20260423 - Claude Code rate limits: Anthropic AI squeezes customers

Hello, I'm David Gerrard and this is Pivot to AI, coming to you daily. Today, the big enterprise AI vendor does that thing enterprise vendors do, especially when they're losing a fortune. Anthropic AI, slogan, we're second because we AI doom harder, has a great business. Every bad computer coder and aspiring bad computer coder loves Claude Code, their very favourite pile of vibe-coded trash. Anthropics' revenue is through the roof. Except the minor detail that Anthropics sells Claude Code at a massive loss. It looks like Anthropics spending $8 to $13 for each dollar that comes in, which is a vastly greater loss rate than even I thought they were running up. Anthropic touts annual recurring revenue of $14 billion to $20 billion a year. That's a very fudged number for marketing, how cool they are. Anthropic's actual revenue that they're willing to state in legal filings where they cannot lie is a bit over $5 billion in the company's entire history up to March 9th, 2026. Quote, Although the company has generated substantial revenue since entering the commercial market, exceeding $5 billion to date, unquote, and Anthropix spent at least $10 billion on training and inference. So, it's time to cut costs. Enterprise Software as a Service knows how to handle this one. You make your pricing as obscure and contradictory as you can get away with, and then you put the squeeze on the customers. What are they going to do? Learn to code So what Anthropic did was they quietly removed Claude code from the per month pro plan They made it only for the $100 a month max plan. Anthropic even changed the support documents to match. The Vibe coding world exploded in outrage. How can enterprise SaaS happen to us? Anthropic eventually had to walk it back a bit. This was just a test, see? Anthropic head of growth, Amol Avasari, tweeted. Quote, For clarity, we're running a small test on 2% of new prosumer signups. Existing pro and max subscribers aren't affected. Unquote. So it's just stealth gouging. See if they can get away with it. They didn't get away with it. And they'd put it on their public pricing page for everyone to see. So I think Avasari is declaring a retrospective small test. Anthropik's problem is that people are using their product. Avasari tweets, quote, Long-running async agents are now everyday workflows. The way people actually use a Claude subscription has changed fundamentally. Unquote. See, most businesses, that's a good thing. Avasari's title is literally head of growth. Mate, you're getting growth organically. People love replacing their brain with your clockwork mouse. But Anthropik's losing a packet on every use. $8 to $13 out for every dollar in. making up the difference by setting venture capital cash on fire. Anthropik already been moving to charge Claw enterprise customers per token on top of their monthly fee Anthropik did this quietly but they confirmed it when the information asked them Quote, It made the price change because under the prior system, some customers would hit usage limits that interrupted their work, while others didn't use all the capacity they'd paid for. This change better reflects how customers are actually using Claude as workloads shift from seat-bound productivity into agentic use. Unquote. So it's actually good for you if we charge you more. See? Hope you're comforted. Not that it matters if you aren't. What happens if you use too much clawed code? Well, Anthropic switches off your account and it sends you to fill in a Google form to appeal with that they never seem to look at. Pato Molina from Bello, a finance app in Argentina, has a whole company that runs on Claude Code. We'll skate over the bit where they're vibe coding a finance app. Ew. Molina tweeted how Anthropica'd cut off the company because an automatic system said it had detected abuse and now 60 people could not work. Our automated systems detected a high volume of signals associated with your account which violate our usage policy. Melina appealed and Anthropic emailed back that it was a false positive and switched them back on. So that's nice. It helps when the story is a hit on Twitter. Anthropic isn't the only company setting money on fire and worrying about it. Microsoft's been clamping down on GitHub Copilot and moving individual users to token-based charging. Ed Zitron was leaked internal Microsoft documents about the change and GitHub confirmed some of the leaks in a blog post later that day Microsoft is moving all GitHub Copilot customers to token charging as of June. GitHub Copilot's always been a money burner, but now it's getting a bit much even for Microsoft. All the AI vendors are just setting money on fire. Their biggest problem is that people keep using their services when there was never a path to profit, even marginal profit. What can you, the professional vibe coder, do about this? Well, learn to code? You're going to have to. Thanks for tuning in to Pivot to AI. This evening, I'm off to the closed preview of the new anti-AI documentary, Ghost in the Machine by Valerie Veitch, which I'm in. We're doing a panel too. It should be recorded and up on social media in due course. You can get tickets for the first showing in London. Link in the show notes, and I urge you to get in there. The film's hitting New York next month, and PBS a bit after that. Don't forget to forward this episode of Pivot to AI to the forced workplace vibe code sufferer in your life. Like, subscribe, podcast review, and please do drop your $5 into the Patreon in the show notes. Thank you all. I'll see you tomorrow and bye for now.