OpenAI and SoftBank have announced the largest AI infrastructure investment in history — "Stargate."
Scale
- Total investment: $500 billion (over 4 years)
- Initial investment: $100 billion
- Plans to build 10 hyperscale data centers
- Each data center equipped with 100,000–200,000 GPUs
- Once fully built, total computing power will be more than 5 times the current global AI computing capacity
Participants
- OpenAI (providing technology and demand)
- SoftBank (primary investor)
- Oracle (data center operations)
- MGX (UAE sovereign wealth fund)
Why So Much Computing Power Is Needed
According to Sam Altman, achieving AGI requires exponentially growing computing power. GPT-5's training compute is already 10 times that of GPT-4, and future models will only be more compute-hungry.
Energy Challenges
A $500 billion AI data center buildout means astronomical electricity consumption. It's estimated to require around 50GW of power — equivalent to the output of 5 large nuclear power plants.
OpenAI has already begun investing in new energy sources like nuclear fusion and geothermal. The future of AI is not just a competition of algorithms, but also a competition of energy.
References:
Comments
Comments are closed.