What Tech Leaders Think About DeepSeek’s Rise in The AI Game
The AI industry has centred on US technology firms since ChatGPT's release in 2022, with companies including OpenAI, Google and Microsoft leading development.
Yet recently, China’s AI start up company, DeepSeek, made waves in the AI market, leading Nvidia's stock price to decline by 17% on 27th January.
Such a significant rise from a start up company has sent ripples and questions throughout the global tech industry, particularly in the US considering the political tensions between the two AI leaders.
Now, Silicon Valley is questioning if efficiency matters more than raw power and whether cost-efficient AI development is the future.
DeepSeek’s development raises environmental questions
Sustainability is a constant challenge in the tech industry, yet DeepSeek says it has reduced the computational cost of training and running models of its model.
This addresses environmental concerns surrounding AI development, especially considering that data centres running AI models consume substantial electricity and water resources for server cooling – such as ChatGPT, that generates 260 tonnes of carbon dioxide monthly, equivalent to 260 London-New York flights, according to the BBC.
Additionally, part of DeepSeek’s sustainability and efficiency gains breakthrough, is that it centers on test-time scaling, a technique that dynamically manages computing resources during operation, rather than relying on fixed training parameters.
At CES 2025, Nvidia CEO Jensen Huang highlighted this approach as one of three key scaling methods shaping AI development, alongside pre-training and post-training scaling.
A Nvidia spokesperson told Technology Magazine: “We now have three scaling laws: pre-training and post-training, which continue and new test-time scaling.
DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling.
“DeepSeek’s work illustrates how new models can be created using that technique, leveraging widely-available models and compute that is fully export control compliant.
Inference requires significant numbers of Nvidia GPUs and high-performance networking.”
Furthermore, alongside its sustainability achievements, DeepSeek says it has been able to train its model at a much lower cost – claiming R1 cost US$6m to train, a fraction of the “over $100m” referred to by OpenAI CEO Sam Altman when discussing GPT-4.
Sam posted a congratulations to DeepSeek on X, saying it had developed “an impressive model, particularly around what they’re able to deliver for the price.”
DeepSeek’s technical innovation in relation to cost reduction
Adding to the AI competition between China and the US, at the core of DeepSeek’s strategy is a clever workaround to US export restrictions on China.
According to the BBC, the company’s founder, Liang Wenfeng, had accumulated an estimated 50,000 Nvidia A100 chips before they were banned from export to China in September 2022.
Rather than relying solely on these high-end processors, experts suggested that DeepSeek has paired them with less expensive chips to create its AI model.
However, at the London TechEx event on 5th February, Indul Hassan, Head of Engineering at BT Group, said in a panel discussion regarding AI costs for the future - and particularly when asked how DeepSeek’s become so successful - that success can only be measured over time and DeepSeek’s rise is new.
He said: “Is it an academic breakthrough? Yes [...] does that translate to success? [...] that remains to be seen.”
Furthermore, despite DeepSeek’s cost efficiency and sustainability achievements, not all of the company's cost-cutting techniques are new – as some have been used in other LLMs.
For example, in 2023, Mistral AI released its Mixtral 8x7B model which was on par with the advanced models of the time - yet Mixtral’s and DeepSeek’s models both leverage the "mixture of experts" technique - where smaller specialised models handle specific tasks within the larger system, improving efficiency, according to the BBC.
DeepSeek has also admitted to its unsuccessful attempts at improving LLM reasoning through other technical approaches.
Concerns about DeepSeek’s censorship and data storage
"Relatively speaking" the Chinese government has been "hands off" with the app, says Kayla Blomquist, researcher at the Oxford Internet Institute and director of the Oxford China Policy Lab.
"I would say there's a shift as we've seen an announcement in huge investment from the central government just in the last week - so that is probably going to signal a change moving forward”, she told the BBC.
The BBC reports that according to DeepSeek's own privacy policy, it collects large amounts of personal information from users.
The company says it uses this information to improve DeepSeek by enhancing its "safety, security and stability", but will then share this information to others, like service providers, advertising partners and its corporate group, which will be kept "for as long as necessary".
The app also maintains content restrictions aligned with Chinese regulations, declining to address certain topics and this self-censorship distinguishes it from US-developed alternatives.
Global competition intensifies in AI development
DeepSeek’s rise underscores the potential for sustainability, cost reduction, algorithmic innovation to compete with raw computational power in AI development and the possibility for startups to rise to the top.
Discussing DeepSeek’s impact on the tech industry in the same panel as Indul, Franny Hsiao, AI Architect at Salesforce, said regarding questions about how other AI companies can be as ‘successful’ as DeepSeek, that she hopes to see “more fierce competition globally”.
“One of the biggest opportunities in the AI sector is the open arms race of all these AI models and all the technologies that will help generate more AI powered solutions,” she said.
Explore the latest edition of Technology Magazine and be part of the conversation at our global conference series, Tech & AI LIVE.
Discover all our upcoming events and secure your tickets today.
Technology Magazine is a BizClik brand
