OpenAI President Reveals Company Will Spend $50 Billion on Infrastructure in 2026 Alone

--- headline: "OpenAI President Reveals Company Will Spend $50 Billion on Infrastructure in 2026 Alone" slug: openai-brockman-50b-infrastructure-2026 category: llms-genai story_number: "09" date: 2026-05-08 sources: - name: Bloomberg url: https://www.bloomberg.com/news/articles/2026-05-05/openai-to-spend-50-billion-on-computing-in-2026-brockman-says domain: bloomberg.com - name: Quartz url: https://qz.com/openai-computing-spending-50-billion-brockman-trial-050526 domain: qz.com - name: SiliconANGLE url: https://siliconangle.com/2026/05/04/openai-president-grilled-30b-personal-stake-duty-humanity/ domain: siliconangle.com - name: Technobezz url: https://www.technobezz.com/news/openai-president-greg-brockman-testifies-company-will-spend-50-billion-on-computing-this-year domain: technobezz.com ---

OpenAI expects to spend $50 billion on computing infrastructure in 2026, President Greg Brockman disclosed under oath this week, a figure so large it rivals the annual capital expenditures of Google, Microsoft, or Meta individually and lays bare the staggering financial demands of building frontier artificial intelligence.

Brockman made the revelation during his second day testifying in the Musk v. Altman trial in Oakland federal court, where Elon Musk is suing OpenAI over allegations that its leadership betrayed the company's founding nonprofit mission. The $50 billion number emerged not as a boast but as a factual data point about the resources required to compete at the cutting edge of AI, and its disclosure under oath gives it more weight than any press release or investor deck ever could.

From $30 Million to $50 Billion in Under a Decade

The trajectory is almost incomprehensible. OpenAI's computing costs have rocketed from approximately $30 million in 2017 to tens of billions of dollars today, according to Bloomberg's reporting on the testimony. That represents a roughly 1,600-fold increase in less than nine years, driven by the development of increasingly sophisticated AI models and the need to serve them to a rapidly expanding user base that now numbers in the hundreds of millions.

To put the $50 billion in perspective, it exceeds the annual GDP of more than 80 countries. At current pricing, that sum could theoretically acquire roughly 1.25 million NVIDIA B200 GPUs. The power draw alone for that kind of computing capacity would exceed 1.5 gigawatts, enough to supply electricity to a small city.

The figure is not exclusively a training budget. It encompasses the full compute envelope: GPU clusters for training frontier models, inference infrastructure for ChatGPT's massive user base, API compute for enterprise customers, and research experimentation capacity.

The Impossible Math

The financial arithmetic behind that $50 billion raises immediate questions. OpenAI's revenue, while growing aggressively, does not come close to covering infrastructure costs of this magnitude. The company generated an estimated $3.7 billion in revenue in 2024 and is projected to reach roughly $11.6 billion in 2025. Even with optimistic projections of $25 billion to $30 billion in 2026 revenue, compute spending alone would consume nearly double the company's income.

Layer on an estimated $5 billion to $10 billion for headcount, research, and operations across a workforce exceeding 4,000 employees, and OpenAI faces an annual loss potentially in the range of $25 billion to $30 billion. The company has raised approximately $20 billion cumulatively, including Microsoft's $13 billion investment and a $6.6 billion round in October 2024 that valued the company at $157 billion. Private investors now value OpenAI at over $850 billion.

The spending gap only makes sense through the lens of Stargate, the $500 billion infrastructure joint venture announced in January 2025 with SoftBank, Oracle, MGX, and other partners. Under this arrangement, SoftBank and its partners provide the capital expenditure while OpenAI rents the resulting capacity, effectively keeping the compute spend off OpenAI's own balance sheet. What Brockman's testimony revealed is that Stargate is not merely an ambitious side project. It is an existential necessity without which OpenAI's current trajectory would be financially impossible.

The Trial's Broader Revelations

Brockman's infrastructure disclosure was just one of several striking revelations from his time on the witness stand. He also confirmed for the first time that his personal equity stake in OpenAI is worth close to $30 billion, a fortune that would place him among the 100 wealthiest people on Earth according to Forbes, despite never having personally invested any money in the company.

When pressed by Musk's attorney Steven Molo about whether that windfall contradicts OpenAI's stated duty to humanity, Brockman pushed back. "Compensation was certainly secondary to the mission," he testified. He later added: "No, I believe that we have developed the most well-capitalized nonprofit in human history."

Molo compared Brockman to someone who robs a bank, a comment Judge Yvonne Gonzalez Rogers struck as argumentative. Brockman also confirmed that OpenAI is exploring an initial public offering, and disclosed a $471 million investment in Stripe and a stake in CoreWeave, the cloud computing provider with a major OpenAI contract.

The trial itself centers on Musk's allegations of unjust enrichment, fraud, constructive fraud, and breach of charitable trust. Musk's legal team has argued he is owed up to $134 billion in wrongful gains. OpenAI has maintained the lawsuit is an attempt by a competitor to inflict damage.

What It Means for the AI Industry

The $50 billion figure carries implications far beyond OpenAI's courtroom drama. At that level of annual spending by a single company, the barrier to competing at the AI frontier becomes effectively insurmountable for startups and most established technology firms alike. The compute cost floor for training a frontier model may already sit between $5 billion and $10 billion, and inference at ChatGPT's scale costs billions more on top of that.

The disclosure also adds granularity to a broader spending trajectory OpenAI has outlined elsewhere. In February, the company told investors it was planning roughly $600 billion in total compute spending through 2030, with separate commitments totaling more than $1.4 trillion earmarked for AI infrastructure. If the current trajectory holds and spending continues to roughly double each year, 2027 could see $75 billion to $100 billion in compute expenditure.

The AI industry is entering a phase where building at the frontier requires the financial backing of sovereign wealth funds or partnerships with the world's largest technology conglomerates. The era in which a scrappy startup could plausibly challenge for frontier AI leadership is, by the numbers Brockman disclosed under oath, definitively over.