Broadcom is working to arrange more than $50 billion in financing for the custom AI chip it is developing with OpenAI. The Wall Street Journal reported on October 7, 2026 that Broadcom has spent recent weeks advancing the plan and has talked with firms including Apollo and Blackstone about taking part. The report also cautions that the talks are early, the size could still change, and neither the borrower nor the terms have been disclosed.
This money has not landed yet
The nature of the deal matters first: this is a financing under discussion, not a signed loan, and certainly not cash in the bank. The figure above $50 billion is the package Broadcom hopes to assemble, and the presence of Apollo and Blackstone points to private credit and large asset managers rather than a conventional public bank syndicate. What can be confirmed today is limited to three things: the chip is being co-developed by Broadcom and OpenAI, the financing effort is real, and the scale runs to tens of billions of dollars. The interest rate, the maturity and who ultimately carries the repayment obligation are all undisclosed, so treating the money as a done deal would be premature.
Why even chips now run on borrowed money
The backdrop in the same report is just as telling. Large cloud providers used to pay for servers and chips out of cash flow, then turned to bond issuance; with the public debt market now pushed close to its limits, buyers are turning to Wall Street firms for private arrangements worth tens of billions apiece. In the same wave, Oracle is in talks with Apollo and Goldman Sachs to fund a large chip purchase, and SpaceX has spoken with lenders about roughly $40 billion tied to Nvidia chips. On the other side sit newer buyers such as OpenAI and Anthropic: they lack the cash reserves of the hyperscalers, have historically rented most of their compute, and now want to own more of their infrastructure to cut long-term costs and reduce dependence on suppliers. A custom chip is one piece of that strategy, and its bill first has to be borrowed.
Who ends up holding the risk
The real question behind the financing is where the risk lands. A chip takes years to move from design to deployment, and model roadmaps, compute prices and customer demand can all shift in the meantime; if a package of this scale closes, the lenders are betting that OpenAI's future revenue can carry the cost of the hardware. For OpenAI, borrowing to build its own silicon buys supply security and cost control at the price of tying part of its future to debt arrangements. For the wider industry, once compute expansion depends on private credit, the pace of the build-out is set not only by technology and demand but also by how long credit markets keep the door open. What to watch next is whether the talks turn into a signed deal, whether the final size shrinks, and who the borrower turns out to be.
