September 27, 2026
Anthropic has committed $11.6 billion over seven years to run infrastructure on Akamai's cloud, with the deal structured so it could grow to roughly $20 billion. What makes this notable isn't the size -- big AI infrastructure deals are routine now -- it's what Anthropic is actually buying: CPU capacity, not the GPU clusters everyone assumes AI companies are hoarding. Akamai is also handing Anthropic a stake that could reach 5% of its stock as spending increases, which is an unusual way to structure a vendor contract. It looks less like a customer relationship and more like a partial merger of incentives.
For business buyers, the lesson isn't about Anthropic specifically. It's that the infrastructure race is diversifying in ways that will eventually show up in your bill, one way or another. Compare this to British neocloud Nscale, which just secured $3.36 billion in convertible financing from Third Point, Nvidia, and others ahead of a US IPO, purely to keep building AI data centers. And then look at Crusoe, which walked away from a $1.25 billion plan to use Boom Supersonic's turbines at its AI data centers -- a reminder that not every infrastructure bet pans out, even when the checks are already being written. We covered the Akamai arrangement in more depth in our analysis of what the deal really signals. The takeaway for anyone evaluating vendors: infrastructure spending at this scale eventually gets passed through in pricing, and it's worth asking any AI vendor what's actually underneath their product before you sign a multi-year contract.
Blue Cross Blue Shield says hospitals using AI tools drove an additional $942 million in healthcare spending over two years. That's a striking number, and it cuts against the tidy pitch that AI simply makes operations cheaper and faster. The insurer's framing deserves some skepticism of its own -- insurers have obvious incentives to blame rising costs on someone else, and hospitals could reasonably argue AI is catching billable issues or improving care in ways that cost more up front but save money later. Still, the core claim is worth taking seriously: when AI systems are layered into billing, diagnostics, or documentation without careful oversight, they can just as easily inflate costs as cut them.
This is the pattern business leaders should watch for in their own operations, not just healthcare. An AI tool that speeds up a workflow can still increase total spend if it generates more line items, more follow-ups, or more edge cases that need human review. We dug into this dynamic already in our piece on who actually pays when hospital AI runs up the bill. The general principle applies broadly: adopting AI without clear metrics on what it's supposed to save -- and monitoring whether it actually does -- is how you end up with a tool that looks efficient on paper and expensive in practice.
Put the Akamai deal and the Blue Cross Blue Shield findings side by side and you get a clearer picture of where AI's costs actually land: not in the sticker price of a subscription, but in the infrastructure commitments behind the scenes and the downstream operational effects once a tool is live. Businesses evaluating any AI-powered platform, whether it's a CRM or a billing system, should ask two questions vendors rarely volunteer answers to: what's the actual compute cost structure behind this tool, and what's the plan for measuring whether it saves money once deployed, not just whether it looks impressive in a demo.
If your organization is running AI tools without a clear read on total cost of ownership, or without a way to prove the tool is actually reducing spend rather than just adding activity, what would it take for you to trust that number?
Sources