Google, OpenAI, and Anthropic are ready for DeepSeek’s upcoming key AI release.

Despite the export ban in the US, a Chinese AI startup known as Deepseek has apparently trained its advanced AI model on Nvidia's most powerful Blackwell chips. According to Reuters, it is cited by a senior official from the Trump administration. DeepSeek is considered the next big release and is expected to drop in the upcoming week. Rumours about chip smuggling had been circulating since 2025.

As per the information shared by the official, ‘the Blackwell chips are considered to be in a data centre in Inner Mongolia. Deepseek appears to scrub technical fingerprints of US chip manufacturing before release. However, the information on how Deepseek obtained the chips is not shared. NVIDIA declined to comment, and neither Deepseek nor the US Department of Commerce answered Reuters.

If the timing of the leaks is any pointer, the DeepSeek AI model might be on the verge of one more key splash. OpenAI, Google, and Anthropic have all been complaining about refinement attacks on their models by Chinese companies, and OpenAI recently changed to relativize a renowned coding benchmark. Mutually, the moves imply Deepseek is about to give strong results at rock-bottom cost once again. Look back in January 2025, China's top AI startup sent shockwaves through US tech stocks moving to the AI bubble.

The upcoming launch is broadly considered a turning point for disruptive AI models, potentially reshaping cost and innovation standards globally. Analysts consider that the DeepSeek AI competition will speed up fast advancements, positioning this next generation LLM release as a defining time in the growing global artificial intelligence race.