On Sunday, I wrote that DeepSeek-R1 was a revolutionary, good and cheap AI product from China. But I had no idea that a day later Wall Street would react as if aliens had launched an attack on our planet.
Yesterday, the technology sector experienced a sharp downturn, to put it mildly, with the chip sector hit the hardest. Nvidia's share price fell 16.9%, resulting in a loss of $593 billion in market capitalization. Broadcom saw a 17.3% drop, accounting for a loss of $198 billion. Advanced Micro Devices (AMD) lost 6.3%, a loss in value of $12.5 billion. Taiwan Semiconductor Manufacturing Company (TSMC) fell 13.2%, down $151 billion, and shares of Arm Holdings fell 10.2%, a $17 billion hit. Marvell Technology experienced the steepest decline, losing 19.2%, a whopping $20 billion.
Apple smiling third
The drop pushes previously high-flying chipmaker Nvidia to third place in total market cap, behind Apple and Microsoft. Last Friday, it was still in first place. Apple now has the highest market value with $3.46 trillion ($3,460 billion ), followed by Microsoft with $3.22 trillion and Nvidia with $2.90 trillion. Shares of Apple, which has less exposure to AI, rose 3% Monday, while the tech-heavy Nasdaq fell 3%.
Apple's rise is similar to a house rising in value because the neighbor's roof is on fire. Intrinsically, of course, nothing has changed in Apple's value, and a trade war with China still lurks, which would hit Apple hard. Bur that's for another day.
DeepSeek hits Wall Street
Investors blame the sell-off on the rise of DeepSeek, a only one year old Chinese company that last week unveiled a revolutionary Large Language Model (LLM), named DeepSeek-R1. DeepSeek's model is similar to existing models such as OpenAI's ChatGPT 4o or Anthropic's Claude, but is said to have been developed at a fraction of the cost. It also costs a fraction for customers compared to ChatGPT and Claude.
This has rightly led to concerns in investor circles that the U.S. strategy of heavy investment in AI development, often referred to as a "brute force" approach, is becoming obsolete. This brute force method uses extensive computing power and large data sets to train AI models, with the goal of achieving higher performance due to its massive scale. It is a billion-dollar approach that I wrote about earlier.
DeepSeek 'the Sputnik moment for AI'
A Wall Street Journal editorial clearly summarizes the competitiveness of DeepSeek-R1 with a catchy example:
"Enter DeepSeek, which last week released a new R1 model that claims to be as advanced as OpenAI's on math, code and reasoning tasks. Tech gurus who inspected the model agreed. One economist asked R1 how much Donald Trump's proposed 25% tariffs will affect Canada's GDP, and it spit back an answer close to that of a major bank's estimate in 12 seconds. Along with the detailed steps R1 used to get to the answer."
Venture capitalist and former entrepreneur (Netscape) Marc Andreessen described the launch of DeepSeek-R1 as the Sputnik moment for AI; similar to the moment the world realized the Soviet Union had taken a lead in space exploration.
OpenAI reacts anxiously
OpenAI CEO Sam Altman put on a brave face:
"DeepSeek's R1 is an impressive model, particularly around what they're able to deliver for the price. We will obviously deliver much better models and also it's legitimately invigorating to have a new competitor! We will pull up some releases."
(I myself took the liberty of inserting the capital letters Altman avoids, because otherwise I find it too annoying to read.)
Altman puts on a brave face, but in the last sentence it appears that OpenAI is accelerating product releases under pressure from DeepSeek. Or without capital letters just like him: altman blinked. Legit, you know, bro.
Sweat-soaked body warmers
They need to exhale and tuck in their body warmers up: a
on Wall Street. The claim that DeepSeek developed their R1 model with only a $5 million investment is not verifiable, and the Chinese media are not known for their transparency, nor for their critical approach to Chinese initiatives.
Western companies are unlikely to adopt Chinese AI technology, with all the geopolitical tensions and regulatory constraints, especially in critical sectors such as finance, defense and government. Chief Information Officers are increasingly cautious about integrating Chinese technology into critical systems. In fact, it only happens now if there is no other alternative.
Despite recent market volatility, major technology companies such as Microsoft, Google, Amazon and Oracle continue to rely on high-performance chips for their AI initiatives. No company is canceling its orders with Nvidia because DeepSeek has a different approach.
Because there is currently no Western equivalent of DeepSeek's R1 model that is fully open-source (as opposed to "open-weight," a term I discussed in my Sunday edition ), these companies will continue to invest in expensive hardware and huge data centers.
This means that shareholders in companies such as Nvidia and Broadcom can expect a recovery in stock prices in the coming months, perhaps weeks.
Panic at the vc's
The real impact of DeepSeek's innovation will likely be felt more profoundly by venture capital funds that have poured billions into AI startups without clear revenue models or, as VCs always so delightfully know how to put it from the comfort of their armchairs: a clear path to profitability.
Earlier I highlighted the precarious financial situation of OpenAI, which seems headed for a $15 billion loss this year: that's $41 million per day, $1.7 million per hour and $476 per second. Partners at Lightspeed, which last week invested $2 billionin Anthropic, the developer of DeepSeek-R1 competitor Claude, at a valuation of as much as $60 billion, will have slept terrible last night.
Does DeepSeek dare a frontal attack?
DeepSeek's approach to AI development, especially its emphasis on efficiency and the possibility of local implementation, running the model on your own computer, is remarkable. But globally, the AI community can only benefit from this methodology if DeepSeek chooses to release its underlying code and techniques. So far, DeepSeek-R1's codebase has not been made public, raising questions about whether it will ever be. It's the difference between open-weight and open-source I wrote about on Sunday.
If China decided to fully open-source DeepSeek-R1, it would pose a massive challenge to the U.S. tech industry. An open-source release would allow developers worldwide to access and build upon the model, greatly reducing the competitive advantage of U.S. companies in AI development.
This could lead to a democratization of advanced AI capabilities, reducing reliance on closed models such as from OpenAI and Anthropic and expensive infrastructure such as from Nvidia, Oracle, Microsoft and Amazon Web Services. Such a move would totally disrupt the current market dynamics and force U.S. companies to completely change their strategies in funding AI research and development.
China rules, Wall Street pays
How big an impact the technology sector has on the U.S. economy was once again evident yesterday when the total loss on Wall Street was estimated at a trillion dollars: a staggering thousand billion dollars.
It leads to the ironic conclusion that the first week of "America First" President Trump ends with a moment when China can determine whether to throw the U.S. economy into disarray. Wall Street is now watching every move from Beijing like a dear in the headlights.
