Claude's Growing Pains: Enterprise Customers Voice Concerns as Anthropic's IPO Nears

Anthropic is approaching what may be the most consequential stretch in its history. Its valuation has crossed $900 billion in secondary markets, a potential October IPO is drawing closer, and the company is managing the most complex product lineup it has ever had. Some of its most important enterprise customers, according to a Bloomberg opinion piece published Tuesday, are watching that complexity with a degree of unease.

The concerns span nearly every part of Anthropic's business: perceived declines in Claude performance, pricing changes that caught developers off guard, capacity constraints, and the controlled rollout of Claude Mythos, the security-focused model the company has described as too dangerous to release publicly. None of these issues has slowed Anthropic's revenue growth. But they have arrived at the same time, which is a different kind of problem.

Where the Complaints Are Coming From

The first wave of concern emerged from developers who noticed what they described as declining output quality from Claude Opus 4.6 after a silent update. Anthropic has not confirmed a deliberate capability change, but the perception was enough to trigger a broader examination of its product reliability record. When the company then removed Claude Code access from the $20-per-month Pro plan, a second wave followed. For developers who had structured their workflows around that combination, the change was not a minor inconvenience. It altered the cost structure of their work overnight.

Around the same time, enterprise clients in financial services and healthcare began raising questions about capacity commitments. Anthropic's rapid growth has strained its infrastructure in ways that periodically show up as elevated error rates or slower response times during peak hours. For companies running mission-critical applications on Claude, that creates reliability risk that their security and operations teams have to account for.

The Mythos Problem

The most pointed source of tension is Mythos itself. Anthropic announced in April that the model had demonstrated nation-state-level cybersecurity capabilities, specifically the ability to autonomously discover and exploit software vulnerabilities at a scale and speed no human team could match. The company chose not to release it publicly, opting instead to run Project Glasswing, a coalition of operating system vendors, browser makers, and security firms who are using Mythos to patch critical infrastructure before the model's capabilities are more broadly available.

For enterprise customers who want a predictable product roadmap, the Mythos situation presents a genuine dilemma. The model is real, its capabilities are apparently as described, and it is being used, but not by them. Anthropic's stated rationale is that a public release would give attackers the same access as defenders, compressing the window between vulnerability discovery and exploitation. That reasoning is coherent. It is also, from the perspective of a CIO trying to plan next year's AI budget, not a product schedule.

"Our goal is to deploy Mythos-class models safely at scale, but first we need safeguards that reliably block their most dangerous outputs." Anthropic statement, May 2026

Competitive Noise Around the Rollout

OpenAI CEO Sam Altman called the Mythos rollout "fear-based marketing" in a podcast appearance, suggesting that Anthropic's "too dangerous to release" framing conveniently serves as a competitive positioning tool ahead of its IPO. The accusation has a surface logic: announcing a model so powerful it cannot be deployed publicly, right before an S-1 filing, does create a useful narrative about technological moat. Whether that motivation is real or not, the effect is that Mythos has become entangled in both the product concerns and the investor story simultaneously.

Anthropic has pushed back on the framing, pointing to the independent assessment from the UK AI Safety Institute, which reviewed Mythos and confirmed its cyber capabilities, as evidence that the safety concerns are legitimate rather than manufactured. The UKAIS review does not resolve the scheduling question for enterprise clients, but it does complicate the "fear-based marketing" narrative.

What the IPO Window Means for Customers

The timing of all this matters because Anthropic's ability to command a premium valuation in an IPO depends partly on enterprise customer trust. Investors will look past headline revenue numbers and examine renewal rates, contract expansions, and whether the major accounts that drove Anthropic's 80x growth in Q1 2026 are deepening their commitments or starting to hedge with alternatives. A company closing a $30 billion funding round at a $900 billion valuation is pricing in continued dominance. Product quality concerns and pricing surprises are exactly the kind of signal that can complicate that story.

Anthropic's response to the Bloomberg analysis, to the extent it has made one, is visible in its recent enterprise announcements. The company has continued expanding its partner network, adding KPMG, Fujitsu, NEC, and several other major enterprise commitments in the last few weeks. Revenue growth remains steep. But growth can mask customer friction for a while. The IPO roadshow is where the questions will be asked directly, and where the answers will need to hold up under scrutiny from institutional investors who have seen a lot of software companies grow fast and then stumble.

For now, the picture is not a crisis. It is a set of compounding pressures on a company that is simultaneously navigating its most ambitious commercial expansion, its most complex safety challenge, and a public markets debut. Whether Anthropic handles them cleanly over the next five months will say something important about whether it is ready for what comes after the IPO.