The new AI trade emerging after DeepSeek shock
Add Axios as your preferred source to
see more of our stories on Google.
The AI bubble didn't burst on Monday.
Why it matters: What initially looked that morning like a broad selloff in AI stocks turns out in hindsight to have been a more measured recalibration of where the market thinks AI profits are going to flow in the future.
The big picture: The consequences of AI becoming the dominant driver of the global economy are impossible even for a super-intelligence to model, but are certainly more complex than the naive yet highly profitable belief that "Nvidia will be able to retain 90% gross margins more or less indefinitely."
- Nvidia was the defining meme stock of 2024, powered by an army of retail investors working hand-in-glove with their momentum-trade brethren.
- Profitable trades don't last forever, however. The traders always move on eventually, often after having made enormous sums of money.
Follow the money: Nvidia closed Thursday with a market capitalization north of $3 trillion, putting it comfortably in the top three most valuable companies in the world. It's worth 70% more than Saudi Aramco and 24% more than either Alphabet or Amazon.
- That's not what a burst bubble looks like.
Between the lines: The best guide to the highly complex world of AI economics is probably the instantly famous "short case for Nvidia" note that analyst Jeffrey Emanuel first posted on Saturday.
- While DeepSeek certainly features in Emanuel's 12,000-word thesis, it's a great deal more complex than the "DeepSeek has singlehandedly disrupted the world of AI" idea that gripped headline writers on Monday.
- It's worth point out that the sell-off came a full week after DeepSeek's R1 AI model was released, and more than a month after the release of the V3 model that competes directly with the likes of OpenAI's GPT-4o.
- Emanuel's thesis is not that DeepSeek on its own is particularly disruptive. Rather, the Chinese AI is a proof of concept: It shows how competitive the space can be, and how much opportunity there is for a broad range of players.
Zoom out: Many companies, including Amazon, Apple and even OpenAI, are investing heavily in trying to build chips that can compete with Nvidia. And while Nvidia's competitive moat has proved extraordinarily strong until now, there are many reasons to believe it will be eroded over time, even by AI tools developed using Nvidia chips.
- While Nvidia is extremely profitable at the moment, the big question is how long it can maintain that profitability in the face of competition from the biggest and best-funded tech companies on the planet, all of which have similar access to TSMC, the company that makes Nvidia's chips.
- Even small changes to the growth rate and discount rate plugged into a discounted cashflow model can easily produce a trillion-dollar decrease in Nvidia's present market value.
Where it stands: AI stocks in general didn't have a particularly bad day on Monday, and they certainly didn't have a particularly bad week.
- The market still believes the likes of Meta and Microsoft will spend record amounts of money this year on new data centers filled with Nvidia chips, and is now hopeful those chips will prove to be even more efficient and powerful than originally believed.
- Meta stock in particular has looked very healthy. After its earnings report Wednesday, it closed Thursday at $687 per share, up an impressive 6.1% from where it closed last Friday.
- While it's impossible to calculate exactly how much of Meta's $1.7 trillion valuation can be attributed to AI optimism, it's surely a large chunk.
The bottom line: If this is what disruption looks like, the AI revolution is going to be remarkably painless.
