Claude Code's Weekly Limits Actually Fell 17%, While the Announcement Led With a "Permanent 25% Increase"

From September 14, Anthropic permanently raised the Claude Code weekly limit baseline for Pro, Max, Team and Enterprise by 25% — while simultaneously letting the temporary 50% boost that had run since May, repeatedly extended through September 13, expire. Together, those two changes left users with roughly 17% less capacity than the week before. The arithmetic is simple: with an original baseline of 100, the boost period gave 150, and September 14 set it to 125 — 25% above pre-promotion, about 17% below what had been in hand. The announcement emphasized only the 25% half; developers on X attached a Community Note to the original post, after which Anthropic deleted and reposted it, explicitly acknowledging the 17% reduction and saying it is working on changes that will "make it feel like you're getting more from Claude, while having more visibility and control of your usage." Separately, The Information reported that Anthropic committed roughly $517 billion to compute over the 11 months through the end of August, far above the $180 billion previously disclosed, locking in at least 14.8 gigawatts and bringing total planned capacity to about 16–17 GW. To be clear: these are two separate pieces of reporting, and the causal claim that limits fell because compute is tight has not been established.

Two Real Numbers, One Selective Framing

There are no false figures here. The baseline genuinely rose 25% permanently, and the temporary boost genuinely expired. The problem is that the announcement described only the first half while users experience the sum of both. Run it with a baseline of 100: 150 during the promotion, 125 after September 14. Measured against pre-promotion, that is up 25%. Measured against yesterday, it is down about 17%. And nearly everyone's reference point is yesterday — that 50% boost had been in place since May and was extended four times (through mid-July, then July 19, August 19, and August 31), so after four months of a "temporary" allowance it had long since become the baseline in people's working habits. The resolution is worth recording: once the Community Note went up, Anthropic deleted and reposted, stating the 17% reduction outright in the new version. That is better than holding the original line, but it also confirms the first version was a choice. For anyone who reads release announcements, there is a transferable habit here: **when a company changes a "permanent baseline" and a "temporary boost" in the same move, reading only the number it emphasizes will point you in the wrong direction.**

Put the $517 Billion Beside It, but Don't Wire Them Together

The other report that week came from The Information: over the 11 months through the end of August, Anthropic committed roughly $517 billion to compute, far above the $180 billion previously disclosed. Those agreements, starting October 2025, lock in at least 14.8 gigawatts, which on top of the 1–2 GW it already had brings planned capacity to roughly 16–17 GW. Reported components include about $200 billion over five years with Google and Broadcom, more than $100 billion with AWS over a decade, $50 billion with Fluidstack, $45 billion with Nscale, $45 billion with SpaceX, $35 billion with Lambda and Nvidia, $9.1 billion with Riot Platforms, $1.8 billion with Akamai, and roughly $30 billion with Microsoft Azure for 1 GW. Placing "cut the limits" next to "signed $517 billion" makes it tempting to write "bought all that compute and still rate-limits users." But no party has established that link — the reporting treats them as separate stories, and Anthropic's own framing is that the permanent raise makes part of the summer's capacity expansion stick, language it first used in July alongside a warning that capacity could stay tight. The $517 billion figure needs its provenance labeled too: it comes from non-primary industry reporting, with neither Anthropic nor OpenAI confirming exact numbers on the record; many contracts extend past 2030, so the annualized cost is less staggering than the headline; and it is an upper-bound estimate of potential spending over a decade rather than money already spent. For scale, Bloomberg puts Anthropic's annualized revenue past $65 billion.

Two Things to Do Now If You Work Against a Quota

First, recalibrate your expectations. If your workflow was built around promotion-era allowances — say, letting a strong model sweep an entire repository — it will hit the wall earlier now. Reporting mentions single tasks consuming 80–90% of a weekly limit, and that pattern is the first to break. Second, treat model selection as part of quota management rather than a pure quality decision. This site covered the Elo-per-token paper on September 16, and its conclusion applies directly here: for the same total budget, splitting work across parallel sessions often beats one long run, and the inflection point is measurable. When quotas tighten, measuring that point is cheaper than upgrading the plan. As for cancellations: there are screenshots and switch-to-another-tool posts on social platforms, but those are individual accounts and not an independent measurement of churn. Nobody has published a number, so writing "mass user exodus" would overstate it. Anthropic says changes to visibility and control are coming — that one is worth judging when it ships.

via: BleepingComputer, The Decoder, Investing.com on the $517 billion compute commitments, Analytics India Magazine