News that DeepSeek a Chinese AI startup, had developed an open-source AI model that matched the performance of U.S. AI models from OpenAI, Alphabet, and Meta platforms at a fraction of the cost sent AI stocks reeling today, Monday, January 27. Nvidia (NVDA) shares fell 17%. Broadcom (AVGO) similarly fell 17%.
The Dow Jones Industrial Average added 0.4%. A gauge of the “Magnificent Seven” megacaps slid 3.2%. The Russell 2000 slipped 1.3%. Wall Street’s “fear gauge”—-the VIX—soared 20% the most since mid-December to almost 18.The yield on 10-year Treasuries declined 10 basis points to 4.53%. The Bloomberg Dollar Spot Index rose 0.1%. Bitcoin fell 3.9% to $100,537.23.
Here’s what freaked out the markets today. DeepSeek claims that its breakthrough model cost less than $6 million to train. Compare that with OpenAI’s GPT model, which cost more than $100 million to train.
Suddenly every AI valuation is one to question. And suddenly investors and Wall Street analysts are remembering “oh, yeah, most AI companies don’t have a business models that takes them to profitability. For more detail on this problem, see Part 1 of my Special Report “10 stock picks for the 3 hottest sectors in 2025–and when to buy them: Part 1 AI.”
The timing is also likely to be contributing to the dimensions of the freak out. Big U.S. AI companies and President Donald Trump have been competing to see who can announce the biggest capital spending plans. Last week Meta Platforms CEO Mark Zuckerberg it plans between $60 billion and $65 billion in capital expenditures this year, primarily focused on expanding its artificial intelligence infrastructure. Meta is constructing a 2-gigawatt data center that Zuckerberg described as large enough to “cover a significant part of Manhattan.” The company aims to end 2025 with over 1.3 million graphics processing units (GPUs), and Meta plans to bring online approximately 1 gigawatt of computing power in 2025.
Analysts are likely to focus on the effects of all this spending when they study upcoming earnings reports from the U.S. AI giants over the next few weeks. Today’s selling will focus attention on profit margins. Whatbwill be the return on all this investment and when will it show up on the books?
The real questions that have not been answered. 1) This news release comes from a communist country, and is it reliable? 2) Why now, this model has been out for more than a month, why is the market reacting just now 3) None of the efficiency tricks are used in the model where unknown to American developers, they weren’t used as we have access to more powerful trips. 4) if they just copied (distillation) chat-gpt for $6 million is this really a break through? 5) Even if it is more efficient, is that a bad thing, today’s models aren;t that useful yet and would require a huge improvement, now its cheaper to get to the next level. This will mean more opportunities for people to use AI. This totally speaks to the psychology of the market, and the power of X.
It seems to be reactionary. I do question any “new” thing or approach off the cuff. US or China. I bought NVDA in another account and enjoyed the post gain. Let the experts analyze the approaches and look at the long term impact