Broadcom is negotiating what could become one of the largest debt financings ever assembled around the artificial-intelligence industry, highlighting the extraordinary amount of capital now required to build the computing infrastructure needed by frontier AI companies.

According to Bloomberg reporting cited by Reuters, Broadcom is in talks with lenders to raise more than $60 billion for an AI chip financing arrangement that would benefit Anthropic and other AI companies. The potential structure could ultimately reach as much as $100 billion.

The discussions represent another major escalation in the AI infrastructure race.

As models become more powerful and are used by more customers, the cost of the chips, networking equipment and data-center capacity required to run them is becoming so large that traditional corporate financing is increasingly being supplemented by structured debt.

The potential Broadcom arrangement could include roughly $30 billion of junior debt alongside a senior-secured tranche ranging from approximately $60 billion to $70 billion.

That would bring the total financing to potentially as much as $100 billion.

The proposed debt would reportedly be issued through a special-purpose vehicle, a structure often used to separate a specific project's assets and cash flows from a company's broader balance sheet.

Blackstone and Apollo Global Management are in discussions about participating in the financing.

The structure illustrates how AI infrastructure is beginning to resemble large-scale project finance rather than conventional corporate technology spending.

Broadcom is one of the most important suppliers behind the industry's shift toward custom AI chips.

The company helps major technology companies design specialized accelerators and networking systems, allowing firms such as Alphabet and Meta to reduce their dependence on Nvidia's general-purpose GPUs.

Broadcom also has chip supply relationships with Anthropic and OpenAI.

That puts Broadcom in an unusual position.

It is not simply selling chips.

It is helping design the infrastructure architectures that AI companies will use for years.

The AI industry initially centered on Nvidia's GPUs because they provided the flexibility needed to train increasingly large models.

But as AI workloads become more specialized, technology companies have stronger incentives to develop custom chips.

Custom processors can be optimized for specific workloads and can potentially improve performance per watt, reduce operating costs and provide greater control over infrastructure.

Alphabet and Meta have both invested heavily in internal silicon efforts.

Broadcom's role is to help turn those designs into deployable hardware.

The company therefore benefits from the industry's desire to diversify beyond Nvidia.

Broadcom, Apollo and Blackstone already partnered on a $35 billion financing arrangement announced in June to expand Anthropic's computing capacity using Broadcom's custom chips and networking equipment.

That earlier agreement was designed as part of a broader plan to provide more than 20 gigawatts of computing capacity to major AI labs by 2028.

The new financing discussions suggest that the first project may be only the beginning.

If frontier AI companies continue expanding rapidly, they will need substantially more computing power.

That means the infrastructure supporting them will require financing at a scale comparable to major industrial projects.

The financial model of AI is changing.

Early-stage software companies could once scale primarily by hiring engineers and developing products.

Frontier AI companies are different.

They need enormous quantities of advanced chips, electricity, cooling capacity, networking equipment and data-center space.

As a result, the path from revenue growth to profitability can be complicated.

Companies can generate billions of dollars in sales while simultaneously committing even larger sums to future infrastructure.

That creates a financing gap.

Debt can help bridge it.

Blackstone and Apollo are exactly the type of investors that can fill this gap.

Large private-equity and credit firms control enormous amounts of institutional capital and are increasingly looking for exposure to infrastructure linked to the AI boom.

The attraction is potentially different from investing directly in an AI company.

Rather than betting entirely on whether one AI model becomes dominant, investors can finance the physical infrastructure required by multiple customers.

That can create a more diversified investment proposition.

The strategy also raises an important question: who bears the risk if AI demand slows?

Debt holders expect repayment regardless of whether a particular AI model remains commercially dominant.

If computing demand grows as projected, the infrastructure can generate substantial cash flows.

If demand disappoints, the fixed obligations associated with debt can become a burden.

That makes these financing structures highly dependent on long-term customer commitments.

Investors will therefore want to understand the contracts supporting the debt, the credit quality of customers and the residual value of the equipment if AI demand changes.

The Broadcom deal is another sign that AI data-center infrastructure is evolving into a major investment category.

Just as telecommunications infrastructure, power plants and transportation networks attract dedicated pools of capital, AI computing capacity may increasingly be financed as a long-lived infrastructure asset.

That could bring enormous amounts of institutional money into the sector.

It also reduces the amount of capital that AI companies must fund entirely from operating cash flow.

The timing is especially notable because broader debt markets are facing rising long-term interest rates.

U.S. Treasury yields have recently climbed to their highest levels in many years, increasing financing costs across the corporate sector.

At the same time, the Treasury has been expanding its own buyback operations to stabilize the long end of the government bond market.

AI borrowers are therefore entering the debt market at a time when investors are already demanding greater compensation for long-term credit risk.

That could make large AI financings more expensive.

The underlying demand story remains strong.

Alphabet, Amazon and Microsoft have all indicated that AI investment will remain elevated through 2026, while other companies are expanding their own data-center plans.

The scale of those commitments means chip suppliers, cloud providers and infrastructure financiers have substantial opportunities ahead.

Broadcom's position could be particularly strong because it sits at the intersection of custom silicon and networking.

Broadcom's expansion should not be interpreted as a replacement for Nvidia.

Nvidia remains the dominant supplier of AI accelerators and has an enormous software ecosystem built around CUDA.

The more likely scenario is a more diversified industry in which Nvidia remains the largest platform while custom-chip providers capture an increasing share of workloads suited to specialized architectures.

Broadcom is well positioned for that transition.

If large-scale structured debt becomes an effective way to finance AI infrastructure, it could accelerate the pace at which companies build computing capacity.

It could also lower the barrier for AI labs that have strong customer demand but cannot finance all of their infrastructure independently.

That could increase competition among AI model providers.

At the same time, it could increase financial-system exposure to the industry's future growth assumptions.

The more AI infrastructure is financed with debt, the more important the industry's cash flows become.

Broadcom's potential financing is therefore about much more than one debt deal.

It demonstrates that AI is becoming a capital-intensive industrial ecosystem in which technology companies, semiconductor suppliers, cloud providers and financial institutions increasingly operate together.

The total potential size — as much as $100 billion — is extraordinary.

If finalized, the deal would reinforce the idea that the next phase of AI development will depend not only on better models, but on the financial system's ability to fund enormous quantities of physical infrastructure.

That could be a major opportunity for Broadcom.

It could also become a test of whether investor appetite for AI-related debt can remain strong as borrowing costs rise.

For now, negotiations remain preliminary.

But the numbers alone send a powerful message: the AI race has become so expensive that the next generation of computing capacity may increasingly be built with Wall Street's balance sheet as well as Silicon Valley's.