China’s artificial intelligence lab DeepSeek vaporized $586 billion and counting from Nvidia‘s (NASDAQ:NVDA) market capitalization in midday trading Monday.
The unveiling of its DeepSeek R1 large language model (LLM) using inferior Nvidia chips to achieve superior results to comparable models on the market, and doing so at dramatically lower costs, upended the AI market.
Nvidia stock, which opened the day at a 10% deficit, continued falling well into the afternoon, and is down 17% at 2:00 p.m. The semiconductor stock went from being the most valuable stock to third place behind Apple (NASDAQ:AAPL) and Microsoft (NASDAQ:MSFT).
What happened
Because of the Biden administration’s ban on high-end technology to China, Nvidia developed a one-off of its popular and powerful H100 AI chips that power most data centers and hyperscaler operations.
Nvidia significantly reduced chip-to-chip data transfer rates on the H800 making it much less powerful. Yet DeepSeek was able to incorporate FP8 mixed precision training and leveraged low-level PTX instructions specific to Nvidia CUDA graphic processing units (GPUs) to make its model work.
It was a massive revelation. The original DeepSeek V3 model performed comparably OpenAI’s GPT-4o and Anthropic’s Claude Sonnet 3.5. It handily beat Meta’s Llama 3 as well as Alibaba’s (NYSE:BABA) Qwen2.5. The latest R1 model outperformed OpenAi’s late ChatGPT o1 model across various benchmarks to immediately become the industry frontrunner.
What it means for the stock
What made DeepSeek even more remarkable was its cost. Trained on just 2,048 H800 GPUs, it took about one-tenth the computing power that Meta Platforms (NASDAQ:META) comparable Llama 3.1 model took to train, which required 16,384 of the more expensive H100 GPUs. It was estimated to be 27 times cheaper to train as well.
That’s a big problem for Nvidia as it calls into question its pricing power going forward as well as the demand for its soon-to-be-released Blackwell chips. The chipmaker was already having difficulty with the advanced accelerators overheating, which delayed its release last year and caused customers to cancel or change orders this month.
The stock carried a premium on the basis of the money flowing into AI and the demand for Nvidia chips, but the market is seemingly resetting its outlook.
Key takeaway
On the one hand, the Chinese AI app surprised everyone, which accounts for Nvidia’s stock plummeting. It wasn’t alone as the entire AI sector had all the oxygen sucked out of the room and over a trillion dollars worth of value vanished.
Yet as Nvidia has been the face of the AI boom and the focus for much of the spending in the space, it’s not surprising it was one of the worst hurt.
Still, hyperscalers like Amazon (NASDAQ:AMZN), Alphabet (NASDAQ:GOOG)(NASDAQ:GOOGL), and Meta will continue to demand Nvidia’s latest, most powerful chips.
It will take some time for the dust to settle and the market to make sense of what this means for AI, but Nvidia might not be able to command the sort of lofty valuations it previously enjoyed.
Top Gaining Stocks
Top Losing Stocks