Excess Returns · · 1h 10m
Who Pays for the AI Buildout? Two Bond Investors Take Us Inside the Debt Market
With Alex Morris, Marcin Zdunek
ChorusBring the market's conversation to your AI. Learn more
Summary
Two bond investors, Marcin Zdunek and Alex Morris, argue that AI capex has outgrown internally generated cash flow, with roughly $1.5 trillion needed in 2027 across the largest hyperscalers and OpenAI and only about 30% covered by free cash flow, pushing the buildout into public and private debt markets. They say the market is now pricing AI borrowers differently: Oracle's CDS has risen most sharply, Microsoft's has stayed calm, SpaceX's new-issue spread has widened from 150 to about 260 basis points, and Meta, rated AA, pays a surprisingly high 128 to 130 basis points. Concessions are rising, order books are filling more slowly, and Marcin expects Anthropic to land in high yield, which would shrink the buyer base, while both see AI issuance rising in index weight and competing with Treasuries for the same capital. The discussion then turns to a US fiscal squeeze, with Alex warning about debt above $40 trillion, entitlement trust funds running out in the mid-2030s, and a government refinancing constantly at higher rates, and both favour short-duration Treasuries, TIPS and bond ETFs as practical ways to hold fixed income through this environment.
Key takeaways
- Alex Morris (bullish on Microsoft): Microsoft's credit insurance cost stays low and stable, so its debt carries little worry.
- Alex Morris (bearish on Oracle): Oracle's credit insurance got far more expensive, so it looks like a riskier AI borrower.
- Marcin Zdunek (bearish on Space Exploration Technologies): SpaceX's borrowing spread has widened sharply since its June bond, showing rising credit cost.
- Marcin Zdunek (bearish on Meta Platforms): Meta, rated AA, pays 128-130 bps over Treasuries, an unexpectedly high borrowing cost.
- Marcin Zdunek (bearish on Anthropic): Anthropic will likely be rated high yield once it starts issuing debt.
What was said, by company and theme
33 views from 2 speakers. Every quote is checked against the episode's transcript (how).
Microsoft
Across the shows: 47 speakers in other episodes, 50 bullish and 12 bearish calls, past 30 days
Where everyone stands on Microsoft
Alex Morris · F/m InvestmentsBullish
Microsoft's credit insurance cost stays low and stable, so its debt carries little worry.
Microsoft is an old, established company whose CDS has shown low volatility while peers have spiked. Alex attributes this to its diversified business and deep integration into the ecosystem, which he says reduces the worry about its ability to repay. It is the contrast point against the more volatile AI names.
“Microsoft, very old established company, pretty low volatility in the cost of that insurance. Microsoft has a pretty diversified business structure.”at 9:39
Oracle
Across the shows: 32 speakers in other episodes, 10 bullish and 30 bearish calls, past 30 days
Where everyone stands on Oracle
Alex Morris · F/m InvestmentsBearish
Oracle's credit insurance got far more expensive, so it looks like a riskier AI borrower.
Oracle's CDS cost rose the most among the names shown, a sign the market is pricing higher default risk. Alex asks whether it has bitten off more than it can chew, reflecting the AI-borrower concern that some companies are overextended relative to peers. He also notes Oracle has said it does not need more debt, which creates scarcity but does not reduce the underlying risk he is pointing to.
“Oracle probably leading that charge here, you see that one, things were more expensive, and they got way more expensive.”at 9:39
Space Exploration Technologies
Across the shows: 55 speakers in other episodes, 67 bullish and 26 bearish calls, past 30 days
Where everyone stands on Space Exploration Technologies
Marcin Zdunek · F/m InvestmentsBearish
SpaceX's borrowing spread has widened sharply since its June bond, showing rising credit cost.
SpaceX priced a ten-year bond in June at a spread of 150 basis points over Treasuries, and it now trades around 260 basis points. Marcin presents this as evidence that borrowing has become much more expensive for lower-quality AI-linked issuers such as Oracle and SpaceX. He expects the cost to keep rising.
“On June they priced a ten year bond at a spread of 150 basis points and it's now trading closer to [260 basis points]”at 13:57
Meta Platforms
Across the shows: 66 speakers in other episodes, 64 bullish and 34 bearish calls, past 30 days
Where everyone stands on Meta Platforms
Marcin Zdunek · F/m InvestmentsBearish
Meta, rated AA, pays 128-130 bps over Treasuries, an unexpectedly high borrowing cost.
Meta is rated AA yet pays 128 to 130 basis points over the ten-year Treasury, which Marcin calls unbelievably expensive for a company of its quality. He expects that cost to increase, reflecting the broader rise in borrowing costs for AI-related issuers.
“Meta is [an] AA rated company, and it's paying 128 or 130 basis points over a ten year treasury, which is really unbelievable how expensive it is for them to borrow”at 14:59
Anthropic
Across the shows: 142 speakers in other episodes, 251 bullish and 116 bearish calls, past 30 days
Where everyone stands on Anthropic
Marcin Zdunek · F/m InvestmentsBearish
Anthropic will likely be rated high yield once it starts issuing debt.
Marcin expects Anthropic to be rated high yield, or junk, when it begins issuing debt. He says this would limit investor capacity because there is not enough high-yield capital to absorb large issuance. The view rests on Anthropic's expected rating, not on any specific financial data in the discussion.
“And Tropic [Anthropic] will very likely be a high yield when [it] starts issuing debt”at 26:31
Credit markets
Where everyone stands on Credit markets
Marcin Zdunek · F/m Investments
Credit default swaps lead corporate credit moves, so they're a useful gauge for corporate bond ETF decisions.
CDS are insurance against default, and their cost reflects institutional views of credit risk. Marcin says CDS typically blow out or tighten before the actual corporate bond market moves, so they can give early warning. Because they are widely available, investors can use them as an indicator when deciding whether to hold or trade corporate bond ETFs.
“CDSs are the leading indicator of credit risk, right? You see them always kind of such a blow up or tighten ahead of what happens in the actual credit market”at 13:13
Alex Morris · F/m Investments
A broader credit problem would show as ratings downgrades, missed coupons and workouts.
Rating agencies are not shy about downgrading weaker companies, so a downgrade wave would be a real warning sign. Missed coupon payments and requests for workouts would be the next signals. Alex says the newer issuers such as SpaceX have not been around long enough to miss payments yet, and a loss of capital from those spending the most would also signal trouble.
“I think you'd see a problem if all of a sudden the ratings started to drop.”at 16:06
Alex Morris · F/m Investments
AI credit stress is not endemic; high yield defaults remain contained and the market is healthy.
Alex says the stress shown is specific to AI-linked names rather than a sign of trouble across industries, sectors or the wider world. High yield is holding up with no runaway defaults, and rates have ticked up only modestly from low levels. He concludes that the first cracks in fixed income are not a broad credit problem.
“So far, things are pretty good even in the high yield space. We don't see defaults running away from us.”at 17:48
Marcin Zdunek · F/m Investments
Rising new-issue concessions mean investors get paid more, but AI bond funds get more volatile.
As AI issuance grows, concessions and coupons rise, so bond fund income should increase. The trade-off is that the AI-heavy share of the index adds volatility to bond funds. Active and equal-weighted funds should cut some of that index concentration risk.
“you certainly will get paid more. Right? These concessions, and to the issuance, and the coupons will be higher”at 20:40
Marcin Zdunek · F/m Investments
AI-related credit weight in passive corporate ETFs will rise from about 12% to around 16% by 2027.
Marcin says the AI share of passive corporate ETFs was in single digits in 2025 and is about 12% now. He expects it to reach roughly 16% in 2027. The result is higher volatility in bond funds but also higher yield income, while active and equal-weighted funds can offset some of the index concentration risk.
“certainly for passive ETFs, the allocation to AI related credits will be in low to mid teens. We expected that in 2027, that allocation to AI credits will be around 16% or so, versus 12% currently”at 21:02
Alex Morris · F/m Investments
Equal-weighting bond funds avoids over-weighting heavy debt issuers, which cap-weighted indexes do.
In cap-weighted bond indexes, the more debt an issuer sells, the bigger its index weight, which concentrates exposure in heavy borrowers. As AI issuers grow, this effect increases. Equal-weighting sectors gives a truer cross section and steadier returns, though Alex admits this has not always helped in equities, and adds that more debt is not necessarily a good thing.
“They get sort of cap weighting. So the more debt you issue, the more of the index you take out.”at 23:08
Marcin Zdunek · F/m Investments
Investor demand can absorb AI debt if concessions rise, but a SpaceX downgrade to high yield changes everything.
Consensus from a Goldman capital markets call was that rising new-issue concessions would draw enough buyers, as long as issuers stay within the BBB-plus range. A downgrade to high yield would shrink the buyer base because little capital is allocated to high yield. Marcin notes the risk that capital could leave emerging markets, high yield or Treasuries if issuance continues at this pace.
“subject to rating changes, right? Like if SpaceX drops to, I'm not suggesting it will, but if it drops to high yield, obviously that changes everything, right? Because there's just not enough capital allocated to high yield to absorb it”at 26:19
Alex Morris · F/m Investments
New-issue concessions for AI bonds are above non-AI levels and will keep rising until issuers balk.
AI-name concessions have been higher than for non-AI issuers, which means investors get a built-in discount for buying new AI bonds. The trend has been up, with ebbs and flows. Alex expects it to keep growing, until issuers find the concession too costly and underwriters refuse the deal. The concession rising also raises the coupon burden on the issuer over time.
“you see that it's just higher than the average for non [AI] names. So you're getting this sort of discount built in. This will continue to get bigger”at 27:14
AI infrastructure & data centers
Where everyone stands on AI infrastructure & data centers
Marcin Zdunek · F/m Investments
Only about 30% of 2027 AI capex will come from free cash flow; the rest must be externally financed.
AI capex that was largely funded from internally generated cash flow is now too large for that. Estimates for 2027 are around $1.5 trillion across the five biggest hyperscalers and OpenAI, with roughly 30% covered by free cash flow. Everything else has to be financed through public or private capital markets, and that excludes whatever Anthropic and OpenAI themselves will raise.
“And the estimates for 2027 are, you know, aggressively around [$1.5 trillion], of which, you know, 30% or so will be done through free cash flows. Everything else will have to be financed.”at 3:34
Alex Morris · F/m Investments
AI-related debt is now priced as riskier than when the borrowing began, unlike bank debt.
Banks have historically issued more debt and still been judged creditworthy, but the market is now decoupling AI-related borrowers from that pattern. Their debt is priced as inherently riskier than at the start of the cycle, and some AI companies look like they may be overextended. Alex says the AI names are not all equal in credit quality, and the spread gap shows the market recognizes that.
“you can see the decoupling that's starting to happen as AI related debt is inherently riskier today, according to the market, than it was at the beginning.”at 12:12
Alex Morris · F/m Investments
Companies are exhausting cheap US dollar debt and turning to foreign markets, which will run out.
Concessions keep rising and order books fill slowly at the desired levels, showing the limits of investor demand. To find new buyers, AI-linked borrowers have issued in Singapore, Canada and EU markets rather than only in dollars. Alex expects this capacity to run out eventually, which points toward other financing sources.
“So many of these companies have gone. Singapore, Canada, EU debt, you know, across the way in an effort to find fresh buyers. They will eventually run out of that capacity.”at 30:36
Marcin Zdunek · F/m Investments
Credit markets will push AI capex onto equity holders as debt becomes too expensive.
If new-issue concessions keep rising, issuing bonds becomes too expensive, so companies may reissue equity or do add-on equity deals to fund capex. Marcin says equity issuance is generally positive for credit because it reduces leverage. He suspects the credit market will force equity holders to bear the capex and its cost, though he hedges that he will have to see how it plays out.
“that the credit market is going to force the equity market to be the bearer of CapEx and that cost”at 32:26
Marcin Zdunek · F/m Investments
AI debt is becoming costlier and competes with US Treasuries for the same investor capital.
Marcin says there are a lot of similarities over the last year or so between AI debt and US government debt. Both have seen borrowing costs rise. He notes that they now compete for the same pool of capital, and he asks what will happen as a result. He does not give a specific forecast, and frames it as something that will be interesting to watch.
“Certainly a lot of similarities in this... In the last year or so between AI debt and US debt. Right? Like, you can see those, That there's the cost increases”at 38:57
US Treasuries & yields
Where everyone stands on US Treasuries & yields
Alex Morris · F/m Investments
The AI wobble pushed Treasury yields lower as money fled AI deals back into Treasuries.
When AI deals stumbled, Treasury rates came down. Alex says Treasuries were not more certain that week; the capital that had gone to chase AI deals returned to Treasuries as the safe asset. This shows that AI-related financing competes with Treasuries for investor capital.
“And the interesting side effect of the AI wobble a few weeks ago was treasury rates came down.”at 30:56
Alex Morris · F/m Investments
A failed Treasury auction is practically impossible because primary dealers would step in.
Investors fear that a failed auction would signal the limit of debt capacity. Alex says there is no mechanism for a failed auction in America today because primary dealers are required to step in and bank CEOs are asked to bid. He expects auctions to finish successfully, which means the fiscal limit is not reached through auctions.
“Practically speaking, there's no mechanism for an auction to fail in America today. We have the primary dealers who have to step in”at 36:24
Alex Morris · F/m Investments
Lower Treasury yields could come from cutting issuance, though inflation as a tool would cause problems.
Alex says he would not be surprised if the amount of Treasuries the government issues simply came down. If issuance fell by a third, yields would also fall, because demand would be roughly equal for fewer bonds issued each quarter. He sees this as a plausible route to lower long-term rates that the president wants for mortgages, though he treats it as uncertain and distinct from the inflation route.
“I wouldn't be surprised if you saw the amount of tenure we issue just simply come down. If it were to reduce by a third, certainly, the interest rate that it had would also come down”at 37:43
Marcin Zdunek · F/m Investments
Two-, three- and five-year Treasuries stay profitable unless yields sell off about 400 basis points.
Marcin argues the high coupon on the two-year Treasury gives a cushion against rate moves. The two-year bought at about 4.87% still makes money if yields rise 100 basis points, and returns about 2% even after a 300 basis point sell-off if held to maturity. He says a sell-off of roughly 400 basis points is needed before owning the 2, 3 and 5-year Treasuries loses money. The return profile is far less favourable at the 10 and 30-year end, where the risk increases dramatically. He frames this as a roughly twelve-month play.
“you really would have to see a significant sell off of 400 basis points or so to start losing money owning a two, three and five year treasury.”at 40:00
Alex Morris · F/m Investments
Taking more duration is worth it; falling rates would bring large gains that outweigh modest losses if rates rise.
Alex says F/m has encouraged investors to add duration over the past year. A 50 basis point rise in rates would cost even 30-year debt only about 1.4%, while a 50 basis point fall would be worth about 13%. He argues that T-bills pay well today but that cash has been a poor asset in past periods, such as the money market breaking the buck a decade ago. He argues that if investors want to make money they must take duration risk, and says F/m's longest-duration products have grown fastest. He expects rates to come down, but does not give a date.
“There's a lot of upside to rates coming down.”at 44:33
Asset managers, brokers & exchanges
Where everyone stands on Asset managers, brokers & exchanges
Marcin Zdunek · F/m Investments
Bond ETFs reinvest income monthly, letting investors buy in at higher rates when yields sell off.
Marcin argues that a fixed income ETF's monthly income reinvestment is a key advantage over holding cash or waiting for semiannual coupons. If rates sell off, investors can use the monthly income to buy into the higher yields rather than waiting six months. He sees this as a positive offset to the inflation risk of owning bonds, which also affects rates on the 10 and 30-year Treasuries.
“you reinvest that income monthly, right? So you don't have to wait six months to reinvest it. So even if you sell off and your rates are selling off, you are buying in”at 47:26
Alex Morris · F/m Investments
Bond ETFs gave liquidity when the bond market seized in 2020 and trade bonds more cheaply than retail investors can.
Alex argues that in March 2020 the bond market seized up, so the liquid instruments were ETFs, and ETF prices drove net pricing more than bond prices drove ETFs. He says ETFs work as a team sport: market makers now act as bond dealers, price baskets of thousands of bonds several times a second, and can resell bonds through multiple outlets. This added liquidity. He concludes investors can trade bonds inside an ETF far more efficiently and cheaply than they could reconstruct the basket themselves. He also argues portfolio trading is often cheaper than trading bonds one at a time.
“Bond market seized up a bit. So the thing that was liquid were ETFs.”at 57:00
Marcin Zdunek · F/m Investments
Bond ETFs give small investors better trade execution than they could get buying bonds individually.
Marcin says large institutions have big trading desks and get decent executions, but retail investors and small shops do not. For them, fixed income ETFs are better because they outsource trading to market makers such as Jane Street and Citadel, who profit when creations and redemptions are done at the best levels. He says he has seen T-bill ETF prints inside the cost of buying bills outright, all the time. He says the ETF is a superior product to anything an individual can achieve buying individual CUSIPs.
“But if you're a retailer or even a small shop, there's really nothing better than investing in fixed income ETFs because you are outsourcing your trading”at 1:00:15
US deficit & government debt
Where everyone stands on US deficit & government debt
Alex Morris · F/m Investments
Medicare and Social Security trust funds are projected to run dry by the mid-2030s.
Alex says US national debt is over $40 trillion and the government's ability to control spending is weaker than a decade ago. Medicare and Social Security trust funds are on track to be exhausted by the middle of the 2030s, so programs at current size would have to be paid for out of taxes and tariffs. He views this fiscal path as the biggest concern for Treasuries.
“We're looking at by the, say, middle of the twenty thirties, Medicare and Social Security trust funds running dry.”at 34:26
Alex Morris · F/m Investments
US government's borrowing costs are rising from heavy spending, weak restraint and self-inflicted mismanagement.
Alex says the government kept a shorter debt maturity profile when it could have termed out at 30 years, so it must refinance constantly, like a mortgage refinanced every two or three years. He says it is already having a hard time refinancing its debt at current levels, and that the auction backstop from primary dealers only gets it done at an unknown price. He attributes the higher cost of borrowing to a lot of spending and few restraints. If that continues, he expects rates to keep rising until the government takes aggressive action such as yield-curve control, as in Japan. He frames it as hard to run a country that expected to finance everything at 3% and now pays 6%.
“and sadly, the government's ability to borrow has just become more expensive. And a lot of it's self inflicted mismanagement”at 38:21
US inflation
Where everyone stands on US inflation
Alex Morris · F/m Investments
Short-duration TIPS protect cash from inflation better than cash, which loses purchasing power permanently.
Alex argues that long-duration TIPS funds lose money in inflation spikes because the Fed raises rates, and rate-driven losses from only two to three years of duration cancel out the inflation protection. He says that stripping out most duration and holding TIPS with about one year of duration gives real inflation protection. F/m built an index with Bloomberg. Compared with a roughly six-month-duration cash benchmark, the short TIPS product delivered headline CPI when inflation rose from 2020, while the Fed took a long time to catch up. He says the y-axis, the lost purchasing power, is what matters, and that cash losses are permanent.
“the tips product did exactly what it promised you. It delivered all that sort of headline CPI to you, whereas it took a long while for the Fed to come to the party.”at 50:57
Alex Morris · F/m Investments
Inflation is stubborn and generational, with CPI and core PCE running well above the 2% target.
Alex says both CPI and the Fed's preferred PCE measure are well above 2%, and that core CPE and CPI are running hotter than desired. He argues inflation is going the wrong way in many respects and will not be solved quickly. He ties this to voter anger over grocery prices in recent midterms. He expects investors to need to learn to live with it for a long time.
“Inflation is not gonna be solved tomorrow. It's a generational issue.”at 52:09
Private credit
Where everyone stands on Private credit
Alex Morris · F/m Investments
Private credit is opaque but its covenants reduce risk; more data would help judge AI exposure.
Private credit has far less public data than traded debt, so Alex says it is harder to see what is going on in the loans. Those loans, however, carry more covenants, which give lenders more room to work out problems and reduce risk. He would be able to say more if more of this market were visible, and he believes the dispersion among names is greater than it looks.
“One, we don't get all that much data on what's going on... But those loans also tend to have many more covenants with.”at 18:22
Venture capital & startups
Where everyone stands on Venture capital & startups
Alex Morris · F/m Investments
Venture deals financed with hidden debt will produce wipeouts that have not yet been priced.
Large players have done a huge number of venture deals, more than many venture funds do over their lives. Alex says the debt and other instruments used to fund those investments have not yet come to light. He expects some of it to surface and produce losses, which are not reflected in public information.
“eventually, a lot of that will come to light. I hope it's in a positive way, but I think you'll see some wipeouts”at 19:59
US rates & the Fed
Where everyone stands on US rates & the Fed
Alex Morris · F/m Investments
The Fed will struggle to raise rates enough to offset worsening inflation.
Alex says the Fed has done a good job leveling inflation out, but if inflation gets worse the Fed has said it will raise rates only a little. He thinks inflation could get much worse than that, so the Fed's rate increases would not be enough to compensate savers and investors for lost purchasing power. He suggests long-term cash allocations as a result.
“I think the Fed's gonna have a hard time raising rates enough to compensate you.”at 52:22
Across the shows
What other episodes Chorus covered said about the same companies in the 30 days up to this one.
- Anthropic142 speakers · 251 bullish, 116 bearish calls
- Meta Platforms66 speakers · 64 bullish, 34 bearish calls
- Space Exploration Technologies55 speakers · 67 bullish, 26 bearish calls
- Microsoft47 speakers · 50 bullish, 12 bearish calls
- Oracle32 speakers · 10 bullish, 30 bearish calls
Topics
- AI borrowers' credit risk
- AI debt issuance
- AI share of corporate bond indexes
- CDS as credit indicator
- CDS pricing on AI borrowers
- Fed response to inflation
- TIPS as cash inflation protection
- Treasury auction risk
- Treasury demand vs AI debt
- Treasury supply
- bond ETF income reinvestment
- bond ETF liquidity
- bond ETF trading costs
- bond index construction
- bond spreads on AI-linked issuers
- credit crisis warning signs
- credit market stress breadth
- credit ratings
- debt financing capacity limits
- duration positioning
- entitlement trust fund solvency
- equity financing of AI capex
- government borrowing costs
- hyperscaler capex financing
- inflation persistence
- investor demand for AI debt
- new-issue bond spreads
- new-issue concessions
- private credit transparency
- short-end Treasury returns
- venture-linked debt risk
More from Excess Returns
- The Fed Is Hiking. Earnings Growth Is Slowing. Is This 2018 Again? | Barry Knapp · November 18, 2026
- From Poker Pro to Elite Hedge Fund Manager: David Orr on Investing, Risk and AI · November 17, 2026
- Show Us Your Portfolio: Meb Faber · November 15, 2026
- Fidelity’s Jurrien Timmer on AI Spending, Bond Yields and the Next Phase of the Bull Market · November 13, 2026
- Stock Investors Fear an AI Bubble. Bond Investors Fear Inflation. How Can You Diversify Without Selling Them? · November 12, 2026
Ask your AI about this episode
Chorus gives Claude, ChatGPT and other assistants every view from this episode and every other show Chorus follows. Ask who agrees, who changed their mind, and what else was said about the same companies.
Chorus reports what speakers said. It isn't investment advice. How Chorus works · About Chorus