Enterprise artificial intelligence adoption is expanding rapidly, but the financial rewards are heavily concentrated. New data from fintech firm Ramp reveals that just 1% of customers account for 80% of enterprise revenue at both OpenAI and Anthropic—a level of concentration risk “unseen in any other software category,” according to Ramp lead economist Ara Kharazian.
Revenue Concentration Raises Red Flags
The finding, shared by Kharazian on X and LinkedIn this week, underscores a precarious dependency for two of the world’s most valuable private AI companies. While thousands of businesses are now paying for generative AI tools, the vast majority of revenue flows from a minuscule cohort of power users.
“This is a level of concentration risk unseen in any other software category we track,” Kharazian wrote, noting the pattern has held steady even as the customer base widens.
Anthropic’s Two-Customer Dependency
At Anthropic, the concentration has a name: coding assistants Cursor and GitHub Copilot. Together, these two tools drove approximately $1.2 billion of the company’s $5 billion revenue milestone last year—nearly a quarter of total revenue from just two customers, VentureBeat reported.
One of those customers, GitHub, is owned by Microsoft, which has also invested $13 billion in Anthropic’s chief rival, OpenAI, creating a complex web of competitive and financial ties.
Anthropic’s broader metrics tell the same top-heavy story at scale. The company disclosed to investors it now has roughly 6,000 customers spending at least $100,000 annually, up sevenfold in a year. Meanwhile, customers spending $1 million or more annually more than doubled to over 1,000 in about two months.
Enterprise and startup API calls, priced by usage rather than flat fees, drive roughly 80% of Anthropic’s total revenue, according to research firm Sacra—mirroring the concentration Ramp captured independently.
OpenAI’s Disclosures Show Wide Usage, Not Revenue Breakdown
OpenAI’s picture is less transparent on revenue concentration but shows broad usage. More than 9,000 organizations have processed over 10 billion tokens through its API, and nearly 200 have exceeded 1 trillion tokens each, the company said in its State of Enterprise AI 2025 report.
However, OpenAI has not disclosed what share of total revenue any single customer represents, leaving the 80%-from-1% figure from Ramp as the clearest signal of dependency.
Market Share Shifts as Anthropic Pulls Ahead
The spending structure helps explain uneven growth between the two labs. Enterprise API spending has shifted sharply toward Anthropic, which now commands around 40% of the market against OpenAI’s 27%, according to Menlo Ventures’ 2025 State of Generative AI in the Enterprise report.
Anthropic’s share has climbed from just 12% in 2023, when OpenAI held 50%. Google has risen to 21% over the same period, meaning three companies now account for roughly 88% of enterprise large language model (LLM) API usage combined.
Ramp’s more recent data shows Anthropic holding roughly 44% of US business AI spend as of July 2026, compared to OpenAI’s 40%, though OpenAI is growing faster in the current quarter.
AI Pricing Model Drives Uneven Spend
The distinction matters because AI costs aren’t set simply by headcount. Conventional software contracts are usually priced per seat, but AI adds another variable: how much the software actually does. That dynamic allowed two coding tools to become a quarter of Anthropic’s business almost overnight.
Financial Services: Next Wave of Heavy Users?
Financial services may show where the next wave of heavier users could emerge outside tech. Ninety-five percent of financial firms report broad or embedded use of newer AI in their data and technology work, according to a PYMNTS Intelligence August 2026 Enterprise AI Benchmark Report.
Eighty percent of those firms expect spending to rise over the next 12 months, with none surveyed expecting to cut it. The firms have also reached majority adoption in 27 of 75 tasks tracked—more than healthcare and media combined.
Open Question: Will Pattern Spread Beyond Tech?
The open question is whether that pattern spreads beyond companies built around technology, or stays concentrated in firms like Cursor, built entirely around AI from day one. Banks, payments firms, and retailers handle enormous transaction volumes, and even modest AI usage there could add up fast once systems move from pilots into production.
But higher adoption doesn’t guarantee the same spending curve, and Anthropic’s own Cursor-and-Copilot dependency is a reminder of how much risk can sit inside a fast-growing revenue line.
For investors eyeing potential IPOs from either company, the 80%-from-1% dynamic presents a math problem: can these labs diversify revenue fast enough to justify sky-high valuations, or will concentration remain a structural vulnerability?
