AI’s Efficiency Wars Have Begun

The DeepSeek shock may reshape a global race.

By , a researcher at University College London, and , an associate professor of digital government and a co-deputy director of University College London’s Institute for Innovation and Public Purpose.


In this photo illustration, the DeepSeek search page is displayed on a mobile phone in front of a laptop screen displaying the DeepSeek homepage in London on Jan. 29.
In this photo illustration, the DeepSeek search page is displayed on a mobile phone in front of a laptop screen displaying the DeepSeek homepage in London on Jan. 29.

In this photo illustration, the DeepSeek search page is displayed on a mobile phone in front of a laptop screen displaying the DeepSeek homepage in London on Jan. 29. Leon Neal/Getty Images



The rapid release of DeepSeek-R1—one of the newest models by Chinese AI firm DeepSeek—sent the world into a frenzy and the Nasdaq into a dramatic plunge. The reason is simple— DeepSeek-R1, a type of artificial intelligence reasoning model that takes time to “think” before it answers questions, is up to 50 times to run than many U.S. AI models. Distilled versions of it can also on the computing power of a , while other models require several of Nvidia’s most expensive chips. But what has really turned heads is DeepSeek’s that it only spent about $6 million to finally train its model—much less than OpenAI’s o1. While this figure is misleading and does not include the substantial costs of , refinement, and more, even partial cost reductions and efficiency gains may have significant geopolitical implications.

So, why is DeepSeek-R1 so much cheaper to train, run, and use? The answer lies in several computational efficiency improvements made to the R1 model. First, R1 used a different machine learning architecture called “mixture of experts,” which a larger AI model into smaller subnetworks, or “experts.” This approach that when given a prompt, RI only needs to activate the experts relevant to a given task, greatly decreasing its computational costs.

The rapid release of DeepSeek-R1—one of the newest models by Chinese AI firm DeepSeek—sent the world into a frenzy and the Nasdaq into a dramatic plunge. The reason is simple— DeepSeek-R1, a type of artificial intelligence reasoning model that takes time to “think” before it answers questions, is up to 50 times to run than many U.S. AI models. Distilled versions of it can also on the computing power of a , while other models require several of Nvidia’s most expensive chips. But what has really turned heads is DeepSeek’s that it only spent about $6 million to finally train its model—much less than OpenAI’s o1. While this figure is misleading and does not include the substantial costs of , refinement, and more, even partial cost reductions and efficiency gains may have significant geopolitical implications.

So, why is DeepSeek-R1 so much cheaper to train, run, and use? The answer lies in several computational efficiency improvements made to the R1 model. First, R1 used a different machine learning architecture called “mixture of experts,” which a larger AI model into smaller subnetworks, or “experts.” This approach that when given a prompt, RI only needs to activate the experts relevant to a given task, greatly decreasing its computational costs.

Second, DeepSeek improved how efficiently R1’s algorithms its computational resources to perform various tasks. For example, R1 uses an algorithm that DeepSeek previously called Group Relative Policy Optimization, which is less computationally intensive than other commonly used algorithms. Beyond these areas, DeepSeek other computational optimizations as well. For example, it used fewer decimals to represent some numbers in the calculations that occur during model training—a technique called mixed precision training—and improved the curation of data for the model, among many other improvements. Together, these computational efficiency improvements a model that was more cost-efficient than many other existing ones.

These efficiency gains are significant and offer, among many others, four potential—though not guaranteed—implications for the global AI market. First, these efficiency gains could potentially drive new entrants into the AI race, including from countries that previously lacked major AI models. Until now, the prevailing view of frontier AI model development was that the primary way to significantly increase an AI model’s performance was through ever larger amounts of compute—raw processing power, essentially. Smaller players would struggle to access this much compute, keeping many of them out of the market.

However, R1, even if its training costs are not truly $6 million, has convinced many that training reasoning models—the top-performing tier of AI models—can cost much less and use many fewer chips than presumed otherwise. The result, combined with the fact that DeepSeek mainly domestic Chinese engineering graduates on staff, is likely to convince other countries, firms, and innovators that they may also possess the necessary capital and resources to train new models.

Indeed, such perceptions are already taking root. In the wake of R1, Perplexity CEO Aravind Srinivas for India to develop its own foundation model based on DeepSeek’s example. Governments such as France, for example, have already been supporting homegrown firms, such as Mistral AI, to enhance their AI competitiveness, with France’s state investment bank in one of Mistral’s previous fundraising rounds. With the perception of a lower barrier to entry created by DeepSeek, states’ interest in supporting new, homegrown AI firms may only grow.

These lower barriers to entry may also add additional complexity to the global AI race. In recent months, many that AI would become a footrace between Washington and Beijing. But now, while the United States and China will likely remain the primary developers of the largest models, the AI race may gain a more complex international dimension. Both U.S. and Chinese firms have heavily courted international partnerships with AI developers abroad, as seen with Microsoft’s with Arabic-language AI model developer G42 or Huawei’s in the China-ASEAN AI Innovation Center. With more entrants, a race to secure these partnerships might now become more complex than ever.

Furthermore, efficiency could soon join compute as another central focus of state industrial policies in the global AI race. Prior to R1, governments around the world were racing to build out the compute capacity to allow them to run and use generative AI models more freely, believing that more compute alone was the primary way to significantly scale AI models’ performance.

India’s Mukesh Ambani, for example, is to build a massive 3-gigawatt data center in Gujarat, India. However, R1’s launch has spooked some investors into believing that much less compute and power will be needed for AI, a large selloff in AI-related stocks across the United States, with compute producers such as Nvidia seeing $600 billion declines in their stock value.

Despite these recent selloffs, compute will likely continue to be for two reasons. First, there is the classic economic case of the Jevons paradox—that when technology makes a resource more efficient to use, the cost per use of that resource might decline, but those efficiency gains actually make more people use the resource overall and drive up demand.

There has been some evidence to the Jevons paradox in energy markets, whereby total compute demand might go up in any scenario. The drop in Nvidia’s stock price was significant, but the company’s enduring $2.9 trillion valuation that the market still sees compute as a vital part of future AI development. Second, R1’s gains also do not disprove the fact that more compute to AI models that perform better; it simply validates that another mechanism, via efficiency gains, can drive better performance as well.

These reasons suggest that compute demand could actually increase, not decrease—but at the same time, improving efficiency will likely be a priority for both firms and governments. In particular, firms in the United States—which have been spooked by DeepSeek’s launch of R1—will seek to adopt its computational efficiency improvements alongside their massive compute buildouts, while Chinese firms may try to on this existing advantage as they increase domestic compute production to bypass U.S. export controls.

Governments in both countries may try to support firms in these efficiency gains, especially since documents such as the Biden administration’s 2024 made having the world’s most performant AI systems a national priority.

R1’s lower price, especially when compared with Western models, has the potential to greatly drive the adoption of models like it worldwide, in parts of the global south. This kind of rapid AI adoption might accelerate AI’s benefits to economic growth in these countries, potentially increasing their long-term geopolitical heft and posing new challenges for U.S. policymakers concerned about the global use of Chinese AI tools.

However, as DeepSeek sees this vast global market, many of America’s powerhouse AI developers might also double down on more computationally efficient and lower-price models to make competitive offerings in the AI markets in these countries, suggesting an AI race across the global south—at the level of adoption, in addition to partnerships—may occur.

Very little can be guaranteed in a competition as fast-moving as this one. However, DeepSeek’s efficiency gains have provided a challenge to existing assumptions of the global AI race and may change its competitive dynamics in a way previously unpredicted. Across much of the world, it is that DeepSeek’s cheaper pricing and more efficient computations might give it a temporary advantage, which could prove significant in the context of long-term adoption.

However, it may not also be long before both U.S. and homegrown or regional alternatives enter the fray as well, triggering further competition over who will use which platforms. With more models and prices than ever before, only one thing is certain—the global AI race is far from over and is far twistier than anyone thought.



Sarosh Nagar is a researcher at University College London. His work on AI has previously been published by the United Nations and in the Hill, Newsweek, and the Diplomat. X: 

David Eaves is an associate professor of digital government and a co-deputy director of University College London’s Institute for Innovation and Public Purpose. Bluesky:  X: 

Read More On
| |

Join the Conversation

Commenting on this and other recent articles is just one benefit of a Foreign Policy subscription.

Already a subscriber?
.

Join the Conversation

Join the conversation on this and other recent Foreign Policy articles when you subscribe now.

Not your account?

Join the Conversation

Please follow our , stay on topic, and be civil, courteous, and respectful of others’ beliefs.

You are commenting as .

Change your username:




Confirm your username to get started.

The default username below has been generated using the first name and last initial on your FP subscriber account. Usernames may be updated at any time and must not contain inappropriate or offensive language.





More from Foreign Policy

  • Russian President Vladimir Putin looks on during a press conference after meeting with French President in Moscow, on February 7, 2022.

    A series of setbacks for Russia is only gaining momentum.

  • The container ship Gunde Maersk sits docked at the Port of Oakland on June 24, 2024 in Oakland, California.

    The White House is threatening a close ally with a trade war or worse—but Copenhagen has leverage that could inflict instant pain on the U.S. economy.

  • Donald Trump speaks during an event commemorating the 400th Anniversary of the First Representative Legislative Assembly in Jamestown, Virginia on July 30, 2019.

    His return to power has been impressive—but the hard work is about to begin.

  • US Secretary of State Marco Rubio greets employees at the State Department in Washington, DC, on January 21, 2025.

    President Trump has an opportunity to unleash underutilized talent in tackling dangers at home and abroad.