Own vs Rent Proof Point
June 6, 2026

Two Cliffs, One Landlord: Why Anthropic Has Two Pricing Events in 35 Days

If you're paying for Claude, you've probably heard about June 15. That's when Anthropic splits your subscription into two credit pools — one for chat, one for agents — and the agent pool has a hard cap.

But there's a second cliff that almost nobody is talking about. And the 35-day window between them tells you everything about where AI pricing is headed.

Cliff #1: June 15 — The Credit Split

Your Claude subscription splits in two. Agent SDK usage (automations, CI pipelines, third-party tools like OpenClaw, Conductor, Zed) gets its own credit bucket with a hard cap. Run out? Your agents stop. Credits don't roll over. Same monthly price, less access.

Cliff #2: July 13 — The Limit Expiry

The temporary 50% weekly usage limit increase expires. When it goes, your weekly cap drops back to the lower baseline. For heavy users, this means hitting your ceiling faster every week — paying the same for less.

The 35-Day Squeeze

Two pricing changes. 35 days apart. Neither adds features. Neither reduces your bill. Both take something away.

June 15: One subscription becomes two capped pools. Agent usage gets a ceiling.
~June 20-25: First wave of users hit their Agent SDK caps. Agents stop mid-task.
July 13: Weekly limits drop 50%. Even interactive usage squeezes.
July 15+: Both cliffs active simultaneously. Monthly bill unchanged. Access significantly reduced.

Why Two Cliffs Instead of One

If you're going to restructure pricing, why not do it all at once? Because two cliffs do something one cliff can't: they make each individual change feel smaller.

June 15 is about structure — redefining what you get. July 13 is about volume — reducing how much you can use. By separating them, each change has its own news cycle, its own Reddit thread, its own moment of outrage that fades before the next one hits.

This isn't new. It's the same playbook Uber used: subsidize aggressively, then raise prices incrementally until each increase feels normal. Business Insider just reported on a startup CEO who described individual AI plans as a "loss-leader" subsidy — $3,000/month for his team on individual plans versus $30,000-$40,000 on enterprise. He's hoping the subsidies don't end. They will.

Who This Actually Hits

User TypeJune 15 ImpactJuly 13 Impact
Casual chat usersMinimalWeekly cap drops 50%
Claude Code developersAgent pool caps automationsFaster weekly ceiling
Agent SDK / CI pipelinesHard cap, agents stopReduced weekly throughput
Third-party tool usersNamed in restrictionsDouble squeeze on usage
Startup teams gaming individual plansSubsidy structure changes10x cliff to enterprise
The startup CEO's math: Pratyush Rai, CEO of Foyer, told Business Insider his 25-person team pays $3,000/month on individual Claude and OpenAI plans. Enterprise API equivalent: $30,000-$40,000/month. That's a 10-13x cliff. His CTO called the individual plan a "blessing." His hope: token prices collapse before subsidies end.

The Pattern: Five Billing Changes in Five Months

April 2026: Anthropic bans third-party agents from Claude subscriptions
May 2026: Tests removing Claude Code from Pro tier
June 15: Restructures pricing — same cost, split pools, less access
July 13: Temporary 50% limit increase expires
Q3-Q4 2026: IPO pricing pressure predicts more interventions

Five months. Five changes. Every single one takes something away from subscribers. Not one adds a feature or reduces a price.

What Owning Your AI Looks Like

When you own your agent — the prompts, the workflows, the decision trees — no company can:

The Twin Agent Kit is $47 once.

Not $20/month that becomes $20/month-for-less.
Not a credit that expires. Not a permission slip that can be revoked.
Not a subsidy that disappears when the IPO needs margins.

Own your agent. →