That shocked the US market, but did China's Deepshek change AI?

US President Donald Trump has been the week when the new Chinese artificial intelligence (AI) app called the Deepseek Jolted Silicon Valley was called Silicon Valley.

Overnight, Deepseek-R1 was shot at the top of the Apple chart as the most downloaded free app in the United States.

The company said at the time the new chatbot was comparable to ChatGPT. Not only that. They claimed that development was costly.

These claims, and a sudden surge in the popularity of the app, reduced the market value of $600 million (£446 billion) or chip giant Nvidia by 17%, marking the biggest loss in the history of the US stock market.

Several other high-tech stocks exposed to AI have also been caught up in downdrafts.

Deepseek also questioned the control of American AI. Until then, China was considered to have fallen behind the US. Now it seemed as if China had been attacked by the forefront.

Venture capitalist Mark Andreessen calls the arrival of the Deepseek-R1 “AI's Sputnik moment.” This is a reference to the Soviet satellite that kicked off the space race between the US and the USSR over half a century ago.

It's been six months since Deepseek surprised the world.

Today, China's groundbreaking apps have fallen largely out of the headlines. It's no longer a hot topic at Happy Hour here in San Francisco. However, Deepseek has not disappeared.

Deepseek challenged certain key assumptions about AI that were defended by American executives like Sam Altman, CEO of ChatGpt-Maker Openai.

“We were on a path we thought was bigger,” says Sid Sheth, CEO of AI Chip Startup D-Matrix.

Perhaps making the most of your data centers, servers, chips and electricity and doing it all was not a way to move forward after all.

Deepseek told the BBC that, despite ostensibly lacking access to the most powerful technology available at the time, it showed that “smart engineering can actually build a capable model.”

The surge in interest in Deepseek has now allowed corporate IT personnel to move on to stop employees from huddled up after the weekend in late January.

When the organization got caught up next Monday, many people scrambled to ban the app from using it on whether user data could be shared with the People's Republic of China, the Deepseek home.

However, although the exact numbers are not available, many Americans still use Deepseek today.

Certain Silicon Valley startups have chosen to stick to Deepseek instead of the more expensive AI models of US companies to reduce costs.

One investor told me for a cash-bound company, the money saved by continuing to use DeepSeek helps pay for important needs such as additional personnel.

But they are careful.

The online forum explains how users can run DeepSeek -R1 on their devices rather than online using DeepSeek servers in China.

“It's a good way to use the model without worrying about what it is,” Christopher Kane, CEO of Mill Pond Research, told China.

Deepseek's arrival also marked a turning point for US-China AI rivalry, some experts say.

“Until this point, China was seen as catching up with a large language model with competitive models, but it's always following the best Western models,” Wendy Chang, a policy analyst at the Mercator Institute for Chinese Studies, told the BBC.

A large-scale language model (LLM) is an inference system trained to predict the next word in a particular sentence or phrase.

Deepseek changed perceptions when he claimed to have achieved some key models of common computational resources and costs among its American counterparts.

Openai spent $5 billion (£3.7 billion) in 2024 alone. In contrast, Deepseek researchers said they developed the DeepSeek-R1, which came out on Openai's O1 model across multiple benchmarks for just $5.6 million (£4.2 million).

“Deepseek has revealed the competitiveness of China's AI landscape to the world,” Chang said.

American AI developers were able to take advantage of this shift.

AI-related deals and other announcements trumpeted by the Trump administration and major high-tech companies in the US are often surrounded as important for staying ahead of China.

Trump's AI Czar David Sacks said the technology “has a significant impact on both the economy and national security” when the administration announced its AI action plan last month.

“It's very important that America continues to be the dominant force in AI,” Sachs said.

Deepseek was unable to quell concerns about the security implications of Chinese origin.

As Reuters first reported in June, the US government is assessing its link with Beijing.

A senior US State Department official told the BBC he understood “Deepseek is happy to provide and is likely to continue supporting China's military and intelligence reporting efforts.”

Deepseek did not respond to a request for comment from the BBC, but its privacy policy states that its servers are in the People's Republic of China.

“When you access our services, your personal data may be processed and stored on servers in the Republic of China,” the policy states. “This could be a direct provision of personal data to us or a transfer made by us or a third party.”

Earlier this week, Openai reignited its talk about Deepseek after releasing a pair of AI models.

These were the first free open versions (meaning they can be downloaded and modified) released by the American AI giant five years later, long before being led into the age of consumer AI.

“Deepseek allows you to draw a straight line to what Openai announced this week,” said D-Matrix's Sheth.

“Deepseek proved that smaller, more efficient models can still deliver impressive performance. That's what changed the way the industry thinks,” Sheth told the BBC. “What we're looking at now is the next wave of that mindset: the shift to the right-sized model is faster, cheaper and ready to roll out on scale.”

But for others, for the major AI American players, the old approach appears to be alive and well.

Just a few days after releasing the free model, Openai unveiled the GPT-5. At Landing, the company said it has significantly increased its computing capabilities and AI infrastructure.

Numerous announcements about new data center clusters needed for AI come as American tech companies compete for the talent of top AI.

Meta CEO Mark Zuckerberg has ploughed billions of dollars to meet AI ambitions and tried to lure staff from his rivals with a $100 million wage package.

The fate of the tech giant seemed more connected than ever to its commitment to AI spending, as evidenced by the series of blowouts revealed by this past tech revenue season.

Meanwhile, Nvidia's stock, which plunged into the shortly after Deepseek's arrival, touched on a new high that has become the world's most valuable company in history.

“The first story proved a bit of a red herring,” said Khan of Mill Pond Research.

We are back to a future where AI is superficially dependent on more data centers, more chips, and more power.

In other words, the current state of Deepseek has not continued.

And what about Deepseek itself?

“Deepshek is currently facing challenges to maintain momentum,” said Marina Chang, an associate professor at Sydney Institute of Technology.

This is not only due in part to an operational setback, but also to fierce competition between the US and Chinese companies, she said.

Zhang says the company's next product, the Deepseek-R2, is reportedly delayed. One reason? A shortage of high-end chips.