Anthropic just passed OpenAI in verified business customer counts. Not in hype. Not in Twitter followers. In the number of companies actually swiping a card each month. According to Ramp's AI Index published on May 13, 2026, 34.4% of businesses on the platform now pay Anthropic. OpenAI sits at 32.3%. Twelve months ago, Anthropic's share was 9%. That is a 26 percentage point climb in one year while OpenAI stayed flat. Here is why the company with less money, less brand recognition, and fewer consumer users is winning the metric that actually pays the bills.
I should be honest about what this data does not show. Ramp tracks roughly 50,000 companies, skewed toward digitally mature firms comfortable with modern finance tooling. It is not the whole market. OpenAI still has around 800 million weekly active consumer users. Claude has roughly 18.9 million monthly active users. Those are different leagues. But if you are building a business on top of AI, or selling AI services to other businesses, the Ramp number is the one that matters. Paying business customers are the revenue engine. Everything else is vanity.
The Vertical Wedge Principle
Here is the framework that explains Anthropic's climb: the Vertical Wedge Principle.
The numbers behind the vertical wedge strategy reshaping enterprise AI.
Most companies try to be everything to everyone. They build a horizontal platform, spray marketing across every industry, and hope the best customers self-select. That is the hard way. The easy way is to pick one vertical, build a product so specific it feels custom, win that vertical, then use the credibility to expand into the next one. Wedge in. Expand out.
Anthropic did this in three layers. Layer one: win technical teams in finance, technology, and professional services. Layer two: launch dedicated products for specific job functions, like Claude for Legal with its 20 plus connectors and 12 practice-specific plugins built alongside Harvey, DocuSign, and Ironclad. Layer three: lower the barrier for everyone else with Claude for Small Business, a dedicated SMB tier that lets a 5-person agency get started without an enterprise sales call.
Ramp economist Ara Kharazian confirmed the pattern to TechCrunch. Anthropic started with a technical customer base, focused on their needs, executed well, and then broadened through tools like Claude for Work. That is the Vertical Wedge in action. You do not need to be the biggest. You need to be the most specific.
The Lazy Way to Steal an Enterprise Market
Let me show you exactly how this works in dollars and cents, because the strategy only matters if it prints money.
The old way to win enterprise customers was brutal. You hired a 200-person sales team. You spent 18 months on SOC 2 compliance. You flew to conferences, bought steak dinners, and prayed the procurement committee would sign before the quarter ended. That process costs millions before you close a single deal.
Anthropic flipped the script. Instead of selling top-down, they let developers adopt Claude through API access and coding tools. Cat Wu, Anthropic's head of product for Claude Code and Cowork, helped transform Claude from a chatbot into an agentic coding and task-automation platform. Developers started using it. Then they told their managers. Then their managers told procurement. By the time the enterprise deal hit the table, the product had already proven itself inside the building.
This is stupid easy to understand when you look at the numbers. Anthropic's strongest adoption came in finance, technology, and professional services. These are industries where a single developer or analyst can test a tool on their corporate card for $20 a month. If it saves them 10 hours a week, the ROI conversation is already over. The tool sold itself.
Then Anthropic layered on institutional partnerships. PwC. The Gates Foundation. The University of Chicago's Becker Friedman Institute using Claude Enterprise to study AI labor market effects. Each partnership is not just revenue. It is a signal to every other institution that Claude is safe, serious, and approved by organizations with real compliance standards.
My read on this: the partnership strategy is the most underrated piece. When a 10-person law firm sees that PwC uses Claude, the trust barrier drops to nearly zero. You are not selling the tool anymore. You are selling the association. Sell Maui, not the flights to Maui.
Now here is the part most people miss. Claude for Small Business is not just a product. It is a customer acquisition machine. Every SMB that signs up at the lower tier becomes a potential upsell to Claude for Work or Claude Enterprise. The math is simple. If Anthropic converts even 10% of SMB customers to a higher tier within 12 months, the lifetime value of that cohort explodes. And because SMBs have almost zero procurement friction, the cost to acquire them is a fraction of a traditional enterprise deal.
Whether Anthropic can sustain this growth rate is genuinely uncertain. Kharazian himself expressed skepticism that the lead will last. OpenAI has Microsoft distribution baked into Office 365 Copilot and Azure OpenAI Service. Those channels may not show up in Ramp's direct-spend data, which means OpenAI's real enterprise footprint could be significantly larger than the chart suggests.
There is also a revenue accounting wrinkle that matters. According to a Semafor report from April 10, 2026, when a customer buys $1 of tokens through a cloud partner like AWS, OpenAI may count only its 20 cent cut as revenue. Anthropic may count the full dollar. That discrepancy could distort comparisons by as much as $8 billion in annualized revenue. So even if Anthropic leads on customer count, the revenue picture is still blurry.
But here is the thing. Customer count is the leading indicator. Revenue is the lagging one. If Anthropic keeps adding paying businesses at this rate, the revenue follows. That is just math.
2031
Three signals inside the same shift
Anthropic gained 26 percentage points in paying business share in 12 months.
From 9% to 34.4% on Ramp's platform, Anthropic overtook OpenAI by building vertical products like Claude for Legal with 12 practice-specific plugins and launching Claude for Small Business to eliminate procurement friction. The bottom-up developer adoption model turned every API user into an internal champion.
Token accounting gaps could distort revenue comparisons by up to $8 billion.
Semafor reported that when customers buy tokens through cloud partners like AWS, OpenAI may count only its 20-cent cut per dollar while Anthropic may count the full dollar. This discrepancy means customer count leadership does not automatically translate to revenue leadership. The real financial picture remains blurry.
Microsoft and Google bundling could neutralize standalone AI wedges by 2031.
OpenAI's distribution through Office 365 Copilot and Azure may not appear in Ramp's direct-spend data, potentially understating its true enterprise footprint. If AI becomes a feature inside Google Workspace or Microsoft 365, the specialized vertical strategy loses leverage. History favors specialists, but nothing compounds forever without adaptation.
Pull back five years from now. The Vertical Wedge Principle is not just an Anthropic story. It is the playbook for the entire next generation of AI companies.
The asymmetric advantage in AI is no longer model quality alone. The top five foundation model providers are converging on similar benchmark performance. The gap between the best model and the fifth best shrinks every quarter. By 2031, I think the difference between leading models will be negligible for 90% of business use cases.
What will not converge is distribution. The companies that win will be the ones that built vertical wedges so deep that switching costs become structural. Think about what Anthropic is doing with Claude for Legal. Twelve practice-specific plugins. Integrations with the tools lawyers already use. Once a 500-person law firm has built workflows around those connectors, the cost of switching to a competitor is not the subscription fee. It is the retraining, the workflow rebuilding, the lost productivity during transition. That is a moat made of habits, not technology.
The compounding flywheel looks like this. Win a vertical. Build integrations specific to that vertical. Accumulate usage data that makes the product better for that vertical. Use the credibility from that vertical to enter the adjacent one. Repeat. Each turn of the flywheel makes the next turn easier.
The contrarian risk is that OpenAI or Google simply bundles AI into existing enterprise software so deeply that standalone AI subscriptions become irrelevant. If Claude's value gets absorbed into a feature inside Google Workspace or Microsoft 365, the wedge strategy loses its leverage. That is a real possibility. Anthropic's bet is that specialized, vertical products will always outperform bundled generalist features. History suggests that bet is right more often than not. Salesforce beat bundled CRM. Slack beat bundled chat. Figma beat bundled design tools. But strategies have shelf lives too. Nothing compounds forever without adaptation.
The builders who understand this pattern have an edge right now. Whether you are selling AI services, building on top of foundation models, or choosing which platform to bet your product on, the lesson is the same. Specificity beats scale in the early innings. The nicher you go, the faster you grow.
What to Build This Weekend
You do not need to be Anthropic to use the Vertical Wedge. Here is how to apply it this weekend, even if you are a solo builder or a small team.
Step one: pick one industry you already understand. Not three. One. If you have worked in real estate, healthcare, or e-commerce, that is your wedge.
Step two: find the most painful workflow in that industry that involves documents, data entry, or repetitive decisions. Talk to three people who do that job. Ask them what takes the most time and makes them want to quit.
Step three: build a tiny prototype using Claude's API or a no-code tool. If you want a fast start, try Ignyte to organize your pitch and traction tracking from day one. It bundles AI-powered pitch practice, co-founder matching, and milestone tracking into a single workspace designed for exactly this stage.
Step four: give the prototype to one of those three people for free. Watch them use it. Write down every place they get confused.
Step five: fix the confusion. Ship again. Get your reps in.
You do not need a CS degree for this. You do not need funding. You need one vertical, one painful workflow, and one weekend of focused building. Things will break. That is normal. Test aggressively, learn in public, and remember that 80% done by someone else is still 100% freaking awesome.
The companies winning the AI market right now are not the ones with the best models. They are the ones solving the most specific problems for the most specific customers. You can do the same thing at a smaller scale starting Saturday morning.
Apply the Vertical Wedge to your own product in three steps.
- Pick one industry you already know. Not three. One. If you have worked in real estate, healthcare, or e-commerce, that domain knowledge is your unfair advantage. Specificity beats scale in the early innings.
- Interview three practitioners about their most painful workflow. Find the repetitive task involving documents, data entry, or routine decisions that makes them want to quit. That pain point is your product's entry wedge.
- Ship a tiny prototype using Claude's API or a no-code tool. Solve that single workflow for that single vertical. Once you have one paying customer, use their credibility to expand into the adjacent use case. Each win makes the next one easier.
Customer count is the leading indicator. Revenue follows.
Anthropic's 26-point climb in paying business share proves that specificity, not scale, wins the early innings of enterprise AI. Vertical products, SMB onramps, and institutional partnerships created a compounding flywheel that a better-funded incumbent could not match on adoption speed. The revenue accounting is still blurry and the bundling threat from Microsoft and Google is real. But the builders who wedge deep into one vertical today will own the switching costs that define tomorrow's moats.