On Monday morning, Bloomberg reported that OpenAI was raising $4 billion for a new enterprise joint venture called The Deployment Company, valued at $10 billion, backed by 19 investors including TPG, Brookfield Asset Management, Bain Capital, and SoftBank. A few hours later, Anthropic announced its own enterprise venture, backed by Blackstone, Goldman Sachs, and Hellman and Friedman, valued at $1.5 billion. The two announcements, released within the same business day, used almost identical language about "new channels for enterprise AI deals" and "preferred sales access" to investors' portfolio companies. The timing was either extraordinary coincidence or both labs had been watching each other's fundraising for weeks and neither wanted to let the other go first.
The structural logic of both ventures is the same. AI labs burn through capital at rates that require continuous fresh investment. Enterprise contracts are large, sticky, and predictable in ways that consumer subscriptions are not. By bringing in private equity and institutional investors as founding partners, both OpenAI and Anthropic are creating a sales channel with built-in buyer relationships: investors who own portfolio companies now have a financial incentive to steer those companies toward their preferred AI provider. It is a clever mechanism for converting capital investment into customer acquisition at scale.
The numbers tell the story of where each company is. OpenAI says it now has more than nine million paying business users, with business revenue accounting for more than 40% of total income. Anthropic reported this year that it has more than 300,000 business customers and that its annualised revenue has grown from roughly one billion dollars in January 2025 to thirty billion today. Both are growing fast. Both are spending faster than they earn. The joint ventures are partly about reaching more customers, partly about securing the capital to keep building while the economics catch up.
What the joint venture announcements did not capture is how differently the two labs are thinking about distribution. That gap became visible last week through a separate story involving OpenClaw, the open-source AI agent framework that has accumulated 3.2 million users and 346,000 GitHub stars in under five months. In April, Anthropic blocked Claude Pro and Max subscribers from accessing their flat-rate subscriptions through OpenClaw and similar third-party agent frameworks. The reason was cost: autonomous agents running continuously can generate thousands of API calls per day, consuming far more compute than a human chatting in a browser window. Anthropic judged that unlimited subscription access through an agent framework was economically unsustainable and shut it down.
OpenAI made the opposite call. On 2 May, Sam Altman posted on social media at 2:33 in the morning that ChatGPT users could now log in to OpenClaw using their existing subscription and access GPT-5.4 through the agent framework for a combined cost of $23 per month. For context, using the OpenAI API directly at the volume an autonomous agent generates would cost hundreds of dollars monthly. OpenAI is subsidising the difference, betting that the distribution gain outweighs the margin loss. The logic is similar to how mobile carriers once subsidised handsets to lock in subscription contracts.
The competitive framing is simple enough: Anthropic looked at OpenClaw and saw a cost problem; OpenAI looked at the same product and saw a distribution opportunity. One company locked the door. The other opened it. But the implications are more complex. OpenClaw has had serious security vulnerabilities, including a remote code execution flaw that let malicious websites silently take over users' machines, and a breach that exposed 1.5 million API tokens. The vulnerabilities have been patched in current versions, but a significant portion of installed instances are running older software. By tying its billing and authentication to OpenClaw, OpenAI is accepting that its brand flows through an open-source platform with an unusual security history.
Taken together, Monday's joint venture announcements and the OpenClaw split reveal two companies that are racing toward the same destination by quite different routes. Both need enterprise revenue. Both need distribution. They disagree about whether openness and margin protection are compatible. Which bet pays off will depend partly on whether OpenClaw's 3.2 million users convert to paying customers, and partly on whether the corporate buyers being courted by the new joint ventures care more about safety controls or access flexibility. That question does not yet have an answer, and both labs are spending heavily on the assumption that their read of it is correct.