US AI firms team up to avoid another Deepseek moment
US artificial intelligence (AI) heavyweights are together stepping up to protect their commercial edge, as Chinese rivals increasingly chip away at their pricing power.
Google, Anthropic and OpenAI have begun sharing information on rivals trying to replicate their models through distillation, Bloomberg reported on Tuesday.
The united effort has been coordinated through an industry non-profit, the Frontier Model Forum, founded back in 2023 alongside Microsoft.
US firms have invested dizzying amounts in building these models by betting that their customers will pay for access.
But, distillation has allowed competitors to train similar systems at a much lower price, simply by learning from the outputs of those tech giants’ models.
This reduces the need for the same level of data investment that we have seen from the likes of OpenAI.
The ChatGPT maker has warned the US administration that Chinese developers like Deepseek are using increasingly intelligent models to extract value from homegrown systems/
It also pointed to the practice risks of eroding revenues across the AI sector, all whilst multiplying the number of competing products.
Pricing pressure builds
Chinese models, many of which are cheaper to deploy, are gaining traction with developers and businesses worldwide keen to manage costs.
Firms like Pinterest or Airbnb have already been seen experimenting with Chinese-built models, citing lower operating costs as well as competitive performance.
“Open source techniques that we use to train our own in-house models are 30 per cent more accurate than the leading off-the-shelf model”, said Pinterest chief tech officer Matt Madrigal.
Elsewhere, Chinese models often rank among the most downloaded on developer platforms, pointing to their appeal for startups and SMEs.
A Stanford University report published in late 2025 found Chinese AI models “seem to have caught up or even pulled ahead of their global counterparts.”
This amps up pressure on the business model currently underpinning American AI companies.
While firms such as Anthropic have so far focused on monetising access to their proprietary systems, their cheaper, Chinese counterparts are offering alternatives that can be run more cheaply.
Investors have pointed to the emergence of models like Deepseek’s as a key inflection point, proving that high-level performance does not necessarily require the same level of spending that has defined the US approach.
And at the same time, China’s broader AI ecosystem is expanding. The country is producing a larger share of global AI talent and retaining more of its researchers, while its companies are scaling infrastructure and distribution across emerging markets.