OpenAI will pay Oracle $300 billion over the course of five years to fuel Sam Altman’s AI ambitions by providing five gigawatts of compute capacity.<\/p>\n
At least that’s what unnamed sources familiar with the matter tell the Wall Street Journal, which reports<\/a> that the contract is set to begin in 2027, giving the AI startup a little over a year to figure out how or perhaps who’s going to be left holding the bill when it falls due.<\/p>\n As it stands, OpenAI can’t afford the checks it’s writing. As of June, the AI flag bearer has annual recurring revenue of $10B. To be clear, that’s revenue. The company isn’t expected to turn a profit until at least 2029. Perhaps SoftBank is about to open its wallet after reportedly agreeing to chip in $19B to the OpenAI-led Stargate project that aims to build a series of giant AI Datacenters.<\/p>\n Today’s news comes just over a month after OpenAI announced<\/a> that Oracle had agreed to furnish it with an additional 4.5 gigawatts of compute capacity, bringing its total compute commitment to 5 gigawatts, enough for two million additional GPUs. At the time, we estimated the cost of those GPUs at nearly $100 billion. However, that figure doesn’t take into consideration the cost of the facilities and power plants necessary to support all those chips, and assumed OpenAI and friends would be paying today’s prices for them.<\/p>\n As we reported last week, OpenAI will reportedly<\/a> begin fielding its own in-house silicon developed in collaboration with Broadcom, which may also factor into its cost analysis.<\/p>\n While it remains to be seen how exactly OpenAI plans to pay its bills, it seems that Oracle’s CTO and Founder Larry Ellison is already reaping the benefits of the arrangement. On Tuesday Ellison inched<\/a> closer to overtaking Elon Musk as the world’s richest man after the database giant turned GPU slinger’s share price surged more than 30 percent in after hours trading.<\/p>\n