On October 26, Bank of America, Blackstone, Citi, and Morgan Stanley announced a $63 billion financing structure to enable AI compute capacity for Anthropic, using Broadcom’s technology. The deal runs through a special purpose vehicle that will buy Broadcom XPU accelerators and networking gear and lease them to Anthropic, with deliveries starting in 2027.
That’s not a venture round. That’s asset-backed financing. The same structure used for power plants, ports, and cell towers.
And it wasn’t the only one.
The $63 Billion Blueprint: How AI Infrastructure Gets Financed Now
Here’s what’s actually interesting about the Anthropic deal. It’s not just the size. It’s the structure.
Bank of America, Citi, and Morgan Stanley arranged the first-lien portion. Blackstone provided the committed portion of the second-lien financing. Broadcom is providing residual value support equivalent to the first lien. The chips are bought by an SPV, leased to Anthropic, and the debt is serviced by Anthropic’s lease payments.
That’s how airlines finance planes. How shipping companies finance vessels. How telecom operators built their networks.
AI compute is now being treated as a depreciable, financeable asset class. And the banks are treating it that way because the demand is predictable enough to underwrite.
Jon Gray, Blackstone’s President and COO, put it plainly: “The rapidly growing demand for AI is creating an enormous opportunity to finance compute capacity at scale”.
That sentence matters more than any benchmark.
Anthropic’s $100 Billion IPO: The Balance Sheet Behind the Hype
Two days after the financing announcement, Anthropic filed its S-1 with the SEC. If it raises the expected $100 billion, it will be the largest IPO in history.
The numbers in the filing are the ones worth paying attention to.
Anthropic’s revenue for the first half of 2026 was $16.29 billion, up from roughly $1.1 billion a year earlier—a 13x jump. Q3 revenue hit $16.8 billion, up about 45% quarter-over-quarter.
But the cost side tells the real story. Anthropic spent nearly $9 billion on compute and infrastructure in the first half—more than its entire 2025 spend of $7.3 billion. That’s about 55% of revenue going to compute.
And the long-term commitments are staggering. According to the prospectus, Anthropic has locked in **at least $657 billion in compute contracts** through 2045, across counterparties including Nscale, Lambda, AMD, Broadcom, SpaceXAI, Google, AWS, and Microsoft. If negotiations in progress are included, that number could reach $750 billion.
This is a company that is spending like a utility and raising like a tech startup. The $63 billion financing structure was the bridge between the two.
Naver and Brookfield: The $9 Billion Deal That Unlocked Nvidia
On October 27, Naver signed a contract with Brookfield to secure up to $9 billion in AI computing infrastructure.
The deal was structured through a special purpose vehicle called Asset Co., funded entirely by Brookfield, which will buy GPUs and data centers. Naver Cloud will operate the infrastructure and pay usage fees linked to business revenue, rather than buying the hardware outright.
The critical detail: this $9 billion commitment was a precondition for Nvidia’s planned $1 billion strategic equity investment in Naver. With the Brookfield deal signed, Nvidia’s investment can proceed.
Naver plans to start 55 MW of operations in early 2027 and scale to 200 MW by 2028, with a long-term goal of 1 GW.
This is the same pattern as the Anthropic deal: infrastructure financed by third-party capital, operated by the company that needs the compute, with the hardware supplier taking an equity stake. The capital structure is becoming as important as the technology.
The Telecom Angle: Agents That Actually Run the Network
While the financing headlines grabbed attention, the telecom industry quietly moved closer to autonomous operations.
On October 28, du (the UAE’s second telecom operator) and Comarch launched agentic AI for OSS network operations. The system uses autonomous AI agents that monitor network conditions, analyze multi-vendor logs, and handle routine workflows—including automatically finding and fixing missing configuration data, diagnosing discrepancies between vendor systems, and cutting resolution times from days to minutes.
The deployment starts in secure sandbox environments with human-in-the-loop verification, but the architecture is designed for closed-loop autonomous execution.
du’s CIO, Justin Shields, framed it around trust: “Resilience and control are at the center of how we plan to scale AI”.
That’s the right framing. Telcos aren’t deploying agents because they’re excited about the technology. They’re deploying them because network complexity has outgrown manual operations. The same week, Oracle outlined its blueprint for autonomous telcos driven by agentic AI, describing cross-domain operations that connect networks, IT, and customer journeys.
The telecom network is becoming the first large-scale industrial system where AI agents are trusted to operate production infrastructure. Not because the industry is reckless. Because the alternative—humans manually managing thousands of vendor systems and configuration states—doesn’t scale.
AI infrastructure is becoming financeable, and financeable infrastructure gets built.
The companies winning aren’t the ones with the best model. They’re the ones with the balance sheet, the contracts, and the capital structure to build at scale. Broadcom financing Anthropic’s chips. Brookfield funding Naver’s data centers. Banks underwriting the leases. Nvidia taking equity in its customers.
This is what a maturing industry looks like. Not less ambitious. More structured. The wild ideas are still there. But now they come with a term sheet.
This was the week AI stopped being a venture bet and started being an asset class.
#AI, #Anthropic, #Broadcom, #Blackstone, #Naver, #Brookfield, #Nvidia, #TelecomAI, #AIIinfrastructure, #AgenticAI, #DigitalTransformation, #Innovation, #TechNews, #CapitalMarkets, #EnterpriseAI