DeepSeek’s disruption: How a Chinese startup is redefining AI’s efficiency and cost

The news: Chinese startup DeepSeek disrupted the AI landscape with its open-source R1 model by achieving performance comparable to OpenAI’s latest models with older hardware and presumably at just 3% to 5% of the total cost, per VentureBeat.

DeepSeek’s more-affordable AI staggered Big Tech markets Monday and challenged the notion of the US’ AI leadership. 

• Nvidia’s shares plunged as much as 18%, wiping out $589 billion of the chipmaker’s value.
• The DeepSeek revelation also drove selloffs in other US chip, energy, and infrastructure stocks.

Owned by a small Chinese hedge fund, DeepSeek shared its AI performance breakthroughs a week after the Trump administration announced Project Stargate—a $500 billion AI infrastructure play addressing the future US data server and energy requirements needed to dominate the AI space.

A meteoric rise: DeepSeek’s AI assistant has already become the top free app on the Apple App Store, surpassing ChatGPT. However, the app was hit with a large-scale cyberattack on Monday, limiting new user sign-ups. The company also released an image-generation model comparable to DALL-E 3.

Nvidia called DeepSeek an “excellent AI advancement” that complies with US technology export controls, per Bloomberg.

OpenAI's Sam Altman said DeepSeek's R1 was an "impressive model" while promising that his company would "deliver much better models" in the future.

AI’s Sputnik moment: R1 can handle conversations, math, and coding tasks at the same level as competing products. 

• DeepSeek’s training methods are so efficient that they were built on one-tenth the computing power that Meta’s comparable Llama 3.1 used, per Epoch AI.
• It trained the model at a fraction of the cost of competitors—$5.6 million versus hundreds of millions.
• The company said it used just 2,000 Nvidia chips to train R1, far fewer than the 16,000 used by top models, per The New York Times.

These claims are sparking doubts about the compute- and energy-intensive methods of leading AI firms at a time when investors are wary of escalating AI investments.

A big win for open-source AI: DeepSeek’s startup-built open-source model defies the idea that cutting-edge AI needs to be developed by large companies investing heavily in data centers and expensive server hardware. 

As an open-source project, DeepSeek shares its innovation and methodology, a move that could attract more users and contributors to future builds.

Our take: DeepSeek presents a seismic shift in AI strategy that moves the focus from expensive hardware and data center expansion to efficient open-source AI development. 

It remains to be seen whether DeepSeek’s model can scale under heavy user adoption or how AI companies will realign their efforts around affordable AI competition.

This article is part of EMARKETER’s client-only subscription Briefings—daily newsletters authored by industry analysts who are experts in marketing, advertising, media, and tech trends. To help you start 2025 off on the right foot, articles like this one—delivering the latest news and insights—are completely free through January 31, 2025. If you want to learn how to get insights like these delivered to your inbox every day, and get access to our data-driven forecasts, reports, and industry benchmarks, schedule a demo with our sales team.