In the coming years, OpenAI plans to make significant adjustments to its computing infrastructure providers. The company currently relies heavily on Microsoft for its computational needs, but it anticipates a shift towards SoftBank-backed Stargate, a project dedicated to data center capacity.
By 2030, OpenAI aims to obtain around 75% of its data center capacity from Stargate. This move represents a notable departure from Microsoft, which currently supports the majority of OpenAI’s power requirements. However, this transition will be gradual, as OpenAI intends to continue utilizing Microsoft’s data centers in the interim.
OpenAI’s overall costs are projected to escalate dramatically. The company estimates a $20 billion cash outlay by 2027, considerably higher than its reported $5 billion expenditure in 2024. Moreover, OpenAI anticipates that the costs associated with running AI models (inference) will surpass the expenses of training AI models by 2030.
Original source: Read the full article on TechCrunch
