An article by DeepSeek Founder Liang Wenfeng revealed the Chinese AI startup's R1 AI model cost just $294,000 (₹2.6 crore) to train, much lower than its US rivals. The reasoning-focused model used 512 Nvidia H800 chips, he noted. OpenAI CEO Sam Altman had said in 2023 that his startup's foundational models' training cost "much more" than $100 million (₹880 crore).