DeepSeek, the AI you’ve never heard of that’s winning the race

by | Jan 27, 2025 | E-commerce News

DeepSeek, a relatively unknown AI research lab from China, released an open source model that's quickly risen in popularity to become the #1 free download on Apple's App Store, leaving ChatGPT behind in second place.

According to a paper authored by the company, DeepSeek-R1 beats the industry's leading models like OpenAI o1 on several math and reasoning benchmarks. It's also significantly cheaper to use because it was significantly cheaper to develop.

Marina Zhang, an associate professor at the University of Technology Sydney, explains: 

“Unlike many Chinese AI firms that rely heavily on access to advanced hardware, DeepSeek has focused on maximizing software-driven resource optimization. DeepSeek has embraced open source methods, pooling collective expertise and fostering collaborative innovation. This approach not only mitigates resource constraints but also accelerates the development of cutting-edge technologies, setting DeepSeek apart from more insular competitors.”

CNBC published a great video report a few days ago that dives deeper into DeepSeek's history, including how it's been able to outperform America’s best AI models despite only taking two months and less than $6M to build, all without access to the top of the line processors that American AI companies are hoarding and the government is withholding from Chinese competitors. 

Liang Wenfeng, a former hedge fund founder who pivoted into the AI research that later became DeepSeek, told the Chinese tech publication 36Kr that the decision to enter the AI space was driven by scientific curiosity rather than a desire to turn a profit. He explained:

“I wouldn't be able to find a commercial reason even if you ask me to. Because it’s not worth it commercially. Basic science research has a very low return-on-investment ratio. When OpenAI’s early investors gave it money, they sure weren’t thinking about how much return they would get. Rather, it was that they really wanted to do this thing.”

When Liang put together DeepSeek's research team, he was not looking for experienced engineers to build a consumer-facing product, but instead focused on finding PhD students from China's top universities who were eager to prove themselves, but lacked industry experience.

In October 2022, when the US government started putting together export controls that restricted Chinese AI companies from accessing chips like Nvidia's H100, DeepSeek had to come up with more efficient methods to train its models. Born out of necessity, the firm optimized their model architecture using a mix-and-match approach of various engineering models and new ideas the team came up with, which they eventually combined to produce a cutting edge-model.

DeepSeek’s willingness to share these innovations with the public has earned it considerable goodwill within the global AI research community, which put Meta into panic mode

An anonymous Meta employee shared:

“Engineers are moving frantically to dissect DeepSeek and copy anything and everything we can from it. I'm not even exaggerating. Management is worried about justifying the massive cost of GenAI org. How would they face the leadership when every single ‘leader’ of GenAI org is making more than what it cost to train DeepSeek V3 entirely, and we have dozens of such ‘leaders’… DeepSeek R1 made things even scarier. I can’t reveal confidential info, but it’ll be public soon.”

Yann LeCun, Meta's chief AI scientist, looks at things differently. He said in a post on Threads, it's not that China's AI is “surpassing the US,” but rather than “open source models are surpassing proprietary ones.”

Days after DeepSeek's announcement, Meta CEO Mark Zuckerberg said Meta planned to spend more than $60B in 2025 as it doubles down on AI. That would pay for a lot of DeepSeek models!

Never miss important e-commerce news

Our weekly newsletter is read each week by 15,000+ e-commerce professionals.

Loading...