Claude Code Limits Doubled: Anthropic's SpaceX Colossus 1 Deal
Prefer to watch? Watch on YouTube · Subscribe for new builds · Read the transcript
In this video I go through Anthropic's announcement that it's doubling Claude Code usage limits, and the SpaceX compute deal behind it. It's for anyone who uses Claude Code, OpenClaw, or any AI agent workflow every day and wants to know what changed and why. I recorded this in May 2026. Limits and plans change fast, so check Anthropic's current docs before you plan around any of these numbers.
What I cover in this video
- 0:00: The headline. Anthropic signed a deal to use all of SpaceX's Colossus 1 compute capacity.
- 1:31: What the announcement changes for Claude Code users on Pro, Max, and Team.
- 3:01: Why hitting a 429 limit error halfway through a refactor hurts so much.
- 3:46: Why the whole industry is running into a compute wall.
- 5:17: The other compute deals Anthropic has lined up (Amazon, Google, Microsoft, FluidStack).
- 6:05: The odd part: SpaceX owns xAI, one of Anthropic's public rivals.
- 6:50: The real-world cost of Colossus 1 in Memphis, Tennessee.
- 10:38: The orbital compute line in the announcement, and what builders should take from all this.
What changed for Claude Code users
Anthropic's announcement was titled "Higher usage limits for Claude and compute deal with SpaceX." The part users care about right away:
- Claude Code 5-hour rate limits are doubling for Pro, Max, and Team accounts.
- The peak-hours limit reduction is gone on Claude Code for Pro and Max accounts.
- API rate limits for Claude Opus models are going up.
Before this, you could get noticeably further on an OpenAI subscription than on a Claude subscription. If you build with Claude Code, you know the pain. You're in the middle of a bug fix or a refactor, the agent finally understands your codebase and file structure, and then you get a 429 limit error. Coming back later breaks your flow, and in agent workflows where one tool hands off to another, it can break the whole chain. So doubling the limits matters.
Why Anthropic needed a SpaceX data center
The deal covers all of the compute at Colossus 1, not just part of it. Anthropic says that means more than 300 MW of new capacity and over 200,000 Nvidia GPUs coming online within about a month.
The reason is that the whole industry is hitting the same wall: compute. Once a model is good enough, the bottleneck moves. Can you serve it to millions of people? Can developers run long coding sessions without getting cut off? Can agents run for hours instead of one prompt at a time? All of that takes a huge amount of power and hardware.
SpaceX isn't Anthropic's only deal. The announcement also mentions:
- An agreement with Amazon for up to 5 GW, with nearly 1 GW coming online by the end of this year
- A 5 GW agreement with Google and Broadcom that starts coming online at the beginning of next year
- A Microsoft and Nvidia partnership that includes $30 billion of Azure capacity
- A $50 billion American AI infrastructure investment with FluidStack
Reuters also reported that The Information says Anthropic committed to spend $200 billion with Google Cloud over five years. Reuters said it couldn't independently verify that, so treat the number with care. Even without it, the pattern is clear: Anthropic is locking down compute everywhere it can, across AWS Trainium, Google TPUs, Nvidia GPUs, Azure, FluidStack, and now SpaceX.
The rival angle and the real-world cost
SpaceX acquired xAI earlier this year, so Anthropic is getting capacity from the ecosystem of one of its public rivals. CNBC reported that Elon Musk posted in February that Anthropic "hates Western civilization," and he has attacked its safety posture. A deal like this only happens when the need for compute outweighs the rivalry.
Colossus 1 is in Memphis, Tennessee. CNBC reported that xAI and a subsidiary ran dozens of natural gas turbines to power the facility, claiming no federal permit was needed because the turbines were temporary. There have been protests over pollution, and the NAACP sued xAI over Clean Air Act allegations. "The cloud" is really someone else's building pulling electricity from somewhere. That's concrete, steel, gas turbines, power grids, water, land, and the local communities around them.
What this means for agent builders
This matters a lot for OpenClaw and AI agent builders. A real agent workflow isn't one prompt. It might read files, search memory, call tools, inspect code, spawn a coding run, review the output, summarize the changes, update tasks, and then message you. That's a chain of requests, and agent workflows use much more compute than a normal chat. Rate limits aren't there to annoy people. They come from that load, and we're still early, with only a small share of businesses using AI so far.
The wildest line in the announcement is about orbital compute. Anthropic expressed interest in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity. That doesn't mean there's a signed contract to launch data centers into space. But the fact that the sentence is in the announcement at all shows how far ahead these companies are thinking about power, land, and cooling.
Key takeaways
- In the short term this is good news: higher Claude Code limits mean less waiting, fewer interrupted coding sessions, and a better chance your agent workflows actually finish.
- The AI race is turning into an infrastructure race. The winners will need the best model, distribution, compute, and energy strategy.
- Compute has real physical costs, and Colossus 1 shows that clearly.
- If you build with agents, watch these compute deals. Your workflow's reliability depends on the infrastructure underneath it.
- Claude didn't just get more generous. Anthropic bought more runway.
FAQ
Did Anthropic double Claude Code usage limits?
Yes. In May 2026 Anthropic announced it was doubling Claude Code's 5-hour rate limits for Pro, Max, and Team accounts. It also removed the peak-hours limit reduction for Pro and Max and raised API rate limits for Claude Opus models.
What is the Anthropic and SpaceX Colossus 1 deal?
Anthropic signed a deal to use all of the compute capacity at SpaceX's Colossus 1 in Memphis, Tennessee. Anthropic says that gives it more than 300 MW of new capacity and over 200,000 Nvidia GPUs coming online within about a month.
Is Anthropic putting data centers in space?
Not yet. In the announcement, Anthropic expressed interest in partnering with SpaceX on multiple gigawatts of orbital AI compute capacity, but that isn't a signed contract to launch data centers.
Why do AI agents hit rate limits so often?
An agent workflow is a chain of requests, not a single prompt. It reads files, calls tools, inspects code, and reviews output, so it uses far more compute than a normal chat.
Transcript
Read the full transcript
0:00Anthropic just doubled Claude code limits, and there is some major news that happened this morning. Anthropic has basically teamed up with SpaceX to basically 10x their computing power. So, if you haven't noticed recently, you can get a lot further on an OpenAI subscription than you could on a Claude subscription, but now everything has changed. And in this video, I'm going to break down why this matters, especially if you are Claude code user. They didn't just optimize the model, they didn't just tweak a pricing page. They literally signed a deal with SpaceX to use all of the compute capacity at Colossus 1, not just a portion of it, but the entire amount. And according to Anthropic, that gives them more than 300
0:45MW of new capacity. So, over 223,000 Nvidia GPUs coming online just within the next month. That is the part that people are going to scroll past, but it is the whole part of the story. It's not really just a Claude update, it's an entire infrastructure change. So, this is a signal that the AI race has officially moved from just simple model announcements, somehow has been for the last year, to actually raw physical infrastructure, power, land, GPUs, and cooling. Basically, data centers, right? Apparently, maybe one day, compute in orbit, which SpaceX is already kind of dialing down. You know, data centers in space, we're already seeing it. So, I'm going to break down what this actually
1:31means and why it happened, why it matters, and what this actually means if you use Claude code every day, or even open Claude, or really any AI agent workflow. Anthropic posted an announcement called higher usage limits for Claude and compute deal with SpaceX. That was just this morning. The first part is what users care about right away. Claude code 5-hour rate limits are doubling for pro, max, and team accounts. So, So are removing the peak hours limit reduction on Claude code for pro and max accounts, which is great news. And they're raising the API rate limits for Claude Opus models. So if you use Claude code heavily, this is a pretty big deal because anyone who builds with Claude code knows the pain over this last month. Real quick, before we keep going, if you're watching this
2:16and you want to actually build with some of these tools, not just watch the videos about them, then you're going to want to check out our community down below, Shipping School. We have a full Claude code course, a full open Claude course, and four live boot camps every single week where we actually help you get set up from scratch. Like actually set this thing up, not just watch a tutorial and figure it out by yourself. And we also provide one-on-one coaching so you could book a call with me, we could share screens, and I can help you get Claude code or open Claude running on your machine. That's it, no fluff. I built this community because watching YouTube's only gets you so far. We launched it just 3 days ago and we have over 55 members. You need people around you who are actually building, people
3:01who hold you accountable, and coaches who can help you when you get stuck. I'll put the link in the description down below. Get in now before the price goes up. You're in the middle of a bug fix or refactoring some part of your code base and you finally got the agent to understand your code base. It even has the context, it knows the file structure, and halfway through refactoring something, and then you hit your usage limits. 429 API limit error. You come back later and that is not just annoying, but that breaks the whole flow, especially when you're in the flow state. So, if you're using Claude code inside of an agent workflow where one tool hands off to another, that's going to happen. So yes, doubling those limits really does matter, but the biggest question is is why did Anthropic need a
3:46SpaceX data center to do this? Because the entire industry is running into the same wall. The wall is computing power. Everybody talks like the AI race is only about who has the smartest model. It's not. Claude versus GPT versus Gemini versus Grok, right? But once the model gets good enough, the bottleneck changes. Can you actually serve it to millions of people? Can you let developers run long coding sessions without cutting them off? Can you actually support agents that run for hours instead of one prompt at a time? Can you let businesses build workflows where AI is not a chatbot, but an always-on employee. That takes an insane amount of computing power. And Anthropic is basically saying this out loud now. They are not hiding behind vague
4:31wording. They said this SpaceX deal gives them access to more than 300 MW of new capacity. Like I said, over 200,000 Nvidia GPUs within 1 month from now. And that's pretty crazy. So, to put it into this picture and to paint some context for you, this is not only the compute deal Anthropic has that they stacked up, they also referenced an up to a 5 GW agreement with Amazon. Nearly 1 GW coming online by the end of this year. They referenced a 5 GW agreement with Google and Broadcom that starts coming online at the beginning of next year. They referenced a Microsoft and Nvidia partnership that includes 30 billion dollars of Azure capacity. They
5:17referenced a 50 billion dollar American AI infrastructure investment with FluidStack. And Reuters reported that the information says Anthropic committed to spend 200 billion dollars with Google Cloud over the next 5 years. Reuters said that they could not independently verify that report, so be careful with that number. But even if you ignore the 200 billion dollar number, the direction is pretty obvious. Anthropic is trying to lock down compute everywhere. A to AWS Trainium, Google TPUs, Nvidia GPUs, Azure fluid stack, and now SpaceX. That tells you the demand is not normal. Demand is overwhelming the entire infrastructure. And this is where the story gets pretty weird right here is because this SpaceX deal is not just any data center deal. SpaceX acquired xAI
6:05earlier this year. So, Anthropic is effectively getting capacity from the ecosystem of one of its its public rivals. Elon Musk has been openly critical of Anthropic. CNBC reported that he posted in February that Anthropic hates Western civilization. He has called them hypocritical. He has attacked their safety posture. And now Anthropic is using SpaceX compute to improve Claude. That is the kind of thing that only happens when the need is bigger than the rivalry. When you need 200,000 GPUs and 300 megawatts fast, you just do the deal. This is business, right? This is infrastructure. It's no longer a Twitter drama weird kind of situation. And then there is this Colossus One part. Colossus One is in Memphis, Tennessee. CNBC reported that
6:50xAI and its subsidiary MZX Tech operated dozens of natural gas-burning turbines to power the facility, claiming that no federal permit was needed because they were temporary. There have been protests around pollution concerns. Obviously, there always has been. The NAACP sued xAI over Clean Air Act allegations. So, there is a real-world cost to this compute race. People hear AI data center and think it's just some clean abstract cloud in the sky, and it's not. It is concrete. It is steel. It is gas turbines. It's power grids, real local communities, things like water, cooling, and land. I mean, the cloud is just someone else's building pulling electricity from somewhere. And that is why the story matters for more than just Claude users. This is the whole AI
7:36industry showing its actual bottleneck in real time. We spent the last few years asking which model's smarter, right? The next few years are going to be which company can actually run these models at scale. And that is the difference between all of them from Google, Anthropic, OpenAI, and now xAI. The model demo can now go viral on X, right? We've always seen it. But a real business that is an AI business needs capacity every single day, and we're not talking enough about it. It needs uptime, it needs predictable limits, it needs enough compute that developers do not get kicked out halfway through their work. It needs enterprise customers to trust that system will actually be available when the work matters. That is why Claude code limit part is so important. This is not just a consumer feature. This is Anthropic trying to
8:22make Claude code usable as infrastructure. If you're struggling to keep up with content, what I'm about to save you about 40 days worth of work. I built something [music] called Content Machine. It's 10 AI agents that run on the Open Claw orchestration, and they handle everything. Scripts, thumbnails, X posts, [music] blogs, outreach, clips, news leaders, all of it. So, I went from 1,000 subscribers to 4,000 subscribers on YouTube in 7 days using this exact system. Every single morning I wake up and the content's already done. I spend maybe 15, 20 minutes reviewing and approving them, and I move on with my day. [music] It works for any niche, fitness, finance, real estate, marketing, whatever you are building,
9:07and it is 100% completely customizable to your use case. [music] So, you get the mission control dashboard, all of the cron jobs, everything I've built over the last 40 days helping me gain more and more people to subscribe and join the community. So, you plug in your own thing and it molds it to you, it learns how you talk and it writes so it doesn't sound like AI slop. $97 one-time, it's not a subscription, [music] I'll put the link down below and you'll thank me later. If Claude code is going to become the coding layer for companies, it cannot feel like a toy that just randomly taps out in the middle of a sprint. And I know you experienced this in the past. It has to actually feel dependable. And dependability costs GPUs. It also matters a lot for open claw and agent
9:53builders in general. Because agents are not cheap little prompts. A real agent workflow might read files, search memory, call tools, and inspect code, even spawn a coding run and review the output. It may summarize the changes you made and update tasks. And then it messages you. This is not one request. It's a chain of commands. And if everyone starts running agent chains, model demand is going to go crazy. Remember, we're still in that one to two percent phase where people that are actually using AI in their businesses. We are still so early. This is why rate limits have been such a big pain point even with us using it, builders, every day. I mean, imagine when the whole world catches on. It is not just because Anthropic wants to annoy people, either. It is because agentic workflows use way
10:38more compute than a normal chat. So, when Anthropic doubles Claude code limits, that is a sign that they are preparing for a world where people are not asking Claude just one question. They're giving Claude actual work. And there's a huge difference between the two. The craziest line in the announcement might be the orbital compute part. Anthropic already said as part of the agreement, they expressed interest in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity. Now, be careful here. That does not mean they signed a contract to launch data centers into space tomorrow. But, the wording is expressed interest. And that's crazy. The sentence even being in the Anthropic announcement is pretty cool. And it is the future that we are looking into
11:23right now. It tells you how far ahead these companies are actually thinking. If terrestrial power, land, and cooling become the bottleneck, the next frontier becomes anything that expands that capacity. Different countries, new power deals, custom chips, and even dedicated data centers. Apparently, one day we may have that in orbit. That sounds ridiculous until you remember that literally 10 years ago renting 200,000 GPUs from a rocket company would have sounded like a crazy person. So, what is the takeaway of all this? For builders specifically, this is great news short-term. Claude code limits doubling means more room for you to work, to build out autonomous agents, to build out simple workflows that run on 24/7. There's less waiting. There's fewer interrupted coding sessions, better chance that your agent workflows can
12:08actually finish. But long-term, this tells us that there's something bigger around the corner. AI is not just software anymore, it's actual infrastructure. The winners are not just going to have the best model anymore. They're going to have the best model, the best distribution, and the most compute. Even the best energy strategy and the ability to turn that into actual reliable products. That is why OpenAI is locking in down compute, right? And that is why Anthropic is locking down compute. You see it. Also Google, Amazon, Microsoft, Nvidia, Broadcom, Coreweave, Fluidstack, SpaceX, and xAI all keep showing up in the same story. The AI race is becoming an infrastructure race. And today, Anthropic basically admitted that Claude code needed a rocket company supercomputer to raise its actual limits. That is the headline, not
12:54because it's funny, I mean it is, but not really. It's because it shows where everything is going. The next wave of AI will not be limited by whether the model can answer your prompt or not. It will be limited by whether the company behind it has enough power to let that model work for everyone at once. And if you are building with AI agents right now, this is why you should be paying attention to these compute deals. Because the quality of your workflow depends on the invisible infrastructure under it. When the compute gets better, your agent gets more reliable. When the limits go up, your workflows can run longer. And when the infrastructure gets serious, the products built on top of it stops feeling like demos. That is what this SpaceX deal is really about. Claude did not just get more generous, Anthropic bought more runway. And for
13:39anyone building with Claude code, Open Claude, or AI agents, that runway really does matter. And if you need help building out AI agents using Open Claude, Hermes, Claude code, Codex, come join us in the community. We have 200 builders building in there right now. I'll put the link down below. We have unlimited one-on-one calls where you could book a million calls if you want to help share your screen and get these setups and automations working. We have nine boot camp calls a week from local LLMs to project sharing to landing page redesign. And we also have full courses on Claude code and Open Claude in that community. I'll put the link to it down below. Check it out. And if you haven't subscribed already, please do cuz we release videos just like this every single day. See you in the next one. Be blessed.
Ready to start building with AI?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week and weekly-updated courses.
Join Shipping SkoolKeep reading
Claude Code Just Doubled Your Limits. Here Is How I Am Using Every Last One.
Anthropic just doubled Claude Code usage limits during off-peak hours. This is how I restructured my entire AI agent pipeline to take advantage before it expires March 27th.
Claude Fable 5.2 Explained: Pricing, Benchmarks, and Effort Levels
Anthropic's Claude Fable 5.2 costs $8 in and $40 out per million tokens, down from $10 and $50. At High effort it roughly matches Opus 5.5 at Max on FrontierCode for about half the cost per task. I'm keeping Opus 5.5 as my daily driver and using Fable for big, hard projects.
Anthropic S-1 Explained: $16.3B Revenue and the $80.7B Loss
Anthropic's S-1 shows $16.3B in revenue and an $80.7B net loss for the first half of 2026. About $72.4B of that loss is a paper charge on Amazon's convertible notes. The operating loss was only $693M, and cash from operations was positive. For Claude Code users, the main things to watch are rate limits and pricing.
