Welcome to Runtime! Today: Faced with limited options, Anthropic kicks off a perilous partnership with Elon Musk, Cloudflare lays off 20% of its staff to "[create] value in the agentic AI era," and the latest enterprise moves.
Please forward this email to a friend or colleague! If it was forwarded to you, sign up here to get Runtime each week, and if you value independent enterprise tech journalism, click the button below and become a Runtime supporter today.
Delta blues
Anthropic had to do something; six months after the release of Claude Opus 4.5 jump-started the AI coding agent craze, the company risked squandering all that enterprise goodwill with the reliability issues, rate limits, and rising prices caused by a shortage of computing capacity. But going into business with Elon Musk was certainly a choice.
SpaceX announced Wednesday that Anthropic has agreed to purchase all the computing capacity at xAI's Colossus 1 data center in Memphis, Tenn. "Colossus 1 features over 220,000 NVIDIA GPUs, including dense deployments of H100, H200, and next-generation GB200 accelerators," SpaceX said in a statement, and Anthropic immediately increased usage limits for Claude Code.
- "First, we’re doubling Claude Code’s five-hour rate limits for Pro, Max, Team, and seat-based Enterprise plans," and Anthropic will also remove limits on Claude Code usage during peak hours, it said in a blog post.
- It also significantly increased the limits on how often customers can access APIs for Claude Opus models, which will allow companies to build more complex agents that depend on Opus.
- Anthropic had already been struggling to serve Claude customers before Opus 4.5 caused a huge surge in enterprise demand in the first quarter, which CEO Dario Amodei put at "80-times growth" when the company had been planning for a 10x surge, according to The New York Times.
But Colossus 1 isn't just another massive data center. It has quickly earned a reputation as one of the most notorious projects of the recent AI boom thanks to xAI's casual regard for environmental rules and standards and its inability to make efficient use of all those GPUs.
- Local Memphis residents have been complaining about air pollution from the facility pretty much from the moment it opened for business, pointing to xAI's decision to use unpermitted methane gas turbines to power those chips in order to get up and running quickly.
- xAI later received permission from Shelby County to use the turbines, but the NAACP sued the company claiming violations of the Clean Air Act.
- Just last month, The Information reported that xAI stopped work on a wastewater treatment facility designed to recycle water used at Colossus 1 as well as Colossus 2 in nearby Southhaven, Miss. (which SpaceX will continue to use), and isn't saying why.
- And earlier this year Business Insider reported that xAI executives were aghast at the "embarrassingly low" 11% efficiency rating of its AI model training operations, compared to an industry standard of around 35% to 45%.
It's not clear how much operational control Anthropic will be allowed to exercise over Colossus 1, but as the first customer of what appears to be SpaceX's neocloud era, it should not hesitate to throw its weight around to improve the air and water pollution situation in the Memphis area. Still, the bigger long-term risk for Anthropic is that it just entered into a business relationship with Elon Musk.
- Musk has not exactly been a fan of Anthropic over the last year, and remains close to Trump administration officials that labeled it a supply chain risk for refusing to let the Pentagon do whatever it wanted with Claude models.
- Anthropic was careful to note that Colossus 1 is just one element of its compute strategy, which relies more heavily on proven hyperscalers like AWS and Google Cloud.
- But that element is subject to the whims of a very unpredictable man, who assured his followers Wednesday that "we reserve the right to reclaim the compute if their AI engages in actions that harm humanity," which, of course, he didn't bother to define.
- As Seattle investor Charles Fitzgerald likes to put it, "Never take a dependency on Elon Musk."
And just like that
While reasonable people can disagree on the scope of the impact, there's no question that software companies will go through a lot of changes over the next few years as AI coding agents mature. The economics of building software may never be the same, which could reduce demand for software as well as the number of people required to produce world-class products.
Cloudflare announced Thursday that it was laying off about 20% of its staff, or 1,100 employees. "Today’s actions are not a cost-cutting exercise or an assessment of individuals’ performance; they are about Cloudflare defining how a world-class, high-growth company operates and creates value in the agentic AI era," co-founders Matthew Prince and Michelle Zatlyn wrote in a memo published by the company.
However, Cloudflare also disappointed investors Wednesday by issuing revenue guidance below analyst estimates, which sent its stock plunging nearly 17% in after-hours trading. Add Cloudflare to the list of companies who have fired workers in recent months citing the productivity gains of AI without acknowledging the effect that this platform shift is also having on their business models.
Enterprise moves
Michel Combes is the new CEO of Lambda, replacing co-founder Stephen Balaban, who will become chief technology officer of the neocloud.
Bill Koefoed is the new chief financial officer at Harness, joining the software tools company after financial leadership roles at OneStream, which went public in 2024, and Blue Nile.
The Runtime roundup
CoreWeave beat Wall Street's estimates for revenue but provided lower-than-expected revenue guidance for the current quarter amid wider-than-expected losses, sending its stock down 12% in after-hours trading.
Datadog, meanwhile, enjoyed a 31% jump in its stock Thursday after reporting blowout earnings that will give the SaaSpocalpyse contingent something to think about.
Thanks for reading — see you Saturday!