OpenAI's $278 Billion Five-Year Burn Projection

A July investor briefing prepared by OpenAI for a compute transaction has emerged as this week's most consequential document for assets orbiting OpenAI chief executive Sam Altman, and Worldcoin (WLD) — the iris-scan identity protocol he co-founded — is the ticker traders keep pulling back into focus. The deck projects cumulative negative free cash flow of $278 billion between 2026 and 2030, meaning OpenAI expects to burn more cash than it generates for five straight years and to plug the gap with fresh capital or debt.

Revenue is forecast to climb from $36 billion this year to $350 billion by 2030 — roughly a tenfold increase that implies about 77% average annual growth and $840 billion in cumulative sales over the window. The catch sits on the cost side: compute and infrastructure spending is projected at $856 billion, $16 billion more than the revenue OpenAI expects to book over the same five years.

The new figures supersede more conservative guidance from February, when the company told investors its compute spend target through 2030 was around $600 billion and 2030 revenue above $280 billion. That marks a roughly 43% uplift on spending, though February covered compute only, while the July figures bundle infrastructure as well. Notably, the burn estimate itself improved: a May version showed about $305 billion in negative free cash flow, some $27 billion worse, because partners such as Oracle finance many data centers and Nvidia is negotiating to guarantee data-center debt, keeping that capital expenditure off OpenAI's own books.

The company raised $122 billion in March at an $852 billion valuation — roughly 44% of the projected five-year gap — with the money expected to run out in 2028. Investors are already sounding out a follow-on round near a $1.2 trillion valuation, about 41% above the March level. OpenAI filed confidentially to go public in June, but Altman says AI safety concerns will keep the company private this year.

Compute Race Re-Rates Power Assets

The same numbers explain why electricity, rather than silicon, is emerging as the binding constraint — and why the energy infrastructure assembled by Bitcoin miners over the past decade is quietly re-rating. According to OpenAI's own announcement, Stargate, launched in January 2025 to install 10 gigawatts of US AI infrastructure by 2029, had already passed its original goal as of April, adding 3 gigawatts in the prior 90 days. The company lists what a single gigawatt actually requires: power availability, land, permits, transmission lines, labor and construction partners. In January, OpenAI and SoftBank invested $500 million in SB Energy to develop a 1.2-gigawatt data center in Milam County, Texas.

Demand estimates underline the squeeze. Data centers are on track to consume 565 terawatt-hours of electricity in 2026, up 26% year over year, with AI-optimized servers accounting for close to 31% of that. The International Energy Agency (IEA) projects consumption roughly doubling to about 945 terawatt-hours by 2030 and warns that grid constraints could delay up to 20% of planned projects. The US interconnection queue stands near 2,600 gigawatts, and New York imposed the first statewide ban on hyperscale data-center construction in July.

Against that scarcity, miners holding energized sites are the natural sellers. Leased AI-ready facilities now value around $27 million per megawatt, versus about $3 million per megawatt for operating but unleased mining rigs, while companies with AI or high-performance computing (HPC) contracts trade near 12.9 times enterprise value to forward sales, against 3.7 for pure miners. Capital is following: Core Scientific paid $41.9 million to cancel roughly 15 exahash per second (a measure of mining computing power) of next-generation mining hardware, IREN targets a completed transition by December 31, 2026, and Cipher Digital is expected to exit mining by end-2027. Crusoe, which sold its Bitcoin mining business to NYDIG in 2025, now books contracts reportedly exceeding $140 billion; Robinhood Ventures Fund I put $25 million into its Series F at a $3.9 billion valuation, lifting post-money value to $30.9 billion. The wider altcoin market reads this as confirmation that AI infrastructure is absorbing capital once earmarked for proof-of-work.

What the WLD Narrative Still Lacks

COINOTAG's reading: nothing in the briefing mentions Worldcoin, and neither OpenAI nor Tools for Humanity, the company that Worldcoin, has disclosed any corporate tie — the linkage is narrative proximity to Altman, not a confirmed partnership, and traders should treat it as such. Attention has nonetheless been persistent: COINOTAG's coverage of six AI misalignment cases moved the token's conversation, as did the GPT-6 Astra report, in which the model cracked a 1941 Enigma message.

With OpenAI's own announcement confirming the Stargate build-out, the $856-billion-versus-$840-billion arithmetic gives bulls scale and bears FUD material in equal measure, while the fear and greed index will likely keep setting WLD's short-term tempo. Traders tracking Worldcoin should watch whether a $1.2 trillion round materializes before year-end — and, on the infrastructure side, whether the grid bottlenecks the IEA flags begin to slow the very build-out that is repricing miners.