Claude Just Stole ChatGPT’s Crown
Anthropic vs OpenAI Is the Headline. But The Real Story Is the Multi‑Model Future.
In just one year, the balance of power in AI chat for businesses has flipped. A payments firm that tracks corporate card spending from over fifty thousand U.S. companies now shows Claude taking roughly 70% of paid AI chat subscriptions, with ChatGPT down to about 30%.
That doesn’t mean “ChatGPT is dead” or that every developer has moved overnight, but it does tell us where serious budgets are flowing right now. The same period has seen Claude rocket up the App Store charts, fueled by word of mouth from engineers and a wave of users who are frustrated with OpenAI’s moral stances, questionable PR, product decisions and defence contracts.
In simple terms, enterprises are quietly rewiring their AI stack, and for the first time since 2022, OpenAI is no longer the obvious default.
TL;DR - What this chart shows.
In just 12 months, Claude has gone from a rounding error to roughly 70% of U.S. enterprise AI chat subscription spend, while ChatGPT has fallen from around 90% to about 30% in this specific dataset.
The chart comes from Ramp’s corporate card and bill‑pay data across 50,000+ U.S. businesses and only counts paid chat products from OpenAI and Anthropic (Claude Team/Max/Enterprise, ChatGPT Plus/Pro/Business).
It shows where serious budgets are flowing today, not total users, not global market share, and not API usage or tools like Copilot, Gemini, or Perplexity.
Bigger picture: what’s really happening ( user-wide)
TechCrunch and others are tracking a visible shift: Claude’s consumer app has surged to the top of the U.S. App Store and stayed there on the back of Super Bowl ads, strong word of mouth from developers, and growing paid conversions.
Sensor Tower’s U.S. mobile app data for ChatGPT on Saturday, February 28, reported by outlets like TechCrunch, Yahoo Finance, and others say U.S. ChatGPT app uninstalls were 295% higher than the previous day (vs a typical 9% day‑over‑day uninstall change over the prior 30 days), after the Pentagon/Department of War deal news broke.
Claude downloads up 37% on Friday 27 Feb and 51% on Saturday 28 Feb, around the time Anthropic rejected the deal. Appfigures estimates an 88% day‑over‑day surge in Claude downloads on 28 Feb, with Claude reaching No. 1 in the U.S. App Store and topping charts in several other countries.
Yahoo Finance frames the App Store spike and the Pentagon‑contract backlash as the first large‑scale “AI boycott” moment: a chunk of power users and enterprises are actively moving their day‑to‑day workflows off ChatGPT and onto Claude.
Underneath the headlines, multiple surveys still show OpenAI models as the most commonly used tools overall, but Claude is rapidly becoming the default choice for deep coding, agents, and long‑context work among serious builders.
What this ‘chart’ doesn’t tell us
It doesn’t include GitHub Copilot, Gemini, or vertical assistants inside IDEs and SaaS tools, which still handle a huge share of real‑world coding.
It doesn’t capture free‑tier usage, hobbyist developers, or the rest of the world outside the U.S., where pricing, regulation, and cloud partnerships look very different.
It doesn’t prove the game is over; market share that moves from 90% to 30% in a year can move again just as fast if the product, pricing, or trust dynamic changes.
While this chart makes the story look like a straight Claude‑versus‑ChatGPT cage match, the real landscape is broader.
Codex‑style capabilities now show up mainly through tools like GitHub Copilot and Cursor, not as a separate “Codex” product, and those assistants still dominate in‑editor usage for many developers. At the platform layer, models from Google (Gemini on Vertex AI) and Amazon (Bedrock‑hosted models like Claude and others) are quietly embedded into cloud stacks and enterprise apps, so a lot of AI work never shows up as a Claude or ChatGPT subscription line item.
In other words, Anthropic and OpenAI are fighting a visible subscription war on the surface, while a whole ecosystem of IDE assistants, cloud platforms, and vertical tools is routing traffic to whichever model best fits the job underneath.
What this means for serious builders and typical users?
For serious builders, the message is clear: you can’t be “an OpenAI shop” or “an Anthropic shop” anymore. you need a multi‑model strategy, with Claude as a first‑class engine for deep work and ChatGPT (plus others) as interchangeable modules where they make sense.
Teams that build their workflows, agents, and internal tools on top of a single vendor’s stack are now carrying real concentration risk; in 2026, resilience means routing, evaluating, and swapping models the way we already swap cloud regions and CDNs.
For typical users, this back‑and‑forth is a net win: switching tools is getting easier, features are shipping faster, and the “best” model will change several times before the year is out—so the smart move is to stay flexible, keep your data portable, and treat LLMs as apps you can swap, not religions you have to defend
The big lesson from this OpenAI–Anthropic tug‑of‑war is not “pick a winner,” it’s “stop betting your stack on a single model.”
If enterprise spend can swing from 90% ChatGPT to 70% Claude in a year, it can swing again just as fast when the next leap in quality, safety, or pricing arrives. For builders and everyday users, the smart move in 2026 is to stay flexible:
keep your data portable,
design your workflows so you can swap models, and
treat LLMs like interchangeable tools rather than brands you swear loyalty to.
The real moat won’t be “we chose the right vendor early,” it will be having the architecture and the habi to adapt every time the landscape shifts.
If you’re a founder, infra/operator lead, or investor working on AI infra, tooling, or real‑world deployments and want to pressure‑test how this maps to your portfolio or roadmap, hit reply with “@AI Unfiltered - ” and one line on what you’re looking at.
I’m also building a small set of operator‑grade briefings and deep dives for teams and investment firms who want the “no‑BS” version of what’s real and what’s noise. Add yourself to the early access list here: - Early Access Waitlist – Fill out form
Subscriber to intelligent founder AI for the deep dives and the podcast.