Claude Usage Limits Crisis: Anthropic’s Huge Fix Finally Revealed

Dr. Mayank Raj
Anthropic doubled Claude Code limits and removed peak-hour throttling after subscriber complaints over fast-draining Claude usage limits.

Quick Take

  • Claude usage limits drained within minutes in March 2026, triggering global developer complaints.
  • Anthropic admitted three engineering missteps in April, reset all subscriber limits, and apologized.
  • A SpaceX compute deal doubled Claude Code limits and removed peak-hour cuts in May.

Anthropic has resolved the Claude usage limits crisis of March and April 2026 by fixing three engineering bugs, doubling Claude Code limits, and adding new SpaceX compute capacity.

Paying subscribers across the world, including thousands of Indian startup founders and developers, had reported quotas vanishing in as little as 19 minutes instead of the promised 5 hours. The backlash forced Anthropic to publish a rare public post-mortem, reset every subscriber’s quota, and sign one of its largest compute deals, announced on the official Anthropic blog on May 6, 2026.

StartupFeed Insight

The non-obvious lesson here is economic, not technical: flat-fee AI subscriptions break the moment autonomous agents can burn compute 24/7, and Anthropic’s 7% peak-hour throttle proved that a tiny user slice can degrade service for everyone. Indian founders building on Claude Code, and anyone reselling AI workflows to clients, should watch the July 13, 2026 expiry of the bonus 50% weekly limit. StartupFeed expects Anthropic to either extend or make that bump permanent before July 13, because OpenAI’s two-month free Codex offer gives enterprise customers a live exit route, and churn is now measurable. By StartupFeed Desk.

What Happened to Claude Usage Limits?

Claude usage limits are the 5-hour and weekly quotas that cap how much compute each Pro, Max, Team, or Enterprise subscriber can consume. From March 2026, Max plan subscribers paying $200 (Rs 16,700) per month began reporting quota exhaustion in 19 minutes instead of 5 hours, according to widely shared Reddit user logs.

Anthropic acknowledged the issue publicly, saying users were hitting limits “way faster than expected” and calling the investigation its top priority. The timeline below shows how the crisis unfolded and was resolved.

Date Event Impact
March 4, 2026 Default reasoning effort cut from high to medium Quality dropped; users re-ran tasks, burning quota
March 26, 2026 Bug made Claude discard its own reasoning history mid-session Erratic outputs; quotas drained faster, Anthropic engineering blog confirmed
April 16, 2026 System prompt capped replies at 25 words between tool calls Coding quality fell; reverted in 4 days
April 20-24, 2026 All three issues fixed; limits reset for all subscribers Public apology drew 1.3 Mn views on X
May 6, 2026 SpaceX compute deal; Claude Code limits doubled Peak-hour throttle removed for Pro and Max

The most striking fact in this table is that none of the three root causes was a billing error. Anthropic confirmed that no bug was secretly over-charging users for tokens.

About Anthropic

Anthropic is an AI safety company founded in 2021 by former OpenAI executives Dario Amodei and Daniela Amodei, headquartered in San Francisco. It earns revenue from Claude subscriptions ($20-200 per month) and API (Application Programming Interface) usage fees. Its Claude Code agentic tool drove a usage surge through 2026. Key backers include Amazon, Google, and Menlo Ventures.

Why Did Claude Usage Limits Drain So Fast?

Claude usage limits drained quickly because of four overlapping factors: deliberate peak-hour throttling, costly 1M-token context sessions, three engineering bugs, and raw compute shortage. Anthropic engineer Thariq Shihipar said the peak-hour reduction affected roughly 7% of users, per company statements. The March 26 reasoning-history bug was the worst offender because it silently re-processed discarded context, inflating token use without any visible change in user behaviour.

“We’re sorry this has been a bad experience,” Lydia Hallie, Anthropic product lead, wrote on X in April 2026.

The apology thread crossed 1.3 Mn views within hours, per X engagement data. Many users read the advice to switch from Opus to Sonnet as cost-shifting rather than a fix. That perception, more than the bugs themselves, is what turned a technical incident into a trust problem.

How Did Anthropic Fix the Problem?

Anthropic fixed the crisis with three product changes announced on May 6, 2026: doubled 5-hour Claude Code limits on all paid plans, complete removal of peak-hour throttling for Pro and Max, and sharply higher API rate limits for Opus models. The company says Tier 1 API customers received a +1500% increase in input tokens per minute and +900% in output tokens per minute, per the official announcement and its updated rate limits documentation.

The fixes were funded by capacity, not rationing. Anthropic signed a deal with SpaceX for the full Colossus 1 data centre: over 300 megawatts and 220,000+ Nvidia GPUs (graphics processing units), per the company announcement. On May 13, 2026, Anthropic added a bonus +50% to weekly limits for all paid plans, valid until July 13, 2026. Two days later it reset every user’s counters to zero so subscribers could feel the new ceilings immediately.

How Does Claude Code Compare With OpenAI Codex Now?

OpenAI moved aggressively during Anthropic’s crisis, offering eligible enterprise customers 2 free months of Codex if they switched within 30 days, announced on May 13, 2026, the same day as Anthropic’s weekly-limit bump. Google’s Gemini-based coding tools remain the third option, though users who tested alternatives during the outage reported that none matched Claude Code’s capability, per TechCrunch interviews.

Tool May 2026 Move Position
Claude Code (Anthropic) Limits doubled, +50% weekly bonus till July 13 Capability leader, rebuilding trust
Codex (OpenAI) 2 free months for switching enterprises Aggressive challenger on price
Gemini tools (Google) No matching limit offer Distant third in agentic coding

What sets Anthropic apart is that it answered a pricing war with raw infrastructure: a 300-megawatt capacity jump inside one month, alongside earlier deals totalling 10+ gigawatts with Amazon, Google, Microsoft, and Fluidstack.

What’s Next

The single milestone to watch is July 13, 2026, when the bonus +50% weekly limit expires unless Anthropic extends it. If Claude usage limits tighten again that week, expect a fresh wave of churn toward Codex, whose free-trial window for enterprises will just be ending. Will Anthropic lock in the higher ceilings permanently, or was May’s generosity a one-time damage-control move?

Frequently Asked Questions

Why did Claude usage limits run out so fast in 2026?
+

Claude usage limits ran out fast because of peak-hour throttling, costly long-context sessions, and three engineering bugs that Anthropic confirmed in April 2026. The worst bug made Claude discard and re-process its own reasoning history mid-session, silently inflating token consumption even when user workloads had not changed at all.

What does Anthropic do?
+

Anthropic is a San Francisco AI safety company founded in 2021 by Dario and Daniela Amodei. It builds the Claude family of AI models, sells subscriptions and API access, and runs Claude Code, an agentic coding tool whose explosive demand through 2026 strained the company’s compute capacity.

What did Anthropic change to fix Claude usage limits?
+

Anthropic doubled Claude Code’s 5-hour limits, removed peak-hour throttling for Pro and Max plans, and raised API rate limits sharply on May 6, 2026. It also added a bonus 50% to weekly limits until July 13, 2026, backed by a SpaceX compute deal adding over 220,000 Nvidia GPUs.

Will the higher Claude Code limits stay after July 2026?
+

The doubled 5-hour limits and removal of peak-hour throttling are permanent changes, while the extra 50% weekly allowance expires on July 13, 2026 unless extended. Anthropic has not committed beyond that date, so heavy users should avoid building workflows that depend on the temporary weekly ceiling.

Should developers switch from Opus to Sonnet to save usage?
+

Yes, switching to Sonnet for routine tasks roughly doubles how long a session quota lasts, according to Anthropic’s own guidance. Opus consumes limits about twice as fast as Sonnet. Reserving Opus for complex reasoning and using Sonnet or Haiku for everyday coding is the most quota-efficient approach.

Written by Dr. Mayank Raj. Published: June 10, 2026. Updated: June 10, 2026. Have a tip? Write to us at editorial@startupfeed.in.