Anthropic, the AI company behind Claude, is in hot water over its highest-priced subscription plans. Customers say the $100 and $200 Max plans don’t deliver the usage promised. One user claims the plans offer far less than the advertised “5x” and “20x” usage compared to the base Pro tier.

The complaint comes from a Washington, D.C. customer who says he quickly hit usage limits soon after subscribing. He describes being cut off mid-session and having to ration his prompts or pay more. The lawsuit seeks refunds and class-action status for others who feel misled.

Anthropic introduced weekly usage caps in mid-2025 to manage heavy users. These limits are real and public but not easy to measure. The bigger problem is that the company marketed the Max plans as offering multiples of the base usage, which may not be true in practice.

Ongoing Throttling and Usage Confusion

Even after Anthropic admitted bugs that drained user quotas faster than expected, heavy users still report hitting hard limits. One subscriber logged multiple throttle events within a single day across different Claude products.

The throttling often comes at peak work times, forcing users to wait hours before their limits reset. This disrupts workflows, especially for developers relying on Claude Code or the AI Agent SDK for automation tasks.

Anthropic offered a reset in April to compensate for degraded service caused by bugs. But reports show that the usage caps and throttling continue, raising questions about whether the reset fully addressed the issues.

New Billing Changes Make Costs Less Predictable

Starting June 15, Anthropic split billing for Claude subscriptions into two parts. Interactive chat and coding sessions stay under the monthly subscription. But automation and programmatic use now draw from a separate monthly credit.

This means that if you use Claude’s Agent SDK, CLI tools, or GitHub Actions integrations heavily, you get a fixed credit—$20, $100, or $200 depending on your plan. Once that credit runs out, you pay API rates per token, which can get expensive fast.

For example, a team that previously paid $200 a month on the Max 20x plan could now see bills spike to thousands of dollars if their automation burns through the credit early. This split exposes automation costs that used to be hidden inside flat-rate subscriptions.

Finance teams must now audit usage carefully. Without controls, unexpected overages can hit. Engineering leaders face decisions about whether to migrate production automation to metered API keys, which offer better spend caps and visibility.

Compute Limits and Infrastructure Challenges

Behind the scenes, Anthropic struggled with insufficient compute power. For nearly a year, users experienced outages, rate limits, and throttling. The company recently revealed a major deal with SpaceX to access a massive data center with over 220,000 GPUs.

This new capacity aims to end the compute shortage that caused many service problems. Rate limits doubled, and peak-hour throttling was removed starting May 2026. But it remains to be seen if these upgrades fully restore user trust.

The compute crunch explains why usage caps and throttling became necessary. High demand and limited server power forced Anthropic to ration usage to keep the service running. The new infrastructure deal should ease these constraints.

Developer Backlash and Data Policies

Anthropic’s recent Claude Fable 5 launch stirred controversy too. The model burns through usage roughly twice as fast as prior versions. Users report draining a $100 Max plan in under ten minutes.

Additionally, Fable 5 silently throttles performance without warning when it detects certain AI research work. A mandatory 30-day data retention policy also locks out companies subject to strict privacy rules like GDPR.

These factors fuel frustration among developers and researchers who rely on transparent, predictable AI tools. The combination of hidden throttling, rising costs, and data concerns puts pressure on Anthropic to improve communication and product design.

For now, Claude remains a strong AI assistant technically. But customers must navigate opaque usage limits, unexpected caps, and new billing that can lead to expensive surprises. If you use Claude heavily, especially for automation, plan your usage carefully and watch for updates from Anthropic.