In May 2024, the Chinese AI startup DeepSeek shocked China’s AI industry by triggering a devastating price war in “inference,” the answers provided by large AI models. That price war is about to erupt across the globe.
At the time, few people in the West paid attention to the obscure Chinese company’s release of its DeepSeek V2 model. But China’s AI tech giants were stunned by DeepSeek’s offer of inference for just 1 RMB per million input tokens (approximately $0.14) and double that for output tokens. That price—one-seventh of Meta’s Llama3 70B and one-seventieth of OpenAI’s GPT-4 Turbo—forced major Chinese players like ByteDance, Tencent, Baidu, and Alibaba to quickly slash their prices in response.
Now Western tech giants face a similar challenge.
DeepSeek, now with models that rival the best of the West, has set the stage for a global war in AI inference pricing that is only now becoming clear to the world.
Commoditizing Inference
Everybody knew AI inference would become commoditized, but no one anticipated it would happen so fast nor that the catalyst would come from China. But the battle is joined and it will reshape the competitive landscape of AI and disrupt the profitability of many of the largest tech giants.
It will also affect us all, by making AI more available to more people, accelerating its adoption by businesses and consumers.
DeepSeek’s 2024 price war demonstrated how a newcomer could force entrenched industry leaders to adapt. Companies such as ByteDance and Alibaba quickly lowered their prices, initiating a cascade of cost-cutting across the Chinese AI market. Meanwhile, the sudden affordability of high-performance AI made these services accessible to smaller enterprises, sparking an explosion of adoption across industries. At the same time, the price cuts decimated AI profit margins, leaving companies scrambling to innovate or consolidate to stay alive.
While the immediate effects were largely confined to China, the implications are far-reaching. With DeepSeek’s cost-efficient, open-source model as a benchmark, global tech giants like OpenAI, Microsoft, and Google are facing pressure to reduce prices in Western markets to retain their dominance.
The Next Battlefront: Global AI Inference
AI inference—powering real-time applications like chatbots, content generation, and autonomous systems—represents a significant cost center for AI providers. The U.S. market, largely dominated by OpenAI, Google, and Anthropic, has so far been able to maintain premium pricing. But the entry of a low-cost competitor like DeepSeek into international markets will trigger a domino effect akin to what unfolded in China.
DeepSeek’s success highlights the growing power of open-source models. If open-source competitors like Meta (with its Llama series) and Stability AI follow suit with similarly aggressive pricing, premium providers such as OpenAI or Anthropic will struggle to justify their costs.
These companies are heavily reliant on AI inference revenue, and will likely see their margins evaporate as competitors undercut them. This will lead to an industry-wide reevaluation of business models.
As prices drop, AI tools will become accessible to smaller businesses and startups worldwide but it could force providers to compete on efficiency rather than innovation.
Lessons from the DeepSeek Playbook
DeepSeek achieved unprecedented cost-efficiency by reimagining its architecture. The company’s MLA (Multi-head Latent Attention) mechanism drastically reduced memory usage, while its DeepSeekMoESparse framework minimized computational overhead. Such architectural innovations may become a competitive necessity.
By embracing open-source strategies, DeepSeek fostered rapid adoption while building goodwill within developer communities. This move challenged the traditional closed-loop models of companies like OpenAI, forcing competitors to rethink their strategies.
Unlike many companies subsidizing low prices at a loss, DeepSeek remained profitable throughout the price war. This underscores the importance of balancing affordability with sustainable business practices.
The Risks of a Global Price War
While price reductions may democratize AI access, a global price war comes with significant risks:
Sustained pressure to lower prices could reduce resources available for research and development, potentially slowing the pace of innovation.
Smaller players without the scale or resources to compete on price may be forced out, resulting in greater market concentration.
As AI pricing becomes a global issue, it may deepen tensions between tech hubs like Silicon Valley and emerging markets like China. Governments may intervene to protect domestic industries or limit the dominance of foreign competitors.
A Precedent for the Future
DeepSeek’s price war in China serves as a warning for the global AI industry that the balance between innovation, accessibility, and profitability is more precarious than ever. As the battle for dominance in AI inference intensifies, companies will need to adapt quickly or risk being left behind.
For now, the question is not whether a global AI price war will happen—it’s when. The lessons from DeepSeek suggest that only those who embrace cost innovation and rethink traditional business models will survive. As the storm brews, one thing is clear: the age of AI affordability is both a promise and a threat, and the consequences will shape the industry for years to come.
