The U.S. stock market is facing a historic shock today, with Nasdaq 100 futures down by over 1,000 points, marking their biggest drop since September 2022. Over 350,000 Nasdaq futures contracts traded pre-market—three times the 30-day average. Nvidia, a leader in AI chip manufacturing, is set to lose over $400 billion in market cap—the largest single-day loss for a company in history.
Retail investors purchased a record $562.18 million in NVIDIA stock during Monday’s DeepSeek-led selloff, according to Vanda Research.
What’s driving this chaos? DeepSeek, a Chinese competitor to ChatGPT, has taken the AI world by storm. Remarkably, it was developed in under two months and is outperforming ChatGPT in multiple benchmarks, including AIME, MATH-500, and GPQA.
DeepSeek’s Rapid Rise
In just a month, DeepSeek became the #1 app in the App Store in both the U.S. and China.
Search interest for DeepSeek has surged, now nearly triple the levels of ChatGPT and Gemini.
The app moved from outside the top 100 list to dominating the global market.
Unmatched Efficiency and Cost
DeepSeek has proven to be a game-changer in terms of cost and efficiency:
It costs an estimated $5.8 million to train the model. Even with development costs of $2 billion, it’s far cheaper than OpenAI’s $6.6 billion fundraising round.
DeepSeek operates with just 200 employees compared to OpenAI’s 4,500.
Output costs are significantly lower, with DeepSeek R1 charging $2.19 per 1 million output tokens versus ChatGPT’s $60.
Why Is Nvidia Falling?
Despite Nvidia’s role in producing AI chips, markets are worried about DeepSeek’s ability to develop stronger AI with weaker, less advanced chips. This raises concerns about Nvidia’s future dominance in the AI sector.
Nvidia’s growth relies on AI driving high GPU demand, backed by its premium hardware and software like CUDA. But DeepSeek’s tech, using 10x less GPU power, could challenge this. If big firms like Microsoft cut GPU orders, Nvidia’s high valuation may be at risk.
Some see DeepSeek as a “black swan,” potentially disrupting Nvidia’s growth if it slows AI investments. Others argue cheaper AI could boost adoption and GPU demand, following Jevons Paradox, where lower costs lead to more usage overall.
Elon Musk and Alexandr Wang suggest that DeepSeek actually has around 50,000 NVIDIA Hopper GPUs, not the 10,000 A100s they claim, due to U.S. export controls.
Update
DeepSeek has limited registration to its services, allowing only users with a mainland China mobile phone number to sign up.
Chinese startup DeepSeek announced on Monday that it is temporarily restricting registrations following a large-scale malicious attack on its services.
Cantor Fitzgerald: DeepSeek V3 is highly bullish for compute and $NVDA:
“The release of DeepSeek’s V3 LLM has sparked concerns about reduced compute demand and peak GPU spending. However, we believe this view is misguided. The announcement signals AGI is closer to reality, and Jevons Paradox suggests the AI industry will likely demand more compute, not less. We recommend buying NVDA shares on any dips.”
The emergence of DeepSeek has reshaped the AI landscape, posing a serious challenge to established players like OpenAI, both in performance and cost. As competition heats up, the financial markets are grappling with the potential implications of this rapidly evolving technology.
JPMorgan (Sandeep Deshpande): Thinks the AI investment boom might be overhyped and that DeepSeek’s efficiency could lead to a more streamlined future.
Jefferies (Edison Lee): Offers two options after DeepSeek: either keep investing in computing power or focus on efficiency, possibly cutting AI spending in 2026.
Bernstein (Stacy Rasgon): Believes the market panic is overblown and keeps a positive outlook on Nvidia and Broadcom.
CITI (Atif Malik): Sees challenges for US dominance but highlights the benefit of access to advanced chips, keeping a buy rating on Nvidia.
Raymond James (Srini Pajjuri): Suggests that if DeepSeek’s technology is adopted, training costs could drop, reducing the need for large GPU clusters.
Cantor (C.J. Muse): Sees DeepSeek’s developments as positive for the demand for computing power and believes GPUs will still be needed, recommending buying Nvidia on dips.
