AI incumbents wrestle with their DeepSeek hangoversAI incumbents wrestle with their DeepSeek hangovers

The speculative bubble around generative AI suffered a major puncture yesterday, leaving its exponents to assess the damage done by the arrival of DeepSeek.

Scott Bicheno

January 28, 2025

2 Min Read

Nvidia, the US chip company that provides much of the processing hardware used to train the current generation of AI models, suffered by far the largest drop in market capitalisation, by company value, in history. Plenty of other chip and tech stocks suffered a similar fate, notably Broadcom, Marvell, TSMC and Oracle, prompting ubiquitous journalistic use of the word “wiped”.

The timing is a bit odd, however, since the news that drove the sell off had been in the public domain for a week. It seems the arrival of DeepSeek at the top of the app charts finally broke through the complacency. Nvidia shares were up 5% in pre-market trading at time of writing as investors recovered their composure and looked for bargains. That should bring its market cap back over the $3 trillion mark so let’s not over dramatise the matter.

View post on X

Indeed, Nvidia was sanguine and magnanimous when questioned on the matter by CNBC. “DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling,” an Nvidia spokesperson told CNBC. “DeepSeek’s work illustrates how new models can be created using that technique, leveraging widely-available models and compute that is fully export control compliant.” That last bit feels like a bit of a dig at US government meddling in the global chip business.

On the whole, DeepSeek’s achievements are being treated with cautious respect. While RFM notes that “DeepSeek’s techniques need to be independently verified,” it concedes that “it is remarkable that DeepSeek claims that its model cost only $6m to train compared to the $500m that OpenAI is thought to have spent training GPT-4o.” It’s the money side of things that has freaked everyone out and which threatens so many current AI business models.

Some excellent analysis is already out there. In addition to the pieces we highlighted previously there’s this from VentureBeat and this from the generally reliable Ben Thompson. The matter of economics remains insufficiently explained and it would be foolish to rule out CCP shenanigans. The Chinese state has a massive incentive to create the impression that US sanctions are at least futile, if not self-harming, but Global Times showed restraint in amplifying President Trump’s appreciation of the DeepSeek development.

We’ll leave you with some more contemplative responses from the US side of things, as they wait for the metaphorical Alka Seltzer to kick in. We haven’t been able to find anything from Elon Musk, however, which is less surprising when you consider the exceptional caution he has about upsetting China, where he does a lot of business.

View post on X

View post on X

View post on X

View post on X

About the Author

Scott Bicheno

As the Editorial Director of Telecoms.com, Scott oversees all editorial activity on the site and also manages the Telecoms.com Intelligence arm, which focuses on analysis and bespoke content.
Scott has been covering the mobile phone and broader technology industries for over ten years. Prior to Telecoms.com Scott was the primary smartphone specialist at industry analyst Strategy Analytics’. Before that Scott was a technology journalist, covering the PC and telecoms sectors from a business perspective.
Follow him @scottbicheno

Subscribe and receive the latest news from the industry.
Join 56,000+ members. Yes it's completely free.

You May Also Like