Microsoft and OpenAI want to make a Stargate AI super computer
Photo: Drew Angerer/Getty Images
- According to a report from The Information, Microsoft and OpenAI are reportedly planning a joint data center project that could reach a value of USD 100 billion or around 1.585 trillion (USD 1 = IDR 15,852)
This project was revealed to culminate in the launch of a very large artificial intelligence or AI supercomputer called Stargate AI in 2028 as reported by detikINET from Gizmochina, Monday (1/4/2024).
This ambitious plan comes amid a growing need for AI data centers that can handle increasingly complex tasks. The emergence of generative AI technology is the main driver behind this demand.
SCROLL TO CONTINUE WITH CONTENT
Microsoft is expected to bear most of the financial burden for the project, according to sources familiar with the discussions. The project budget is estimated to be 100 times more expensive than current data center operations.
The supercomputer, which will be located in the United States, will be the center of a multi-phase initiative lasting six years. The project is reportedly divided into five stages, with Stargate representing the final stage.
Currently, Microsoft is developing a smaller supercomputer for OpenAI, which is scheduled to be released in 2026.
Obtaining the necessary AI chips is expected to be a significant cost for later phases of the project. This particular chip could cost between $30,000 and $40,000, as indicated by Nvidia CEO Jensen Huang.
Microsoft has demonstrated its commitment to advancing AI capabilities by designing its own custom computing chips. The new data center project is designed to work with chips from multiple suppliers, according to the report.
The Information reports that projected spending on the project could exceed USD 115 billion, exceeding Microsoft's previous year's capital spending on infrastructure by a sizable margin.
Last month, OpenAI CEO Sam Altman also revealed insights about the upcoming GPT-5 model during the World Government Summit in Dubai, stating that the model will surpass its predecessor in all aspects.
The upcoming model is anticipated to be a significant leap in AI capabilities, potentially approaching human-level intelligence, with initial testing showing its ability to decipher ancient languages.
GPT-5 is expected to combine features such as image creation via DALL-E and video creation via Sora, with a tentative release in 2024.