Anthropic is on track to post its first quarterly operating profit, projecting $559 million on $10.9 billion in June-quarter revenue, more than double the $4.8 billion it recorded in Q1. The Wall Street Journal reported the same figures earlier on Wednesday. The expected margin sits just under 5.1%, thin by software industry standards but a first for a frontier AI lab that told investors last summer it would not reach full-year profitability until at least 2028. Revenue growth has been driven by developers building with Claude Code and enterprise clients using the Mythos model to find security flaws in production software. These business products yield much more reliable recurring income than chatbot subscriptions by consumers, which explains why Anthropic is making money, whereas OpenAI remains in the red even as its sales soar. SpaceX’s S-1 put a price on the compute SpaceX’s S-1 filing, released Wednesday as part of its IPO process, revealed Anthropic will pay $1.25 billion each month for compute access running through May 2029. The deal covers both of SpaceX’s AI training clusters, Colossus and Colossus II. Business Insider reported the arrangement could deliver more than $40 billion to SpaceX over the full contract term. Either side can terminate with 90 days’ notice, and fees are reduced during the capacity ramp-up this month and next. Anthropic first announced the SpaceX partnership at its May developer conference. Chief Product Officer Ami Vora said the company would gain access to “more than 220,000 GPUs and over 300 megawatts of power capacity.” With utilization fully maxed out, this means about $5,680 per GPU per month. The public cloud spot rates for similar Nvidia H100 machines on AWS, Azure, and Google Cloud tend to range from $1,500 to $3,000 per GPU per month. The discrepancy shows that Anthropic is not only paying for computing power but also guaranteed access for three years, specialized infrastructure, and access to power. As Cryptopolitan reported on May 6, CEO Dario Amodei disclosed that Anthropic grew 80-fold on an annualized basis in Q1, far exceeding the company’s internal projection of 10-fold. The SpaceX deal was announced the same day to address the compute shortage that growth created. SpaceX is turning AI losses into a cloud business SpaceX’s AI division posted a $2.5 billion operating loss against $818 million in Q1 revenue, per the S-1. Musk posted on X that SpaceX was in discussions with other companies about “offering AI compute as a service at significant scale,” signaling the company sees compute leasing as a revenue line beyond serving its own xAI models. SpaceX merged with xAI earlier this year. The S-1 identified AI infrastructure as one of SpaceX’s largest future addressable markets. Research firm Dealroom has described the trend as part of the rise of “neocloud” providers, firms that lease GPU capacity outside the traditional hyperscale cloud market. The Google-Blackstone AI cloud venture announced this week targets the same market with TPU-based compute rather than Nvidia GPUs. Anthropic’s overall compute portfolio now spans multiple providers. It holds contracts for up to 5 gigawatts with Amazon, including nearly 1 gigawatt of new capacity by end of 2026. A separate 5-gigawatt deal with Google and Broadcom begins in 2027. A strategic partnership with Microsoft and Nvidia covers $30 billion of Azure capacity. The SpaceX contract adds 300 megawatts on top of all of that. The profit is real but the bill is enormous The $559 million operating profit includes model training costs but excludes stock-based compensation. The operating profit figure also reflects a quarterly compute bill to SpaceX alone of $3.75 billion (three months at $1.25 billion per month). At $10.9 billion in revenue, that means SpaceX’s compute costs consume roughly 34% of Anthropic’s top line before any other infrastructure spending is counted. The 90-day termination clause in the SpaceX contract reflects how fast pricing and infrastructure needs are shifting. Anthropic may not remain profitable for the full year. Reports suggest that the company plans to increase spending on computing and model training in the second half, which could push margins back into negative territory. From the perspective of the broader artificial intelligence industry, Anthropic’s quarterly results are important because they answer a question that has followed every frontier lab since ChatGPT launched: Can this business model actually turn a profit? The answer, at least for one quarter, seems to be yes. The viability of this will depend upon whether revenue continues to outstrip the compute bill. If you're reading this, you’re already ahead. Stay there with our newsletter .
