According to Epoch AI data, OpenAI’s annualized revenue stands at approximately $24 billion as of early 2026.

Anthropic follows at approximately $19 billion.

Together, the two companies account for the overwhelming majority of commercial AI revenue among the 5 tracked companies in Epoch AI’s database. 

The remaining companies are generating revenue so modest that their lines compress to the bottom of any chart drawn at this scale.

TL;DR

• By early 2026, OpenAI and Anthropic will together have built an estimated $43 billion in annualized revenue.
• xAI’s annualized revenue is less than 5% of OpenAI’s figure at the same point in time.

Epoch AI collects this data via company disclosures, statements from company executives, and media reports.

Three Years, $43 Billion

ChatGPT launched publicly in November 2022. 

Before that moment, the revenue lines for every company on the chart were effectively zero. By early 2026, OpenAI and Anthropic will together have built an estimated $43 billion in annualized revenue. 

No technology sector in recorded history has created that level of commercial output from a standing start in three years.

For context: 

• It took Amazon eleven years from its 1997 IPO to reach $19 billion in annual revenue. 
• It took Netflix over a decade to cross $10 billion. 

The shape of both curves matters as much as the endpoints. Neither line is showing plateau characteristics as of early 2026. 

Both are still in an exponential growth phase, meaning the quarterly dollar increase continues to rise. 

The $110 Billion Question

OpenAI secured $110 billion in new capital in March 2026, one of the largest private fundraising rounds in corporate history. 

The question the raise raises (pun unintended) is less about valuation and more about what a company generating $24 billion annually still needs $110 billion more for. 

The answer is in OpenAI’s own disclosure. 

In a whitepaper shared with potential investors, the company acknowledged that Microsoft provides “a substantial portion” of its funding and processing power, and that any change in that relationship would significantly impact its operations.

That sentence is unusual in tech fundraising materials. 

A company at $24 billion in revenue disclosing infrastructure dependency as a material risk factor is not describing a temporary arrangement it has outgrown. 

It describes a structural condition that scale alone has not resolved. 

Training and running frontier AI models costs more than almost any other software operation in history. The $110 billion raise is, at least in part, capital directed at reducing that dependency and building the compute infrastructure that would make OpenAI less exposed to a single partner’s decisions.

Microsoft has invested approximately $13 billion in OpenAI across multiple tranches. 

Its Azure cloud infrastructure powers a significant portion of OpenAI’s model training and inference workloads. 

The relationship is mutually reinforcing but asymmetric. 

OpenAI needs Microsoft’s compute. Microsoft needs OpenAI’s models to differentiate its cloud and productivity products. 

The Rest of the Field

xAI, the AI company Elon Musk founded after departing OpenAI and filing legal action against it, has publicly positioned its Grok model as a direct competitor to ChatGPT. 

Musk has described OpenAI as having betrayed its original mission and framed xAI as the legitimate continuation of what OpenAI should have been. 

The Epoch AI revenue chart shows xAI with annualized revenue below $1 billion as of early 2026.

That’s less than 5% of OpenAI’s figure at the same point in time.

Mistral AI, Europe’s most prominent AI startup and the company most frequently cited in European policy circles as the continent’s best answer to American AI dominance, is similarly compressed near the baseline. 

Despite: 

• Significant funding rounds
• EU regulatory goodwill
• Stated policy support for AI sovereignty

Mistral’s commercial output does not yet register at a scale that challenges its American counterparts. 

The gap between European AI ambition and American AI commercialisation, at least as measured by revenue, is not closing.

ELI5

OpenAI and Anthropic have compute advantages, enterprise relationships, model quality benchmarks, and brand recognition that compound with every additional quarter of growth. 

The barriers to entry are not regulatory or technical in the traditional sense. They are financial.

The cost of building and running frontier models at competitive quality is now a number that almost no entrant outside the existing leaders can sustain.

Sources: 

Epoch AI Revenue Database | OpenAI investor whitepaper (March 2026)