Even the Nvidia Tree Doesn't Grow to the Sky

On November 30, 2022 OpenAI's ChatGPT provided to the general public an inkling of what Artificial Intelligence can do, and Nvidia stock. then $16.91, has never looked back. NVDA is the clear-cut leader in making the fastest chips that power AI models, and the largest hyper-scalers like Microsoft and Google regularly order tens of $billions of Nvidia's most advanced chips every year.

One week ago China's DeepSeek released two products, DeepSeek-R1-Zero and DeepSeek R1 (aka "Reasoner"), that showed that it is possible to produce AI that is nearly equal in performance to the most advanced models with lesser chips. The stock market reckoning occurred earlier today.

The tech-heavy Nasdaq fell 3.1%, driven by a 16.9% dive in Nvidia shares. Nvidia dominates the market in advanced AI chips. Its stock had surged more than 10-fold since early 2023—achieving a more than $3.3 trillion market valuation until Monday—as tech giants announced hefty outlays on AI.

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.

More startling, DeepSeek required far fewer chips to train than other advanced AI models and thus cost only an estimated $5.6 million to develop. Other advanced models cost in the neighborhood of $1 billion. Venture capitalist Marc Andreessen called it “AI’s Sputnik moment,” and he may be right.

DeepSeek is challenging assumptions about the computing power and spending needed for AI advances. OpenAI, Oracle and SoftBank last week made headlines when they announced a joint venture, Stargate, to invest up to $500 billion in building out AI infrastructure. Microsoft plans to spend $80 billion on AI data centers this year.

If DeepSeek's numbers can be believed, its R1 costs less than one percent (1%) of other models. Even if this was an exaggeration, the cost differential is so vast that Nvidia's valuation fell $589 billion today.

Chips have progressed so rapidly that the industry grew lazy. Why program efficiently when you can "brute force" solutions through advancing hardware? DeepSeek proved that paying attention to software practices can be enormously productive.

Game on!