US-China AI competition intensifies: cost advantage, talent moves, benchmark wins
Key Questions
Which Chinese model beat Nvidia on a physical AI benchmark?
A Chinese startup outperformed Nvidia on its own physical AI benchmark, demonstrating competitive hardware capabilities. This intensifies the US-China AI competition in specialized domains.
How does DeepSeek V4 compare in cost and performance to GPT models?
DeepSeek V4 is reportedly 107x cheaper than GPT-5.5 while achieving strong results, including higher precision in vulnerability scanning at ~$1 vs $22. Many US startups are adopting Chinese models due to these cost advantages.
What talent move occurred between OpenAI and Tencent?
Tencent hired former OpenAI researcher Yao Shunyu as Chief AI Scientist. This reflects ongoing shifts in AI talent between US and Chinese companies.
How is SpaceX involved in AI compute?
SpaceX is entering the AI compute market, including deals such as providing capacity to Anthropic and Google. These moves are reshaping investment and infrastructure dynamics in the sector.
What trend is emerging among US AI startups regarding Chinese models?
Approximately 80% of US startups are now using Chinese models, driven by significant cost and performance edges. This challenges traditional US dominance in the AI landscape.
A Chinese startup beat Nvidia on its own physical AI benchmark. DeepSeek V4 reportedly 107x cheaper than GPT-5.5, with 80% of US startups using Chinese models. Tencent hired OpenAI researcher Yao Shunyu as Chief AI Scientist. SpaceX enters AI compute for Anthropic. New: DeepSeek V4 Pro beats GPT-5.5 Pro on precision in real-world vulnerability scanning, costing ~$1 vs $22, reinforcing cost advantage. These developments challenge US dominance and reshape investment landscape.