Markets were panicked on Monday after Chinese AI firm DeepSeek debuted its new low-cost chatbot. Nvidia suffered the worst one-day stock wipeout in US history, losing $600 billion, amid claims by the Chinese tech firm it could beat US industry leaders for a fraction of the cost. Until this week there was no doubt that the US held a commanding lead in the AI arms race. But the emergence of DeepSeek R1 has sent shockwaves through Wall Street and is being “described as AI’s Sputnik moment.”

It had previously been thought that the cost of training AI models was enormous. The cost of training Chat GPT 4 is over $100 million, according to the Open AI CEO Sam Altman. However, DeepSeek, a small company with fewer than 200 employees, claims its competitor product has been built for just $5 million. If true, that could mean DeepSeek has developed a competitive edge over its American rivals – that would mean assumptions around the value of US big tech firms could need to be adjusted.

However, there are questions over whether DeepSeek is being entirely transparent about the costs of its AI chatbot. On Monday, Gene Munster of Deepwater said “I still think the truth is below the surface when it comes to actually what’s going on,” in comments to the BBC. Others see the advent of faster and cheaper ways to develop AI as a good thing – potentially accelerating the rollout of AI and its many productivity boosting applications.

DeepSeek has changed the future of AI

From nowhere, DeepSeek has in its debut of R1 transformed the future of AI. That’s according to finance chief Nigel Green, who sees an opportunity for investors amid the market turmoil. The deVere CEO said that markets are taking the wrong lessons from DeepSeek’s breakthrough – rather than shrinking demand, efficiency-driven AI models will mean faster and more widespread adoption of the technology. Commenting on Monday, he said:

“DeepSeek’s breakthrough signals a shift toward efficiency in AI, which will redefine both energy and AI markets. The opportunities for investors willing to act now are enormous.

“This challenges the assumption that AI’s growth is tied to ever-increasing energy consumption. While the market is reacting to short-term uncertainty, efficiency-driven AI models will expand adoption into new markets and industries.  “This means more widespread use, deeper integration, and ultimately, sustained demand for energy solutions. DeepSeek’s breakthrough will accelerate AI adoption and create massive opportunities for investors who focus on the sectors driving this transformation.”

The veteran finance boss said investors should ‘ignore the noise’ and focus on the opportunities in AI which become possible in a new era of efficiency:

“Disruption creates opportunities for those who can see beyond the noise.

“Energy and AI markets are entering a new phase, and investors who align with these changes now will be positioned to benefit from the next wave of growth.”

The shock and awe of DeepSeek landing on the App Store on Monday has left many big Silicon Valley players reeling, particularly as it came off the back of a $500 billion investment by Stargate last week. But when the dust settles it may well be that while DeepSeek has established itself as a contender, it doesn’t knock out any of its opponents.

Atif Malik from Citi Group said US dominance in high-tech chips would hold, as major investments in the sector expand its lead over rivals like China. Those chips, Malik told the FT, will remain key to training AI models. While Bernstein’s Stacy A Ragson told the outlet she and her team simply did not believe DeepSeek was being honest about its $5 million budget. She said:

“Did DeepSeek really build OpenAI for $5 million? Of course not…There are actually two model families in the discussion. The first family is DeepSeek-V3, a Mixture-of-Experts (MoE) large language model which, through a number of optimizations and clever techniques can provide similar or better performance vs other large foundational models but requires a small fraction of the compute resources to train.

“DeepSeek actually used a cluster of 2048 NVIDIA H800 GPUs training for ~2 months (a total of ~2.7M GPU hours for pre-training and ~2.8M GPU hours including post-training). The oft-quoted “$5M” number is calculated by assuming a $2/GPU hour rental price for this infrastructure which is fine, but not really what they did, and does not include all the other costs associated with prior research and experiments on architectures, algorithms, or data.”

Is DeepSeek too good to be true?

Some analysts have cast doubt on claims by DeepSeek that its new R1 AI model was trained for a minuscule $5 million. If true that would completely upend assumptions which underpin US tech valuations and potentially threaten to disrupt plans to massively expand energy production in America. Others think while DeepSeek may have pulled off an efficiency coup, the real production cost remains significantly higher than they have so far admitted.

Alexandr Wang, CEO of Scale AI believes DeepSeek has sunk an enormous amount of money into Nvidia H100 chips, which they haven’t declared because such a purchase would violate export controls. Speaking to CNBC he said:

“My understanding is that DeepSeek has about 50,000 H100s…Which they can’t talk about, obviously, because it is against the export controls that the United States has put in place.”

Wang, the world’s youngest self-made billionaire, maintained the AI revolution would still require an enormous increase in energy output, despite what the DeepSeek episode appears to tell us about efficiency:

“The United States is going to need a huge amount of computational capacity, a huge amount of infrastructure…We need to unleash U.S. energy to enable this AI boom.”

Gavin Baker, Managing Partner at Atreides took to social media to express his scepticism over the claims. Posting to X he said DeepSeek R1 was real and in many ways impressive, but added:

“The $6m does not include “costs associated with prior research and ablation experiments on architectures, algorithms and data” per the technical paper…

“DeepSeek obviously has way more than 2048 H800s; one of their earlier papers referenced a cluster of 10k A100s. An equivalently smart team can’t just spin up a 2000 GPU cluster and train r1 from scratch with $6m…”

However rival firms like Nvidia and OpenAI have had nothing but praise for their new rival and have shown no signs of scepticism around DeepSeek’s claims, at least, in public. And even Bernstein’s Stacy Ragson, who does not believe the R1 model was built for $5 million maintains that “DeepSeek’s pricing blows away anything from the competition,”

Is there an opportunity in AI stocks right now?

While some investors have been spooked by DeepSeek, others see opportunity as the AI arms race gets hot. With China now putting a flame under US tech firms, American AI leaders are now rushing to respond. This renewed and intensified competition could bode well for the sector as a whole. The Open AI boss responsible for Chat GPT has already vowed that his company will rise to and beat the challenge. Posting to X, he said:

“We will obviously deliver much better models and also, it’s legitimately invigorating to have a new competitor! We will pull up some releases,”

Meanwhile, Mark Zuckerberg’s Meta has scrambled four ‘war rooms’ of engineers to assess DeepSeek and decide how to respond. The firm’s Llama 4 chatbot is slated for release early this year, which will no doubt have DeepSeek in its sights. Commenting, a spokesperson for the company said:

“We regularly evaluate all competitive models in our development process and have done so since Gen Al was formed,

“Llama has been foundational in establishing the ecosystem for open-source AI models and we couldn’t be more excited to extend this leadership with the upcoming release of Llama 4.”

It came after the Meta CEO announced an enormous $65 billion on AI-related projects in the US this year.

ArKK Investment CEO Cathie Wood is among those who see the emerging battle over AI as a potential boon. Speaking to Bloomberg TV she said:

“Lowering costs is great for the world…They were collapsing anyway, DeepSeek has stepped it up a notch.”

She added Meta and Amazon, in which ArKK holds shares, ‘could benefit from declining costs and may use some of the techniques and algorithms from DeepSeek to improve their platforms.’

This fresh competition could create a furtive environment in the AI sector, with the emergence of new and real threats to US dominance, the incentive to innovate, and fast, will be more intense than ever. But it’s no sure thing they’ll beat off competition from China. Nigel Green, CEO of the deVere Group, told investors:

“This is both a warning and an opportunity—it’s time to rethink traditional tech allocations and seek out new areas of growth….

“Traditional tech giants are no longer the guaranteed winners. The focus must now shift to sectors and regions that are driving the next wave of innovation. This includes not only AI but also the critical infrastructure needed to support and secure it.”

DeepSeek faces challenges in bid to upend US AI dominance

While DeepSeek’s emergence has disrupted the AI landscape in dramatic fashion, the firm, and other potential competitors from China will face challenges not just from their US rivals, but from the American legal framework. Concerns around user privacy and potential regulatory actions in the US could prove significant.

DeepSeek’s data handling practices have come under intense scrutiny. Reports indicate that the platform transmits user data, including chat messages and personal information, to servers in China. The company’s privacy policy permits extensive data collection, encompassing device information and user interactions.

Critics warn that such practices not only risk unauthorised data access but also potential misuse by the Chinese government. Security experts have expressed apprehension about possible content censorship and the broader implications of data privacy. Oxford Professor of AI Michael Woodridge has explicitly warned DeepSeek users that they should refrain from sharing personal information with the chatbot, fearing it could be used by the Chinese state.

Given the heightened focus on data security and geopolitical tensions, DeepSeek could face regulatory challenges in the United States. The U.S. government has previously taken measures against Chinese tech companies over security concerns, including imposing restrictions on firms like Huawei and ZTE.

These actions reflect a broader strategy to limit China’s access to advanced technologies and protect national security interests. Tik Tok could be poised to skirt a US-wide ban if it agrees to a partial sale – and if DeepSeek or companies like it come to be deemed a threat, it could be faced with the same ultimatum.

However, the very emergence of DeepSeek, despite export controls imposed on China by the US, poses questions over whether America can do all that much to quell competition, even if it pulls every lever at its disposal. Some analysts even say DeepSeek could be a Frankenstein’s monster of US export controls, the result of innovation designed to dodge limits on the import of chips and other technologies. DeepSeek’s rapid ascent in the AI sector illustrates both the potential and challenges of global technological competition. Investors should closely monitor these developments, balancing the opportunity of a hotter-than-ever AI arms race with the risk that new efficiencies undermine priced in assumptions about the sector.