Tech Daily - May 11

AI will write 90% of code by the end of 2026, and only 10% of developers will stay relevant.

The engineers in that 10% aren't smarter. They just know how to use AI.

We put together the exact playbook Anthropic engineers use to make sure you're in the top 10%.

Sign up for The Code and get access to:

Join 250K+ engineers

Anthropic just bought every last GPU at one of Elon Musk's biggest data centers, Big Tech's combined capex broke through three-quarters of a trillion dollars, and the line between "AI lab" and "energy infrastructure company" is getting impossible to see. Strange days.

The biggest tech story right now is a deal that almost reads like a typo. Anthropic, the company behind Claude and a fierce public critic of Elon Musk for years, has signed an agreement to use the entire compute capacity of Colossus 1, the massive AI supercomputer in Memphis, Tennessee that sits inside Musk's empire. The deal locks in more than 300 megawatts of capacity and over 220,000 NVIDIA GPUs, including a mix of H100, H200, and GB200 accelerators, all of which will be online for Anthropic within roughly a month.

To put that GPU count in perspective, 220,000 of NVIDIA's top accelerators is more raw compute than most national governments can deploy. The Memphis facility, originally built and branded under xAI's umbrella before being folded under SpaceX during a recent corporate reshuffle, is now a single-tenant Claude factory. The strategic message is louder than the financial one: in 2026, the AI race is no longer mostly about who has the smartest researchers or the best architecture. It is about who can lock in power and silicon faster than the other side.

Bloomberg's full reporting on the agreement: https://www.bloomberg.com/news/articles/2026-05-06/anthropic-inks-computing-deal-with-spacex-to-meet-ai-demand

CoinDesk's breakdown of how the deal positions SpaceX ahead of its planned IPO: https://www.coindesk.com/tech/2026/05/06/anthropic-signs-elon-musk-s-spacex-for-colossus-1-compute-ahead-of-june-ipo

The optics are wild. Musk has spent the better part of two years publicly trashing Anthropic, suing OpenAI, and positioning xAI's Grok as the unfiltered alternative to what he calls overly cautious corporate AI. Anthropic, for its part, has built much of its public identity around safety, alignment, and exactly the kind of guardrails Musk regularly mocks. According to multiple reports, Musk has now publicly praised Anthropic's team after meetings tied to the agreement, a tonal shift that landed somewhere between surprising and surreal.

There is a real reason both sides set the rivalry aside. Anthropic has been struggling to keep up with demand. The company's revenue grew roughly eighty times in the first quarter of 2026, and that growth pulverized its capacity headroom. Claude Pro and Claude Max users have been hitting reliability issues for weeks, and the company has been burning through every gigawatt it can lease. Meanwhile, industry chatter throughout the past several months has suggested that some of the AI capacity tied to Musk's companies, particularly the newer Colossus build-outs, was sitting underutilized while xAI scaled up more slowly than originally promised. The simplest explanation for the deal is the most accurate one: Anthropic needed compute, SpaceX had it, and both sides decided money beats grudges.

There is also a longer-term subplot that is worth flagging. Anthropic has now publicly said it has expressed interest in partnering with SpaceX on multiple gigawatts of orbital AI compute capacity. In other words, Anthropic is at least exploring the idea of putting AI training and inference infrastructure literally in space, where solar power is constant, cooling is free, and the only physical limit is launch cost. Whether that ever happens or stays a research curiosity, the fact that a top-tier AI lab is even floating it tells you how thoroughly the energy bottleneck on Earth is reshaping long-term planning.

TechRepublic on the orbital data center angle: https://www.techrepublic.com/article/news-anthropic-spacex-claude-compute-colossus-1/

StorageReview's detailed look at the Colossus 1 specifications: https://www.storagereview.com/news/anthropic-signs-spacex-colossus-1-deal-for-major-claude-compute-expansion

The deal is not just an abstract infrastructure announcement. It is already producing immediate, concrete improvements for paying Claude customers. Anthropic has doubled the five-hour rate limits on Claude Code for Pro, Max, Team, and Enterprise plans. It has removed peak-hour limit reductions on Claude Code for Pro and Max accounts, which was a significant pain point for developers running long agentic workflows during U.S. business hours. It has also raised API rate limits significantly on the Claude Opus models, with input tokens per minute climbing for both individual developers and enterprise customers.

The free plan, notably, is excluded from the new limits. That is consistent with how Anthropic has been positioning itself for the past year, treating paid tiers as the priority service while the free tier acts as a top-of-funnel and capacity-flexible offering. If you are running Claude Code in production, building agentic systems on the Claude API, or just hitting Max plan ceilings every afternoon, you should already be seeing measurable improvements in throughput and reliability.

Business Standard's detailed breakdown of the new usage limits: https://www.business-standard.com/technology/tech-news/anthropic-claude-usage-rate-limit-increased-paid-users-spacex-compute-deal-new-limits-126050700994_1.html

Pulse 2.0 on what changed in the Claude Code rate caps: https://pulse2.com/anthropic-higher-claude-usage-limits-introduced-following-spacex-compute-partnership/

The SpaceX agreement is only the latest in a list of compute partnerships that, taken together, paint a striking picture of how Anthropic is operating. The company has a roughly 5 gigawatt agreement with Amazon, with nearly 1 gigawatt of new capacity expected to come online by the end of 2026. It has a separate 5 gigawatt deal with Google and Broadcom for next-generation TPU capacity, with first deliveries scheduled for 2027. It has a strategic partnership with Microsoft and NVIDIA covering roughly 30 billion dollars of Azure capacity. It has a 50 billion dollar U.S. AI infrastructure investment with Fluidstack. And now, on top of all of that, a single-tenant takeover of SpaceX's Colossus 1.

Stack those numbers together and Anthropic is effectively the largest single buyer of frontier AI compute in the world. That has implications that go far beyond Claude usage limits. It means Anthropic is now structurally important to the financial models of Amazon Web Services, Google Cloud, Microsoft Azure, NVIDIA, Broadcom, and SpaceX. It means the company has hedged its compute supply across every major hyperscaler, which makes it nearly impossible for any single provider to use leverage against it. And it means that when people ask why frontier AI companies need to raise tens of billions of dollars at a time, the honest answer is that those rounds are not paying for engineers. They are paying for power.

The broader industry context makes this even more striking. Meta, Amazon, Microsoft, and Alphabet have collectively signaled roughly 725 billion dollars in 2026 capital expenditures, almost entirely earmarked for data centers, custom chips, GPUs, and AI models. That is more than a 75 percent year-over-year jump. At the same time, those companies have collectively cut or offered buyouts to well over 150,000 employees in recent months. The trade is explicit: payroll out, compute in.

Data Centre Magazine's analysis of the compute arms race: https://datacentremagazine.com/news/how-spacexai-anthropic-compute-deal-raises-claude-limits

EdTech Innovation Hub on what this means for AI accessibility: https://www.edtechinnovationhub.com/news/anthropic-turns-to-spacex-compute-deal-as-claude-usage-limits-rise

Three threads to keep an eye on over the next few weeks. First, SpaceX's IPO. The company is reportedly targeting a June listing, and Anthropic's takeover of Colossus 1 hands the IPO road show a perfect anchor story: SpaceX is not just a rocket company, it is now a credible AI infrastructure provider with a marquee single-tenant customer. Expect that narrative to be front and center in the prospectus.

Second, the regulatory response. The Pentagon recently designated Anthropic a supply chain risk and gave defense contractors six months to stop using its products, which has created an unusual political dynamic where the U.S. military is uncomfortable with Anthropic's safety posture while major hyperscalers and now SpaceX are pouring billions into Anthropic's infrastructure. Watch for whether the Anthropic-SpaceX deal eases or worsens that friction, especially as Musk's relationship with the current administration shifts.

Third, the orbital compute idea. If Anthropic and SpaceX move past public statements and actually begin technical work on space-based data centers, that will reshape how everyone in the industry thinks about long-term AI infrastructure. Right now it sounds like science fiction. So did 220,000 GPUs in a single Memphis facility, three years ago.

We will keep tracking this and bring you the next chapter as it lands. Stay sharp out there.