The technological advances that Chinese artificial intelligence lab DeepSeek have displayed show the game is on when it comes to U.S.-Sino competition on AI, top tech executives told CNBC.
In a series of interviews at France’s Artificial Intelligence Action Summit, leaders of several major tech companies told CNBC that the emergence of DeepSeek demonstrates that China can’t be counted out as a serious player when it comes to AI innovation.
Last month, DeepSeek shocked global markets with a technical paper saying that one of its new AI models was created with a total training cost of less than $6 million — far less than the billions upon billions of dollars being spent by Big Tech players and Western AI labs such as OpenAI and Anthropic.
Chris Lehane, chief global affairs officer at OpenAI, told CNBC that DeepSeek’s advanced, low-cost model confirms there is a “very real competition between U.S.-led, small D democratic AI and CCP [Chinese Communist Party] China-led autocratic, authoritarian AI.”
Many critics of DeepSeek have pointed to apparent censorship by the model when it comes to sensitive topics. For example, when asked about the 1989 Tiananmen Square massacre, DeepSeek’s AI assistant app responds with: “Sorry, that’s beyond my current scope. Let’s talk about something else.”
“There’s two countries in the world that can build this at scale,” Lehane told CNBC’s Arjun Kharpal on the sidelines of the Paris AI summit Monday. “Imagine if there were only two countries in the world that could build electricity at scale. That’s sort of how you have to think about it.”
“For us, what DeepSeek really reinforces and reaffirms is that there is this very real competition with very real stakes,” Lehane added.
Still, tech bosses largely agreed that even though DeepSeek’s breakthrough shows China being further along in the global AI race than previously thought, the threat it poses to OpenAI remains limited for now.
‘The game is on’
DeepSeek says that its new R1 model, an open-source reasoning model, was able to rival the performance of OpenAI’s own similar o1 model — only using a cheaper, less energy-intensive process.
That led experts to question the prevailing wisdom in the West of the last several years, which is that China is behind the U.S. on AI development because of export restrictions that make it harder for firms in the country to get their hands on more advanced Nvidia graphics processing units, or GPUs.
GPUs are necessary for training and running AI applications because they excel at parallel processing, meaning they can perform multiple calculations simultaneously.
Reid Hoffman, a co-founder of LinkedIn and partner at the venture capital firm Greylock Partners, told CNBC Monday that DeepSeek’s new model is “a big deal in showing that the game is on.”
“The competition is afoot with China,” Hoffman said, adding that DeepSeek’s R1 is “a credible, actionable model.”
Abishur Prakash, founder of strategic advisory firm The Geopolitical Business, told CNBC that DeepSeek shows the West’s understanding of China remains limited.
“America’s assumed place as the technological captain of the world is no longer the acceptable belief,” Prakash told CNBC in a phone interview.
“That is the new status quo now, that the space between the U.S. and China has narrowed almost overnight — but it hasn’t narrowed overnight, it’s been years of progress,” Prakash said.
“If there’s one takeaway for the West, it’s that their understanding of China is incredibly limited — and we don’t know what’s coming next,” he added.
No meaningful threat to U.S. AI — yet
Still, leading AI execs aren’t convinced that DeepSeek poses any sort of meaningful risks to the businesses of AI labs like OpenAI and Anthropic just yet.
While experts on the whole agree DeepSeek’s AI advances have been impressive, doubts have been raised about the startup’s claims about cost.
A report from semiconductor research firm SemiAnalysis last month estimated that DeepSeek’s hardware expenditure is “well higher” than $500 million over the company’s history. DeepSeek was not immediately available for comment when contacted by CNBC.
The report found that DeepSeek’s research and development costs and expenses related to ownership are significant and that generating “synthetic data” for the model to train on would require “considerable amount of compute.”
Some technologists believe that DeepSeek may have been able to achieve such a high level of performance by training its models on larger U.S. AI systems.
This technique, known as “distillation,” involves having more powerful AI models evaluate the quality of answers being generated by a newer model.
It’s a claim that OpenAI itself has alluded to, telling CNBC in a statement last month that it’s reviewing reports that DeepSeek may have “inappropriately” used output data from its models to develop its AI model, a method referred to as “distillation.”
“Most of the market fear around [DeepSeek] is in fact misplaced,” Hoffman told CNBC. “It still requires large models — it was distilled from large models.”
“I think the short answer everyone should take is: game on — but large models still really matter,” he added.
Victor Riparbelli, CEO of AI video platform Synthesia, told CNBC that although DeepSeek challenged the “paradigm that brute force scaling is the only way to kind of build better and better models,” the idea that companies are going to suddenly shift significant amounts of their AI workloads is misguided.
“I still think that when you look at users of these technologies, all the workflows, I think when we look back in three months’ time, I think 0.01% of those is going to be moved to Deepseek from OpenAI and Anthropic,” Riparbelli said.
Meredith Whitaker, president of the Signal Foundation, said DeepSeek’s development doesn’t move the needle much for the industry as market momentum is still broadly in favor of larger AI models. The Signal Foundation is a nonprofit that supports the encrypted messaging app Signal.
“This is not something that’s going to disrupt the concentration of power or the geopolitical balance at this stage,” Whitaker told CNBC. “I think we have to keep our eye on the ball there and recognize that it’s really this ‘bigger is better’ paradigm that is not reduced through efficiency gains historically, that is driving this concentration.”
