NVIDIA loses a record amount of market value in a day
The roughly $589bn one-day fall, the largest for any US company on record, followed DeepSeek's claim that a competitive model cost about $5.6m to train.
- Culture & impact
- Money & business
- Compute & infrastructure
- Era-defining
On Monday 27 January 2025 Nvidia’s share price fell about 17%, wiping out roughly $589bn of market value in a single trading day — by most accounts the largest one-day loss for any company in US market history. The sell-off spread across chipmakers and AI-exposed stocks and was traced to the sudden prominence of DeepSeek, a Chinese lab whose model had reached the top of the US App Store over the weekend.
What unsettled investors was less the model’s existence than its reported cost. DeepSeek said one of its recent models had been trained for around $5.6m in compute, a fraction of the sums Western labs were understood to be spending; OpenAI’s GPT-4, by comparison, was reported to have cost well over $100m. If a capable model could be produced that cheaply, the reasoning went, the enormous capital expenditure that had been justifying Nvidia’s valuation — and the data-centre build-out across the industry — might not be necessary. Nvidia’s chief executive, Jensen Huang, saw his own net worth fall by roughly $21bn on the day.
The figures were contested almost immediately. DeepSeek’s stated cost covered a single training run and excluded the hardware, staff and earlier experiments behind it, so it could not be compared directly with a lab’s total spending. Some analysts also argued that cheaper training would, if anything, increase demand for chips by making AI more widely useful — an application of the observation, often called Jevons paradox, that greater efficiency can raise total consumption rather than lower it. Nvidia’s shares began recovering in the following days.
The episode, quickly labelled the “DeepSeek moment”, mattered as a market and public-attention event distinct from the R1 model release that preceded it. It was the point at which doubts about the economics of frontier AI — how much it really cost, whether the spending was sustainable, and how far export controls had actually held China back — moved from specialist debate to the front pages and the stock ticker.