OpenAI's CFO Sarah Friar published financial details on January 18, 2026, that the company has historically kept close. The blog post reveals annualized revenue exceeding $20 billion and compute capacity reaching approximately 1.9 gigawatts, a near-tenfold increase from 2023.
The disclosure arrives two days after OpenAI began testing advertisements in ChatGPT's free tier and the new ChatGPT Go subscription.
The numbers track suspiciously well
Here's what Friar shared: compute grew from 0.2 GW in 2023 to 0.6 GW in 2024 to roughly 1.9 GW in 2025. Revenue followed an almost identical curve, climbing from $2 billion to $6 billion to more than $20 billion over the same period.
That correlation is the entire point. Friar frames it as proof that demand isn't the bottleneck. Give OpenAI more compute, the argument goes, and revenue follows automatically.
The math is neat. Maybe too neat. A company burning through roughly $17 billion annually on infrastructure has strong incentives to convince investors that spending equals growth. Whether the correlation holds at higher scales remains unproven.
What 1.9 gigawatts actually means
To put that compute figure in context: 1.9 GW is roughly equivalent to the electricity consumption of two million American households. One analysis compared it to about 12% of Israel's peak electricity output.
OpenAI has diversified beyond its original dependence on Microsoft Azure. Friar noted the company now manages compute as an active portfolio, using premium hardware for frontier model training and lower-cost infrastructure for high-volume inference workloads.
The infrastructure commitments backing this expansion are staggering. Sam Altman disclosed in November 2025 that OpenAI has locked in approximately $1.4 trillion in data center deals over the next eight years, spanning partnerships with Oracle, Microsoft, Amazon Web Services, Nvidia, AMD, and Broadcom.
The advertising question
OpenAI announced on January 16 that it would begin testing ads in ChatGPT for US users on the free tier and the new $8-per-month ChatGPT Go subscription. Plus, Pro, Business, and Enterprise tiers remain ad-free.
Sam Altman has expressed reservations about advertising for years. In interviews, he called combining ads with AI "uniquely unsettling." But the financial pressure is real: with only about 5% of weekly active users paying for subscriptions as of mid-2025, the company needed another way to monetize hundreds of millions of free users.
OpenAI's advertising announcement emphasizes that responses won't be influenced by ads and that conversation data won't be sold to advertisers. Ads will appear at the bottom of responses, clearly labeled as sponsored content.
The potential upside is significant. Meta generates nearly $50 in average revenue per user annually in North America through advertising. If OpenAI managed even $20 per free user across its roughly 900 million user base, that's another $18 billion in potential revenue.
The competition for AI revenue
OpenAI's disclosure lands amid an increasingly crowded field.
Anthropic's trajectory has been dramatic. The company grew from approximately $1 billion in annualized revenue at the start of 2025 to over $5 billion by August, reaching an estimated $9 billion by year-end. Anthropic is projecting $20 to $26 billion in ARR for 2026. The company recently closed a $13 billion Series F round at a $183 billion valuation, with reports of a new round potentially valuing it at $350 billion.
Elon Musk's xAI presents a different picture. The company guides to roughly $500 million in standalone revenue for 2025, though it claims 600 million monthly active users across X and the Grok app. xAI just raised $20 billion in a Series E round, bringing total funding to over $40 billion. The company is burning approximately $1 billion monthly on infrastructure while targeting profitability by 2027.
The combined ARR of OpenAI, Anthropic, and xAI likely sits around $30 billion currently. By the end of 2026, if growth projections hold, that figure could reach $75 billion or higher.
What Friar didn't say
The disclosure conspicuously avoids profitability metrics. OpenAI's burn rate reportedly hit $17 billion in 2025, with cumulative losses projected to reach $44 billion through 2028 before the company expects positive cash flow in 2029.
There's also the question of whether the compute-revenue correlation can sustain itself. Each additional gigawatt of capacity requires not just capital for hardware but ongoing energy costs, cooling infrastructure, and operational overhead.
Friar's post gestures toward future revenue models beyond subscriptions and APIs. She mentions licensing, IP-based agreements, and outcome-based pricing for applications in drug discovery, energy systems, and financial modeling. These remain mostly theoretical.
The 2026 priority
OpenAI's stated focus for 2026 is "practical adoption," particularly in healthcare, science, and enterprise applications. The company launched ChatGPT Health in early January and has been expanding its enterprise offerings.
The September 2025 projection had OpenAI targeting $30 billion in revenue for the full 2026 calendar year. If the company maintains its tripling pattern, it would end 2026 closer to $50 billion in ARR. Analyst estimates vary widely, with some industry insiders at the Cerebral Valley conference betting on the more conservative $30 billion figure.
For context, $50 billion in annual revenue would put OpenAI alongside companies like Intel, Uber, and Nike. The difference: those companies are profitable.
OpenAI's path to an IPO, reportedly being explored for late 2026 at valuations approaching $1 trillion, depends on demonstrating that revenue growth can eventually outpace infrastructure spending. Friar's disclosure reads as the opening argument for that case.
Whether investors buy it may determine how much longer the AI infrastructure buildout continues at its current pace.
