DeepSeek AI is next year’s nightmare for Nvidia, today

The rapid emergence and popularity of China’s DeepSeek AI suggests that there may be another way to compete in AI besides jumping into a major chips arms race.

That, if true, would be awful news for the companies that have invested all that money to enhance their AI capabilities, and also hints that those outlays might dry up before long.

As such, Nvidia and Broadcom have tanked more than 10% in early trading, with Oracle, Microsoft, and Alphabet also posting big losses.

It’s unclear exactly how much computing power and chips DeepSeek employed, but a good-enough answer is “way less than any hyperscaler.”

“Whatever the real number, DeepSeek clearly doesn’t have access to as much compute as US hyperscalers and somehow managed to develop a model that appears highly competitive,” Raymond James analyst Srini Pajjuri wrote.

The last time Nvidia suffered a double-digit drop (April 19, 2024), it wasn’t even really about the company. Results from ASML and TSMC had cast doubt on the near-term outlook for semis, but this was really more of a story of the divide between AI vs. ex-AI demand in the space (which persists to this day). The causal factors behind this tumble are of a much more pointed, direct nature when it comes to the magnitude and longevity of the AI spending boom.

Jefferies analyst Edison Lee flagged that tech companies may choose between one of two approaches going forward:

“1) still pursue more computing power to drive even faster model improvements, and

2) refocus on efficiency and ROI, meaning lower demand for computing power as of 2026.”

So at the very least, the emergence of DeepSeek should be casting lots of doubt on 2026 capex estimates tied to AI, and perhaps a flicker of skepticism regarding current-year spending, as well. That’s where Nvidia — and, given its immense weight in many benchmarks, stocks generally — appears vulnerable. Earnings for the $3 trillion chip designer are forecast to grow about 140% this year, and then by over 50% next year.

You have to go from what was the biggest weight in the S&P 500 at the end of last week all the way down to No. 48 to find a company that’s expected to grow earnings by even 30% in 2026 (Advanced Micro Devices).

In the near term, focus turns to the companies that will be the primary determinants of whether those lofty projections are ultimately realized. With a handful of trillion-dollar companies reporting this week, investors will be paying the most attention to what hyperscalers Microsoft and Meta have to say for themselves. Capital spending by those two companies may reach about $150 billion this year, according to commentary from their management teams. 

We wrote this on AI-linked capex in our top five charts to watch for 2025:

“Right now, a shorthand summary of investors’ view is that this is a case of throwing good money after good. This raises the risk that a negative turn in how much companies are willing to spend building out these new capabilities coincides with a more pessimistic view on the returns that will be generated from these capital outlays.”

DeepSeek has seemingly opened up the realm of, “Could we deliver a similar outcome (and returns) with much lower investment intensity?”

The platform’s pricing, which is 20x to 40x cheaper than OpenAI per Bernstein chip analyst Stacy Rasgon, suggests that high adoption, rather than quick commercial viability, is the priority. 

Color me skeptical that the executives who have already dropped tens of billions on AI will be quick to publicly second-guess and pivot from their current courses. Nonetheless, they’ll be challenged to answer questions on how much their end goal (artificial general intelligence) differs from what DeepSeek has been able to produce, why this pursuit will prove more commercially viable, and whether or not this can be achieved with more subdued capital outlays.

“While the model is impressive and it will have a ripple impact, the reality is that Mag 7 and US tech is focused on the AGI endgame with all the infrastructure and ecosystem that China and especially DeepSeek cannot come close to in our view,” Wedbush analyst Dan Ives wrote, deeming this sell-off to be a golden buying opportunity. “The focus of AI right now is the enterprise use cases and broader infrastructure propelling this $2 trillion of capex over the next three years.”

Chip-stock bulls — along with industry bigwigs like Microsoft CEO Satya Nadella — are left hanging their hats on Jevons Paradox. In the 1860s, British economist William Stanley Jevons penned “The Coal Question,” in which he outlined how efficiency gains don’t cause us to use less of something, but rather more: “It is wholly a confusion of ideas to suppose that the economical use of fuel is equivalent to a diminished consumption. The very contrary is the truth.” 

In this case, it doesn’t matter if you can do more with fewer chips. That still means even more chips!

“As semi analysts we are firm believers in the Jevons paradox (i.e. that efficiency gains generate a net increase in demand), and believe that any new compute capacity unlocked is far more likely to get absorbed due to usage and demand increase vs impacting long term spending outlook at this point, as we do not believe compute needs are anywhere close to reaching their limit in AI,” Bernstein’s Rasgon wrote. Rasgon is maintaining outperform ratings on Nvidia and Broadcom.