Last week, chipmaker NVIDIA reported earnings that exceeded expectations, growing revenue at 69% year over year, putting to rest any fears that the AI boom propelling the company and broader market to new heights was slowing down.
To keep the rally going, the company will need to continue meeting Wall Street’s lofty expectations, but for now management and CEO Jensen Huang are being rewarded for their superb execution. At around $3.5 trillion in market capitalization, NVIDIA is once again the most valuable company in the world. Chipmakers like Broadcom and AMD are also pushing to all time highs, while large-cap tech companies, who for a brief time weighed on the U.S. market, are once again lifting it up.
Few predicted that Big Tech and AI names would stage such a dramatic comeback in such a short window. Trump’s tariff threats put enormous pressure on stocks and Big Tech was no exception, but there was a bigger and more existential risk brewing under the surface.
Earlier this year, Chinese startup DeepSeek released R1, a powerful open-source AI model reportedly trained for just $6 million—pennies compared to the billions spent by OpenAI, Google, and others. Under certain benchmarks, R1 performed competitively with GPT-4. The launch rattled nerves across Silicon Valley. NVIDIA stock dropped over 15% in a day.
The revelation of Chinese AI competitiveness raised two big concerns. First, some analysts began questioning whether American Big Tech companies and upstarts like OpenAI and Anthropic could sustain their lead in AI, or if they were just throwing billions of dollars at the problem. Second, they questioned if companies would need to spend as much on computing power and energy, throwing into question NVIDIA’s entire investment thesis as the preeminent chipmaker.
This was something we were curious about and tried to weigh the benefits and risks of DeepSeek and China's entrance into the AI race.
Figure 1: DeepSeek: Risk or Opportunity?
It is still too early to say whether China or America will win in the AI race. Perhaps we end up with two parallel, separate but roughly equal technology stacks and both American and Chinese companies benefit.
But I do think we can say now that fears about spending on chips
is overblown.
The invention of DeepSeek is a classic case of the Jevons Paradox: as the price of a resource drops, demand for it surges, leading businesses to (somewhat counterintuitively) spend
more not
less on that particular resource. By dramatically reducing the cost of cutting-edge AI capabilities, DeepSeek will probably just accelerate global appetite for AI chips, infrastructure, and computing power. We've seen this pattern before. The declining cost of computing in the 1980s and 1990s didn’t cause businesses to spend less on mainframes. It caused them to spend more on PCs for every office worker. The dramatic drop in the cost of energy during the 19
th century didn’t cause people to spend less, rather it fueled the industrial age. When something becomes more abundant and affordable, demand doesn’t drop - it explodes.
Let’s revisit Nvidia’s earnings for signs of Jevons paradox playing out. To be sure, the data center segment was still responsible for most of the revenue - nearly 90% in Q1, driven by “hyper-scalers” such as Microsoft, Amazon, Meta & OpenAI investing in AI infrastructure. But NVIDIA is seeing broader adoption – and faster growth – in other segments, such as Automotive, which grew 103% YoY, increasing to $567 million. NVIDIA’s “DRIVE” platform chips help power autonomous vehicles and advanced driver assistance systems. CEO Jensen Huang expects this segment to potential reach $5 billion in revenue by the end of the year.
And then there’s robotics. NVIDIA announced Isaac GROOT N1, designed to “accelerate reasoning and skill development” in robots. This was accomplished by collaborate with Google Deepmind and Disney Research to develop an open-source engine called Newton, claiming to accelerate robotics machine learning workloads by over 70x. The company also talked about breakthroughs in medicine and other areas.
As AI models become more performant, and as chips become more powerful, AI will become more diffused in the economy, leading to more, not less demand.