President Trump's AI chip export controls hurt DeepSeek's release of its next-gen R2 AI model

DeepSeek was meant to unleash its next-gen AI model known as R2 in May, but the release has been pushed back since US export restrictions have made it much harder for the company to get R2 onto the market.

In a new report from The Information, we're hearing that DeepSeek engineers have been working on refining the new R2 model until DeepSeek CEO Liang Wenfeng gives the green light for release. However, the US government has continued to tighten US export restrictions and it has had a negative impact on the release of DeepSeek's upcoming R2 model.

The report says that the fast adoption of the Chinese company's new R2 model would overwhelm Chinese cloud providers, who are in dire need of NVIDIA's advanced AI chips in order to run the more powerful R2 models. DeepSeek has reportedly been in communication with some of these Chinese cloud companies, providing them with technical specifications to help their plans for hosting and distributing some of these models from their servers.

Some of the Chinese cloud companies are using DeepSeek's current R1, with the majority of them running NVIDIA H20 AI GPUs, but with fresh restrictions handed down by the Trump administration back in April 2025, NVIDIA has been banned from selling its H20 chips to China. This is a major issue, as the H20 was the only AI accelerator the companies could legally export to China at the time.