Chinese startup DeepSeek has disrupted the AI industry with its R1 reasoning model, which was trained in just two months for $5.6 million using less powerful chips. The model is said to be 45 times more efficient than U.S. rivals and quickly became the most downloaded free app in the U.S., even surpassing ChatGPT. Its launch rattled markets, wiping nearly $1 trillion in value from major U.S. tech firms like Alphabet, Microsoft, Nvidia, and Oracle as investors worried about the threat to American AI leadership.
DeepSeek’s model stands out for its use of chain-of-thought reasoning, a method that breaks tasks into logical steps, boosting performance on math, logic, and problem-solving. This efficiency in training is notable, but the approach may actually require more energy during inference, since the model produces longer, more detailed answers. The paradox is that while DeepSeek reduces training costs, the day-to-day use of its system could be more energy-intensive than competitors.
AI tech, as we've discussed, is extremely energy intensive and there is growing concern about the ramp up in required power we are seeing as AI and data storage requirements skyrocket.
We wrote an article for Oil & Energy Magazine detailing the impacts on the markets and the potential energy industry impacts from AI technology. You can read that article here: DeepSeek AI is our Wakeup Call