I’ve been watching the credit side of the AI build-out for months now, and the latest development around Broadcom feels like one of those moments when the numbers stop being abstract and start looking genuinely large. Reports indicate the company is in advanced talks to arrange financing that could reach as high as $100 billion through a special-purpose vehicle structure. Most of that debt would sit off the main corporate balance sheet. At the same time, Broadcom’s credit default swaps have moved sharply higher. The combination is hard to ignore.
Why This Deal Matters Right Now
The scale alone grabs attention. Under the structure being discussed, Broadcom would provide a guarantee on a senior secured portion that could range between $60 billion and $70 billion. An additional junior tranche of roughly $30 billion is also under consideration. Taken together, the package would rank among the largest SPV financings ever arranged for technology infrastructure. The primary purpose is to secure chips and related equipment for artificial intelligence workloads, with companies such as Anthropic expected to be among the beneficiaries.
This is not happening in isolation. Private credit firms including Blackstone and Apollo are said to be involved in the conversations. Those same names appeared in earlier large-scale AI infrastructure arrangements. The model follows a pattern that has become more common: create a dedicated vehicle, issue debt against the assets inside it, keep the bulk of the liability off the sponsor’s balance sheet, and structure the senior pieces so they can achieve investment-grade ratings and lower borrowing costs.
The Broader Context of AI Capital Spending
Anyone following the sector knows the capital expenditure numbers have grown enormous. Hyperscalers and chip suppliers have been locking in multi-year spending plans that run into the hundreds of billions. Much of that spending needs financing. Equity markets have largely cheered the growth story. Credit markets have been more cautious. Spreads have widened at times, and the volume of protection being bought against certain names has climbed to record levels.
I’ve found that the tension between equity optimism and credit caution is one of the more interesting features of the current cycle. Equity investors often focus on revenue projections and market share. Credit investors focus on cash flow coverage, leverage, and the sheer volume of new paper coming to market. When a single company starts talking about a package that could approach $100 billion, those credit concerns become harder to dismiss.
The proposed Broadcom arrangement would add to an already busy pipeline. Earlier deals involving other major technology names have set precedents for how these structures can be built. The goal is usually the same: give the end users of the chips access to capacity without forcing every dollar of the financing onto one corporate balance sheet. In theory it works. In practice it still requires buyers willing to hold the debt at yields that make sense for both sides.
How the Structure Is Expected to Work
Details remain fluid, yet the broad outlines are clear enough. A special-purpose vehicle would issue the debt. Broadcom would backstop a meaningful portion of the senior tranche. The junior piece would absorb more risk and therefore carry higher yields. The chips and related infrastructure purchased with the proceeds would serve as collateral. Leases or capacity agreements with AI companies would provide the cash flows needed to service the debt.
This approach has been used before on a smaller scale. The advantage is obvious: the sponsor can support the financing without showing the full amount of leverage on its own books. Rating agencies have been willing, in previous cases, to assign investment-grade labels to the senior layers when the collateral and guarantees are structured carefully. That lowers the overall cost of capital compared with pure corporate issuance at higher leverage ratios.
Still, size changes the equation. A $35 billion package is one thing. A package that could approach three times that size is another. The market has to absorb the paper. Investors have to feel comfortable with the collateral values over the life of the loans. And the broader interest-rate environment has to remain supportive enough that the all-in yields look attractive relative to other opportunities.
Credit Default Swaps Are Sending a Message
While the equity side of Broadcom has shown resilience, the credit side has been more restless. CDS levels have widened notably. That movement reflects growing concern about the volume of new debt tied to AI infrastructure across the sector. When one large issuer prepares another sizable package, protection buyers tend to react. The same pattern has appeared with other technology names that have announced major funding plans.
In my view, the CDS market is doing what it is designed to do: price the perceived risk of higher leverage and heavier issuance. Equity markets can look through near-term balance-sheet effects if the growth story remains compelling. Credit markets are less forgiving on that point. They care about the absolute amount of debt that will need to be refinanced or serviced in the years ahead.
The fact that CDS volumes for several hyperscaler-related names have reached elevated levels suggests that a meaningful group of investors is actively managing exposure. That does not mean a crisis is imminent. It does mean the market is no longer treating AI-related corporate debt as risk-free or even low-risk by default.
Private Credit’s Growing Role
One striking feature of the current wave of deals is the heavy involvement of private credit. Traditional public bond markets still matter, yet a large share of the newer AI infrastructure financings has been arranged through private channels. Firms with substantial dry powder have been willing to take on the junior and mezzanine risk in exchange for higher returns. The senior pieces can then be structured more conservatively and marketed to a wider investor base.
This shift has practical consequences. Private credit can move faster and accept more complex structures. It can also lock up capital for longer periods. When the deals grow into the tens of billions, the concentration of risk among a smaller group of large players becomes more noticeable. If one or two major private credit vehicles decide to slow their participation, the entire funding pipeline could feel the effect.
I keep coming back to the same question: how much capacity does private credit really have for this particular theme before returns compress or risk limits are reached? The answer will shape how much additional AI infrastructure can be financed at attractive terms over the next few years.
Impact on the Broader Rate Environment
There is a secondary effect that deserves attention. Heavy corporate issuance related to AI has at times competed with government paper for investor attention. When large technology names flood the market with new debt, some of the demand that might otherwise have gone into longer-dated Treasuries can shift. That dynamic can put upward pressure on yields at the margin.
Policy makers have tools to manage supply and demand in the government market, yet the underlying pressure from private-sector borrowing remains. If the AI capital expenditure cycle continues at the currently projected pace, the volume of new corporate debt will stay elevated for several years. That is not a short-term story. It is a multi-year funding requirement.
Perhaps the most interesting aspect is the feedback loop. Higher corporate issuance can contribute to firmer longer-term rates. Firmer rates raise the cost of the very capital expenditure that is driving the issuance. At some point the economics of the next project start to look less compelling. We are not at that point yet, but the direction of travel is visible.
What Success Would Look Like
For the Broadcom package to clear the market smoothly, several conditions need to hold. First, the senior debt must achieve ratings and pricing that attract a broad enough buyer base. Second, the collateral values and lease arrangements must look robust under reasonable stress scenarios. Third, the overall size must not overwhelm available demand in the private and public credit markets at the same time.
If those pieces fall into place, the deal could become a template for further large-scale financings. Other chip suppliers and infrastructure providers might follow similar paths. The AI capacity build-out would continue with less strain on individual corporate balance sheets. Equity valuations could remain supported by the growth narrative.
If the package struggles to find buyers at acceptable terms, the signal would be different. It would suggest that credit markets are starting to push back on the sheer volume of AI-related leverage. That push-back would not stop the capital spending overnight, but it would raise the cost of capital and force more selectivity in project selection.
Risks That Cannot Be Ignored
Every large financing carries risk. In this case the list includes technology obsolescence, shifts in demand for specific chip architectures, changes in the competitive landscape, and the possibility that AI workloads do not scale as projected. Collateral values depend on the usefulness of the underlying hardware over the life of the loans. If that usefulness declines faster than expected, recovery rates in a stress scenario could disappoint.
There is also the question of concentration. A relatively small number of large technology companies and private credit funds are driving a large share of the current activity. That concentration can amplify both positive and negative outcomes. Strong demand keeps the pipeline open. Any sudden caution among the key players can slow the entire process.
I’ve noticed that conversations about these risks tend to be quieter than conversations about growth potential. That is understandable. Growth stories are more exciting. Yet credit markets ultimately price risk, and the current widening in certain CDS levels suggests that at least some participants are already adjusting their assumptions.
- Technology risk remains real even with strong near-term demand
- Collateral values depend on sustained utilization of specialized hardware
- Concentration among a few large sponsors and lenders creates potential bottlenecks
- Interest-rate sensitivity can affect both project economics and investor appetite
- Off-balance-sheet structures still require ongoing market confidence
Equity Versus Credit Perspectives
It is worth pausing on the divergence between equity and credit views. Equity investors have rewarded companies that can show clear exposure to AI infrastructure growth. Share prices have reflected expectations of rising revenue and expanding margins. Credit investors have focused more on the balance-sheet implications of funding that growth. The two groups are looking at the same companies through different lenses.
In practice this means that positive equity reactions to news of large deals can coexist with wider credit spreads. The market is not being inconsistent. It is simply assigning different weights to different factors. Growth potential supports equity valuations. Leverage and issuance volume weigh on credit spreads. Both signals can be true at the same time.
For anyone managing a portfolio that includes both asset classes, the divergence creates both opportunities and challenges. The equity side may continue to perform if the AI narrative holds. The credit side may require more active management of duration, spread, and issuer concentration risk.
Looking Further Ahead
The capital requirements associated with AI infrastructure are not going away quickly. Industry projections still point to several more years of heavy spending. Much of that spending will need external financing. The structures used to raise the money will continue to evolve. Off-balance-sheet vehicles, private credit participation, and creative collateral arrangements are likely to remain part of the toolkit.
What may change is the pricing. As more paper comes to market, investors will demand compensation that reflects both the volume and the underlying risks. Spreads that looked tight a year or two ago may no longer be available. That adjustment process is already visible in certain CDS markets and in the secondary trading of recent issues.
The Broadcom discussions simply bring the issue into sharper focus. A single package that could reach $100 billion is large enough to test market capacity. If it succeeds, it will encourage similar approaches. If it encounters resistance, it will force a recalibration of expectations about how much AI-related debt the market can comfortably absorb.
Either outcome carries information. Markets are good at revealing preferences when the numbers get large enough. Right now the numbers are getting large, and the credit market is starting to speak more loudly than it did a few quarters ago.
Practical Takeaways for Market Participants
For equity investors the key is still the growth trajectory and the competitive position of the companies involved. Strong demand for custom AI chips supports the revenue outlook. Partnerships that lock in long-term volume can provide visibility. Valuation multiples will depend on how durable that growth proves to be once the initial wave of spending matures.
For credit investors the checklist is longer. Absolute leverage levels, the quality of collateral, the strength of any guarantees, the expected life of the underlying assets, and the overall volume of similar paper coming to market all matter. CDS levels offer a real-time gauge of how the market is pricing those factors. Secondary spreads on existing bonds provide another reference point.
For policy observers the interaction between heavy corporate issuance and government funding needs remains relevant. Large private-sector borrowing can influence the level and volatility of longer-term rates. That influence is not always dominant, yet it is no longer negligible.
The real test will come when the next several large packages try to clear the market at the same time. Capacity is finite even in deep markets.
That observation feels particularly relevant right now. The pipeline of potential AI-related financings is substantial. Not every deal will be the same size as the one under discussion, yet the cumulative total is still impressive. Market appetite will determine how many of those deals can be completed on terms that work for both borrowers and lenders.
A Measured View of the Road Ahead
It is easy to get caught up in the scale of the numbers and lose sight of the underlying economics. AI infrastructure has real demand behind it. Companies are willing to commit capital because they see competitive necessity and potential returns. Financing structures have become more sophisticated in response. Those are constructive developments.
At the same time, no market has unlimited capacity. Credit markets in particular tend to reprice risk when issuance accelerates and leverage rises. The recent movement in CDS levels is consistent with that historical pattern. Whether the adjustment remains orderly or becomes more abrupt will depend on the volume of new supply, the evolution of interest rates, and the actual performance of the underlying AI projects.
I expect the next several quarters to provide clearer evidence one way or the other. If large packages continue to clear without significant concession in pricing, confidence will build. If pricing starts to widen more persistently or if certain deals are delayed or resized, the market will have delivered a different message. Either way, the information will be useful.
For now the story remains one of ambition meeting the practical constraints of capital markets. Broadcom’s reported discussions illustrate both the ambition and the constraints. The ambition is to secure massive computing capacity for the next phase of AI development. The constraint is that even well-structured debt still needs willing buyers at yields that make sense. Finding that balance is the challenge of the moment.
The coming months will show how successfully that balance can be struck. Credit markets have already begun to price in higher levels of activity and higher levels of risk. Equity markets continue to focus on the growth opportunity. The gap between those two perspectives may narrow or widen depending on how the next wave of financings is received. Watching both sides carefully still seems like the most practical approach.
In the end, the size of the proposed package is less important than the market’s willingness to fund it on sustainable terms. That willingness is being tested in real time. The outcome will influence not only one company’s funding strategy but the broader trajectory of AI infrastructure investment in the years ahead.
