to the world, described R1 as a “wake-up call” for American tech firms.

Historical resonances were frequent. Andréessen was referring to the pivotal moment in 1957 when the Soviet Union launched the first Earth satellite, demonstrating technological superiority to the US, which sparked the rise of Nasa and, in the end, the internet. The “personal computer” and the mockery imposed upon it by the then-giants of the computing industry, led by IBM and other manufacturers of large mainframe computers, were brought up in other minds. Suddenly, people are beginning to wonder if and its offspring will do to the trillion-dollar AI behemoths of Google, Microsoft, OpenAI et al what the PC did to IBM and its ilk. And of course, there are conspiracy theorists who wonder whether is really just a disruptive stunt designed by Xi Jinping to destroy the US tech sector. Is the training of the model really that affordable? Can we trust the figures in the technical reports its authors have released? And so on.

There are four things to take away from the arrival of DeepSeek, if you’re standing back.

Despite the widespread ( and hubristic ) assumption that the Chinese are not as good at software as we are, the Chinese have surpassed the leading US AI labs in two ways. Even a cursory examination of some of the R1’s and V3’s technical details reveals incredible technical genius and creativity.

Second, R1’s low training and inference costs will exacerbate American fears that the emergence of powerful, affordable Chinese AI could change the industry’s economics in a similar way that the PC’s and 1980s changed the computing industry’s landscape. What the development of DeepSeek indicates is that this technology will eventually become commoditized, like all other digital technology. R1 runs on my laptop without any interaction with the cloud, for example, and soon models like it will run on our phones.

Third, in spite of the fierce technology bans imposed by the first Trump administration and then by Biden, DeepSeek managed to pull this off. According to the company’s technical report, it possesses a cluster of 2, 048 Nvidia H800 GPUs, a technology that has been officially prohibited for sale to China by the US government.

And last, but by no means least, R1 seems to be a genuinely open source model. It’s distributed under the permissive MIT licence, which allows anyone to use, modify, and commercialise the model without restrictions. As I write this, my hunch is that geeks across the world are already tinkering with, and adapting, R1 for their own particular needs and purposes, in the process creating applications that even the makers of the model couldn’t have envisaged. It goes without saying that this has its upsides and downsides, but it’s happening. The AI genie is now truly gone.

What I’ve been reading

When Trump meets tech
A by William Cullerne Bown of what the new regime in Washington means for the UK and Europe.

A dystopia like Philip K Dick’s
explaining why Henry Farrell thinks that our future might be like something written by the great author.

More than just an engineering issue, life is.
A transcript of Ted Chiang’s fascinating interview for the LA Review of Books.

Leave a Comment