OpenAI has cut its 2030 computing power budget from $1.4 trillion to $600 billion, amid challenges with the Stargate project and revenue projections.
OpenAI has significantly slashed its projected computing power budget for 2030 from $1.4 trillion to $600 billion, as reported by CNBC. This revision was made during a recent presentation to investors, where the company outlined a more realistic timeline and expenditure estimates, citing prior ambitions as overly optimistic.
The company anticipates generating revenues exceeding $280 billion by 2030, with contributions evenly split between consumer and corporate sectors. In 2025, OpenAI reported revenues surpassing $20 billion, according to CFO Sarah Friar; however, other sources indicate a lower figure of $13.1 billion, with actual expenses reported at $8 billion, down from an expected $9 billion.
Stargate Project Fails to Launch
In addition to budget cuts, OpenAI is facing challenges with its Stargate project, a $500 billion initiative launched in early 2025 that has reportedly stalled. Thirteen months into the venture, there have been no hires or construction activity, and disputes among partners have hindered progress.
The Information reports that Stargate lacks essential team members and management, and discussions regarding the ownership and development of facilities have been contentious. OpenAI executives have made multiple trips to Tokyo to negotiate with SoftBank’s CEO Masayoshi Son, but disagreements persist over the flagship campus in Texas.
Attempts for OpenAI to proceed independently have been thwarted by a lack of funding from lenders.
Addressing AI's Energy Concerns
During an event hosted by The Indian Express, CEO Sam Altman addressed misconceptions about the environmental impact of AI technologies. He dismissed claims regarding excessive water usage for data processing as "completely fabricated," emphasizing that the current cooling methods used in data centers have alleviated such concerns.
"This was a real issue back when data centers used evaporative cooling. Now that we don’t do that, you can find claims online like: 'Don’t use ChatGPT. It consumes 17 gallons of water per request.' That is completely false," Altman stated.
While he acknowledged the importance of monitoring overall energy consumption, he refuted comparisons between AI usage and personal device charging, suggesting that actual energy costs per request are significantly lower than often portrayed. Altman argued that training AI requires considerable energy, much like the long evolutionary process humans have undergone.
"Training also requires a lot of energy. It’s about 20 years of life and all the food you eat during that time before you become smart," he explained.
He concluded by noting that AI may already be matching human energy efficiency in response to queries, highlighting the ongoing debate over AI's role in energy systems.
