It’s almost exactly one year since DeepSeek arrived and showed the possibility that Large Language Models didn’t have to go the expensive American way of brute force compute. The revelation led to Nvidia making the single day biggest US stock market loss in history.
The release of plugins in Claude Cowork unleashed similar shocks last week. RELX, which owns the legal data analysis and information company Lexis Nexis, saw a drop of 15% on its share price in a day. Many other software and data companies saw sharp falls last week.
What went down
Claude Cowork plugins were released with little fanfare (it wasn’t even announced on the Anthropic newsfeed) on Friday 30 January, triggering a software market selloff the following Monday.
The plugins covered productivity, enterprise search, marketing, sales, data, customer support, the ability to build your own plugins, finance, legal, product management and biology research.
The software selloff was compounded by sudden selling of gold and silver - both had reached record highs in the face of geopolitical uncertainty.
Big Tech companies also reported quarterly results, which showed mega investment in AI infrastructure and further fears of an AI bubble - which led to even more selling. Chip makers also didn’t escape, with AMD losing -15% on Tuesday.
Crypto was hit bad. Bitcoin wilted to -35% of its October peak. Most of the market is considerably lower than last April, when Trump’s ‘Liberation Day’ tariff announcements caused mass market panic.
Why does this matter?
It’s an example of market contagion created by AI product launches. There were many interconnected reasons why markets fell last week - but the Claude plugins release was a major trigger.
Large Language Models can now code with human life proficiency. So called ‘Vibe coding’ means that a difficult to learn skill, and thus a software company ‘moat’ has become increasingly democratised. Other difficult to learn skills, like legal, are also going the same way.
Our take
Software as a Service (Saas) has been one of the key business models of the Web 2.0 era, which stretches back to roughly 2003. There are signals that this model is considerably challenged by AI. Rapid improvements in generative AI present a considerable headwind for any business that monetises information, which is what SaaS, publishers and data companies do.
It’s incredibly difficult to know when, where and how the so called AI bubble ends. What was surprising about this market correction is that an AI release led to existential questions being raised on AI’s effect on the wider economy. However, given the very high level of infrastructure spending by companies like Google, Meta and Microsoft, it seems unlikely that this will be the end of it. I could be wrong of course… but who knows? BlackRock’s CEO has recently dismissed AI bubble concerns.
Another big thing… Claude Superbowl ads throw shade at OpenAI
In another moment of Anthropic’s blistering start to 2026, they took specific aim at OpenAI’s recent decision to place ads in ChatGPT through an ad campaign debuting at Sunday’s Superbowl. Sam Altman put out a defensive post on X on Wednesday, accusing Anthropic of ‘doublespeak’ and calling the ads ‘deceptive’. Cue more snickering on social media and tech headlines.
