Tech stocks have tumbled to the tune of $1 trillion in the wake of DeepSeek launching its new budget Artificial intelligence.

The Chinese firm sent shockwaves through Silicon Valley last week when it launched its free AI assistant which it says was trained at a fraction of the cost of current models on the market and using significantly less chips – approximately 2,000.

On Monday (27 January), DeepSeek’s AI overtook its rival ChatGPT to become the most downloaded free app on the Apple Store in the US.

The development, which the firm claims has performed better than or close to leading Western models, has left investors questioning the levels of spending needed for AI development.  According to DeepSeek, one of its models cost $5.6m to train compared with the $100m to $1bn that is currently the consensus figure for building equivalent models. 

Stocks tumble

In light of these revelations, the value of tech firms in the US markets was slashed during Monday’s early trading.

The S&P 500 slumped more than 2% and the tech-heavy Nasdaq dropped by 3.5%.

The heaviest hit was those at the forefront of AI development, including the chip-manufacturing giant Nvidia – which recorded the greatest one-day loss in US stock market history. Early trading saw its stocks drop by 13% but were as low as 17% during afternoon dealing, a move erasing around $600bn from its market value.

Similarly, other chipmakers, such as Arm, Broadcom and Micron recorded substantial declines in value.

Oracle and SoftBank, two of the leading partners in Stargate, a new Texas-based joint venture alongside OpenAI, which has pledged an initial $100bn investment to build AI data centres, also saw their stocks fall.

In the US, Oracle was hit by an 8% drop while in Tokyo SoftBank shed the same value. 

What is DeepSeek?

DeepSeek is an AI start-up that grew out of the AI unit of the Chinese hedge fund High Flyer-Quant and is led by the fund’s manager, Liang Wendeng.

In essence, the AI functions much like ChatGPT and other large-language models (LLMs) and users can interact with the bot by typing text or uploading files and images.  

On 20 January, the firm released R1, a specialised model designed for complex problem-solving, which it claims performs on par with OpenAI’s o1-mini model released in September. 

Although the claims surrounding the cost of developing DeepSeek’s models have yet to be verified, analysts believe that the breakthrough can have a positive impact on the accessibility of AI for other start-ups – giving them the chance to compete against giants such as Google and OpenAI.

Jon Withaar, Senior Portfolio Manager at Pictet Asset Management, said: “If there truly has been a breakthrough in the cost to train models from $100m+ to this alleged $6m number this is actually very positive for productivity and AI end users as cost is obviously much lower meaning lower cost of access.”

AI arms race

DeepSeek appears to have achieved success while under the constraints of strict US export controls on advanced computing tech to China, forcing the firm to build its models with fewer and less powerful Nvidia AI chips.

Responding to the announcement, US President Donald Trump warned that the Chinese startup should be “a wakeup call” for American AI firms. However, while making an address in Florida, Trump appeared to be buoyed by the prospect of cheaper AI development for the US.

He stated: “I would say that’s a positive that could be very much a positive development. So instead of spending billions and billions, you’ll spend less, and you’ll come up with, hopefully, the same solution.”

This was a sentiment also shared by Nvidia, which responded to the sell-off in its shares with a statement suggesting that the progress demonstrated the usefulness of its chips in the Chinese market while arguing that its chips will be in even more demand in the US.

Given that DeepSeek’s innovation was borne out of necessity due to restrictions on chip exports brought in by former President Joe Biden, Wall Street analyst Jeffries has suggested that Trump may relax the US government’s AI Diffusion policy as he recognises that “the risk of further restrictions is to force China to innovate faster”.