For the first time since ChatGPT launched, Anthropic has overtaken OpenAI in business adoption. According to spending data from Ramp—one of the largest corporate card platforms in the US—Claude now leads GPT among business users, with Anthropic capturing 34.4% of AI tool spending share compared to OpenAI's 32.3%. That gap might sound small, but in a market OpenAI has dominated for three years, any lead for a competitor signals a genuine inflection point.
The revenue trajectories tell an even starker story. Anthropic reached a $30 billion annualized revenue run rate in April 2026, up from $1 billion just fifteen months earlier. In Q2 2026, the company expects $10.9 billion in revenue and its first-ever operating profit of $559 million. OpenAI, by contrast, is projecting a $14 billion loss for 2026, with no path to profitability until 2029 or 2030. Both companies are heading toward IPOs this year, but they're telling public markets completely different stories about AI economics.
This isn't just a data point for AI watchers. If you're a CTO evaluating which AI platform to standardize on, or a CFO approving multi-million-dollar AI budgets, this shift reveals which business model is actually working at enterprise scale.
The Data: What RAMP Spending Actually Measures
Ramp's AI Index tracks real spending by business customers—not survey responses, not downloads, not self-reported usage. It's based on transaction data from companies using Ramp's corporate cards and expense management tools. When a company shows up in Ramp's data, it means they've allocated budget. That filters out free-tier users and casual experimenters, leaving only organizations that have committed dollars.
The 34.4% vs 32.3% figures represent share of business AI spending among Ramp customers, not global market share or total revenue. Ramp's customer base skews toward US-based mid-market and growth-stage companies, so the data reflects that demographic well but may not represent large Fortune 500 deals or international markets equally. Still, it's the most reliable signal we have for how businesses are choosing between AI vendors when spending their own money.
OpenAI held a commanding lead through most of 2023 and 2024. The rise of Claude 3 in early 2024 started closing the gap, and Claude 3.5 Sonnet—released mid-2024—accelerated it significantly. By the time Ramp's 2026 data was published, Anthropic had pulled ahead. The trend line matters as much as the snapshot: Anthropic's share has been rising steadily quarter over quarter, while OpenAI's has plateaued and declined slightly. That's not a random fluctuation—it reflects cumulative decisions made by teams who tested both and chose.
Why Businesses Are Choosing Claude
There isn't one single reason Anthropic is winning business adoption. It's a combination of capability improvements, operational reliability, and a pricing model that scales predictably for high-volume production use.
Claude's Performance on Enterprise Tasks
The tasks that matter most in business settings are different from the ones that dazzle in demos. Businesses care about following long, detailed instructions without drifting, processing large documents reliably, writing clean functional code without constant correction, maintaining consistency across long sessions, and not hallucinating facts in customer-facing or compliance-sensitive outputs.
Claude 3.5 Sonnet and Claude 3.7 Sonnet score well on all of these. On coding benchmarks like SWE-bench, Claude 3.7 Sonnet has posted top results. On long-document analysis, Claude's 200K token context window is a practical advantage that saves businesses from having to chunk documents manually. For legal teams processing contracts, finance teams analyzing earnings reports, or compliance teams reviewing regulatory filings, that context window translates directly into fewer API calls and lower operational cost.
The Trust and Safety Angle
Anthropic markets itself as a safety-focused AI lab, and that positioning resonates with enterprise buyers—particularly in regulated industries like finance, healthcare, and legal. When procurement teams are evaluating AI vendors, "safe by design" is a selling point that moves budget. Whether or not Anthropic is objectively safer than OpenAI at the model level is debatable, but Anthropic has done more public work explaining its approach to AI alignment and Constitutional AI, and that transparency builds confidence with risk-averse buyers.
In conversations with enterprise security leaders, I've heard the same refrain: they want AI vendors who can articulate their safety processes in audit-ready language, not just marketing copy. Anthropic's public documentation on Constitutional AI and model behavior constraints gives compliance teams something concrete to reference when writing internal AI governance policies.
Pricing Transparency and Predictability
Claude's API pricing is competitive and relatively predictable. For businesses running high-volume inference—processing thousands of documents, powering internal tools, running customer-facing features—per-token pricing needs to be stable and calculable. OpenAI has changed its pricing structure multiple times and introduced tiered pricing tiers that some businesses find harder to plan around. Anthropic has generally been more stable on this front, which matters when you're committing to build production systems on a model.
Over 500 companies now spend more than $1 million annually on Anthropic's Claude platform, and eight of the Fortune 10 are customers. That's the foundation of a profitable business. A free-tier consumer base generating massive inference costs without proportionate revenue is not.
The Revenue Structure Difference
Approximately 85% of Anthropic's revenue comes from enterprise and developer customers. OpenAI's mix runs in the opposite direction: roughly 85% tied to ChatGPT consumer subscriptions, with an estimated 95% of those users paying nothing. That structural difference explains why one company is profitable and the other is projecting $14 billion in losses this year.
Enterprise customers generate three to five times more revenue per token than consumer users. Their query patterns are more deterministic and therefore cheaper to serve, and their contracts are sticky. They're not experimenting—they're deploying AI into production workflows that generate measurable business value. When a Fortune 500 company signs a seven-figure Claude contract, that revenue is recurring and budgeted. When a consumer downloads ChatGPT for free, that's a user acquisition cost with uncertain monetization.
OpenAI's computing expenditure will reach $121 billion in 2028 alone, with a projected loss of $74 billion that year. Anthropic, by contrast, projects $17 billion in positive cash flow in 2028 on $70 billion in revenue, with gross margins approaching 77%. The divergence traces directly to the client mix.
Where OpenAI Still Leads
Fairness requires acknowledging that OpenAI hasn't been standing still, and there are areas where GPT models—particularly GPT-4o and o3—still have real advantages.
Ecosystem and Tooling
OpenAI has a significant head start in ecosystem development. The number of libraries, frameworks, tutorials, and integrations built specifically for OpenAI's API is substantially larger than what exists for Anthropic's API. Tools like LangChain, LlamaIndex, and most major agent frameworks defaulted to OpenAI early and still treat it as the primary interface. That legacy infrastructure means development teams can often move faster when building on GPT, simply because more things are pre-built.
Multimodal Capabilities
GPT-4o's multimodal capabilities—voice, image, and video understanding—are broader than Claude's current feature set. For applications that need real-time voice interaction or image generation, OpenAI's integrated offering is more complete. Sora for video generation and DALL·E for images give OpenAI a creative media stack that Anthropic doesn't yet match. For businesses in content production, marketing tech, or media, that difference is material.
Brand Recognition
ChatGPT is still the name most executives know. When a procurement decision goes up the chain to someone who doesn't follow AI news closely, "we're using GPT" often encounters less friction than "we're using Claude." That's a soft advantage, but real. OpenAI's enterprise tier also includes features like dedicated capacity, custom data agreements, and on-prem deployment options that large enterprises sometimes require. Anthropic's enterprise agreements are catching up, but OpenAI's relationships at the Fortune 500 level run deeper.
What This Means for Enterprise Buyers
If you're a CTO or VP of Engineering evaluating AI platforms, the RAMP data and revenue trajectories tell you something important: the market is bifurcating. There's the consumer AI path, where massive scale subsidizes free users and monetization remains uncertain. And there's the enterprise AI path, where revenue-per-customer is high enough to cover infrastructure costs and generate profit.
For technical leaders, the question isn't "which model is smarter?" It's "which vendor has a sustainable business model that will support my production workloads for the next five years?" Anthropic's profitability trajectory suggests they can invest in infrastructure, support, and model improvements without burning through investor capital. OpenAI's path to profitability is less certain, which introduces risk if you're building critical business systems on their platform.
For CFOs and business leaders, the spending data from RAMP is a leading indicator. When peer companies in your industry are shifting budget from OpenAI to Anthropic, that's a market signal worth investigating. It suggests that enterprise buyers—people who care about ROI, operational cost, and vendor stability—are finding more value in Claude than ChatGPT for production use cases.
The practical takeaway: if you're evaluating AI vendors today, test both. But pay close attention to total cost of ownership, not just per-token pricing. Factor in context window efficiency, hallucination rates on your specific tasks, integration complexity, and vendor financial stability. The cheapest API today might not be the most cost-effective platform in production, and the most popular brand name might not be the vendor best positioned to support your business long-term.
The Bottom Line
Anthropic's lead in business adoption is narrow—34.4% vs 32.3%—but the trend is unmistakable. Enterprise buyers are choosing Claude for production workloads, and that choice is driving Anthropic to profitability while OpenAI faces mounting losses. The divergence traces to business model: 85% enterprise revenue vs 85% consumer revenue. One generates margin, the other generates scale without profit.
For the AI industry, this shift settles a fundamental question about AI economics: profitability comes from enterprise discipline, not consumer virality. Both companies are heading toward IPOs this year, but they're selling public markets very different visions. Anthropic is arriving with a profitable quarter already in hand. OpenAI is asking investors to fund additional years of mounting losses in the hope that scale eventually translates to profit.
For enterprise buyers, the lesson is simpler: the vendor with the strongest balance sheet is the safer bet for critical production systems. Anthropic's profitability gives them runway to invest, iterate, and support customers without depending on the next funding round. That's the kind of stability enterprise infrastructure decisions require.
