IDC's June 2026 report says no frontier model is enterprise ready on its own, but Anthropic's January to May pricing and engagement changes make a focused re evaluation worth doing.
For enterprise architects and procurement teams who last tested Claude before this year, IDC's June 2026 report is a reason to revisit the math, not a coronation.
The analyst firm's study, summarized by The Register, lands on a deliberately uncomfortable conclusion: "Currently, no frontier model company is mature enough to be evaluated as an enterprise AI provider on its own." Anthropic, in IDC's reading, is simply the lab "running at full speed to get there before its competitors." That framing is useful precisely because it refuses to settle the question IDC's enterprise clients are now asking their procurement teams: do we restart the Claude evaluation we paused six months ago?
The empirical case for re-opening that work rests on three measurable changes between January and May 2026. Anthropic moved off seat-based pricing toward usage-based pricing across parts of its enterprise catalog that earlier evaluations assumed would be billed per seat — a shift The Register notes began in January and continued through the window. The company produced well over 100 public interactions aimed at enterprise buyers across the January–May period, the highest sustained enterprise-marketing cadence of the three frontier labs tracked by IDC. And capacity-expansion efforts that had constrained earlier pilots moved from announcement into delivery, at least by the characterization carried in The Register's summary. Each of these is the kind of change that, if it had not been visible during a prior evaluation, would have changed that evaluation's output.
The numbers behind the noise are less flattering than the headline implies. According to IDC's FERS Survey from March 2026, reported by The Register, about 19% of enterprises use Claude extensively and another 25% are actively evaluating it, against roughly 42% for OpenAI and 38% for Google. Enterprises, IDC says, "remain largely unsold" on Claude — a phrase that should be read as the analyst firm intends it: not a verdict on capability, but a market-share statement about deployment depth.
The revenue picture sharpens the diagnosis without resolving it. The Information reported, as referenced in The Register's coverage, that enterprise sales made up roughly 86% of Anthropic's projected 2025 revenue, against around 40% for OpenAI. The mix favors Anthropic. The absolute dollars do not: OpenAI's enterprise revenue of about $5.2 billion still exceeded Anthropic's roughly $3.9 billion as of January. A higher share of a smaller pie, in other words, is a defensible enterprise posture rather than a market lead.
What should a buyer actually do? The decision is not whether Claude is good enough. The decision is what to re-test that was either not in scope or not feasible six months ago. The concrete items worth pricing into a fresh evaluation include usage-based billing for the workloads that previously priced out at the seat, the new agent- and tool-use surfaces that arrived across the Jan-to-May release window, the API rate-limit and throughput envelope now that capacity is, by report, less constrained, and the procurement posture Anthropic has signaled in the public enterprise-engagement cadence that IDC highlights.
The honest answer to "is it time to re-evaluate Claude?" is that the prior evaluation is no longer the right document. It is also not the case that any frontier model is ready to be picked up as a sole enterprise provider, by IDC's own judgment. The useful move is a focused re-test, scoped to the specific changes that IDC's report and The Register's coverage actually document, with the prior assumption set explicitly retired rather than recycled.