Listen to this article
Average 4 minutes
The audio version of this article was created by AI-based technology. It can be mispronounced. We are working with our partners to continuously review and improve the results.
Anthropic on Wednesday said it has reached an agreement to use Elon Musk’s SpaceX computing services, marking a tussle with its critic and boosting the two companies in a high-stakes intelligence competition.
Under the agreement, Anthropic will use the full computing power of SpaceX’s Colossus 1 facility in Memphis, Tenn., which houses more than 220,000 Nvidia processors and will provide Claude’s interviewer with 300 megawatts of new capacity within a month.
The deal gives IPO-bound SpaceX a marquee customer as it looks to sell investors on its AI ambitions, while helping Anthropic reduce energy constraints following increased demand for products like its AI coding tool, Claude Code.
The announcement came as Anthropic held a developer day in San Francisco on Wednesday, where it introduced a new Claude AI feature called “dreaming,” which aims to help its AI systems learn by reviewing activity between sessions, seeing patterns and reviewing files that store users’ preferences and other context.
Spurred by the increased volume of the SpaceX deal and other recent deals, Anthropic said it is doubling the Claude Code limits for its premium plans, removing peak-hour usage caps for Pro and Max accounts, and significantly increasing the number of requests developers can make on their Claude Opus models.
Many Canadians are turning to artificial intelligence for questions about their health, but research shows that the answers they get may be wrong or misleading. As CBC’s Jo Horwood explains, while some AI tools are being tested for use in the health industry, researchers are hesitant to put them in the hands of consumers.
The company said it is also interested in working with SpaceX to develop more gigawatts of space-based orbital data centers – one of Musk’s key goals and the main driver behind SpaceX’s initial offering, as the effort is expected to be capital-intensive and technically challenging.
Ryan Mallory, CEO of datacenter operator Flexential, said, “The fact that large companies are even discussing space computing tells you how badly the market wants power and scale.”
Musk reverses stance on Anthropic
In a statement on X, Musk said he made the decision to lease computing power after spending time with Anthropic leaders last week. Their work to ensure that the Anthropic Claude AI is “fit for humanity” impressed him, he said.
“No one has removed my evil detector,” wrote Musk, who is fighting OpenAI and its CEO, Sam Altman, in court for allegedly violating its mission to develop AI for the benefit of humanity.
He added that SpaceX has transferred its AI training efforts to Colossus 2 and that it will provide computing capacity to other AI companies doing similar efforts to benefit humanity, as SpaceX launches satellites to competitors “at the right terms and prices.”
The comments marked a sharp reversal from comments Musk made in February, when he accused Anthropic AI of bias. “Frankly, I don’t think there’s anything you can do to escape the inevitable Anthropic paradox of being the villain,” he wrote at the time in a post on X.
Push to win business customers
Anthropic said Wednesday that its “dream” feature was available as a research preview and comes with its software for managing agents, or AI programs that perform tasks with minimal human involvement.
The move is part of Anthropic’s drive to win over business customers, following the rise in popularity of Claude Code that has intensified competition and prompted OpenAI to shift efforts from products like its Sora video production tool to focus more on the fast-growing market for AI-powered coding.
On stage at the San Francisco event, Boris Cherny, Anthropic’s head of Claude Code, showed how developers can stop the “processes” that organize the Anthropic AI computer programmer to take action.
“The default is not, ‘I’m going to inform Claude Code.’ The default now is, ‘I’m going to ask Claude immediately for Claude Code,'” he said.
Cherny closed the keynote with a message about the productivity potential of AI. “The ability is there. The gap that remains is how we quickly fill it in,” he said.
