Anthropic overtakes OpenAI in enterprise

Ramp said on May 13 that Anthropic had passed OpenAI in paid business AI adoption for the first time in the company’s AI Index, a monthly measure based on American business spend. Anthropic’s adoption rose 3.8 percentage points in April to 34.4% of businesses, while OpenAI’s fell 2.9 points to 32.3%, according to the report by Ramp economist Ara Kharazian. The shift lands as Anthropic and OpenAI compete more directly for developers and enterprise teams building coding tools and software agents. It also comes as both companies adjust how much usage they allow and how they charge for compute-heavy products. ### How did Ramp measure the change in enterprise adoption? Ramp describes the AI Index as a monthly measurement of AI adoption by American businesses using card and bill-pay transaction data from its customer base. The May 13 report said Anthropic “passed OpenAI in business adoption” in the latest release, with April usage showing Anthropic at 34.4% and OpenAI at 32.3%. Ramp had reported five months earlier that OpenAI still led paid AI adoption by businesses at 35.8% in October 2025, while Anthropic stood at 14.3%, showing how quickly the gap closed. (ramp.com) Ara Kharazian, Ramp’s lead economist, wrote that the April reading marked the first time Anthropic moved ahead in the company’s paid adoption data. Ramp’s vendor page for Anthropic separately said 70% of organizations with a generative AI vendor used Anthropic as of May 2026, and that adoption was highest among enterprise companies at 75%. ### Why is Anthropic gaining ground with technical teams? (ramp.com) Anthropic has pushed Claude Code and related developer products more aggressively in recent months. An April 23 engineering post from Anthropic said the company had traced quality issues affecting Claude Code, the Claude Agent SDK and Claude Cowork to three separate changes, underscoring how central those products had become to users. Anthropic’s home page also markets Claude Opus 4.7 for “coding, agents, vision, and complex professional work.” (ramp.com) Anthropic’s own enterprise announcements have focused on large business deployments. On May 14, the company said PwC was deploying Claude to build technology, execute deals and reinvent enterprise functions for clients, and earlier announcements highlighted products for financial services and small business. Those releases do not prove causation for Ramp’s index, but they show Anthropic targeting the same enterprise buyers reflected in spend data. (anthropic.com) ### What changed on Claude usage limits? Axios reported on May 14 that Anthropic had imposed new limits on what paying customers could do with subscriptions, creating an opening for OpenAI to attract heavy users of coding agents. Axios had also reported in April that Anthropic blocked Claude subscriptions from powering some third-party agent tools, including OpenClaw. (anthropic.com) Anthropic said on May 6 that it had raised usage limits for Claude Code and the Claude API after signing a compute partnership with SpaceX. The company said the SpaceX agreement would give it access to more than 300 megawatts of new capacity, or more than 220,000 Nvidia GPUs, within a month, and that the added capacity would directly improve service for Claude Pro and Claude Max subscribers. (axios.com) ### How is OpenAI trying to win the same customers? OpenAI is pitching its platform more explicitly around agents. OpenAI’s API site says its platform helps users “build, deploy, and optimize production-ready agents,” with both a visual Agent Builder and a code-first Agents SDK. Axios reported on April 21 that OpenAI was mobilizing consulting partners and emphasizing its compute edge as it tried to win back enterprise customers from Anthropic. (anthropic.com) OpenAI’s current pricing pages show the company competing on both subscriptions and token-based API access. ChatGPT Business is listed at $25 per user per month when billed monthly, while the API pricing page lists GPT-5.5 at $5 per 1 million input tokens and $30 per 1 million output tokens, with lower rates for cached input and batch processing. Those published prices give enterprise buyers a different cost model from flat-rate subscription products that can be strained by always-on agents. (openai.com) ### What should companies watch next? May 6 and May 14 are the two dates that frame the current tradeoff. Anthropic said on May 6 that new SpaceX-backed capacity would raise Claude Code and API limits, and Axios reported on May 14 that Anthropic’s subscription restrictions were giving OpenAI a chance to court power users. Ramp said its AI Index is updated monthly, making the next release the clearest test of whether Anthropic’s April lead holds. (openai.com) OpenAI’s API pricing page and Anthropic’s newsroom remain the public places where both companies are posting product, pricing and capacity changes for enterprise customers. (ramp.com) (anthropic.com)