Alberto Gallo says the world may have already hit some icebergs, and his hedge fund is short the managers, the insurers and the data centers behind the AI credit boom.
Most credit strategists point to tight spreads and years without a default as reasons for calm. Gallo, whose fund's stated aim is the best risk-adjusted returns in credit, reads the same market as a Jenga tower of government, corporate and private debt with a new, unstable block stacked on top.
"So the iceberg is already, touching the boat."
Gallo co-founded Andromeda Capital Management after running global macro credit strategy at RBS and managing the Global Credit Opportunities Fund at Algebras, and he says the fund is short the AI-linked credit trade across managers, insurers and data centers alike.
I listened to the full interview so you can skip it. 55 minutes of audio, 24 minutes of reading.
Here are the 9 takeaways that matter.
👤 Guest: Alberto Gallo, Chief Investment Officer and co-founder of Andromeda Capital Management, formerly head of global macro credit strategy at RBS
🎙️ Hosts: James Crombie, senior editor at Bloomberg, and Stefan Kovacev, senior analyst at Bloomberg Intelligence
📰 Published: 10 September 2026
🟣 Apple Podcasts | 🔗 Episode page | ⏱️ 55 min | ✅ Time saved: 44 min
Key Takeaways
Governments face an "impossible trinity" of a strong currency, no austerity and controlled yields — something has to give
Gallo's base case is currency debasement first, then credit
Japan, not the U.S. deficit, is his real worry on rates
If the BOJ hikes and Japanese savers stop buying Bunds and Treasuries for JGBs, that pressures the whole developed-market curve
Private credit has returned about 8% net over four years — similar to public high yield, but locked up for five years with no exit
Hyperscalers need to roughly double their revenue to break even on AI capex, refinanced at 5-7% yields over 10 years
Andromeda is short the managers, the insurers and the data centers it likes least, through equity, bonds and CDS
U.S. investment-grade and high-yield positioning sits at the longest in 15 years of DTCC data, with spreads at 30-year lows
10% of private credit is now "pay in kind," which Gallo calls the "Schrödinger's default" — a hidden loss that only crystallizes on a mark or a refinancing
His biggest long-term worry is a reform of the monetary system, not a credit crunch itself — new tools that quietly shift losses onto savers
He still likes Ukraine sovereign and corporate debt, betting on eventual reconstruction and EU accession
1. The impossible trinity
James Crombie opened by asking where rates go from here, with 10-year yields near 5% and 30-year yields being talked about at 6%.
Gallo reached for the Titanic. "We might have already hit some icebergs with government bond yields going so high," he said, comparing it to the two hours of calm after the ship struck ice before passengers began to worry
He said Treasury Secretary Scott Bessent is trying to hold three things at once that cannot all hold. "He's trying to keep the dollar strong or at least afloat," avoid austerity, and keep yields under control. "But at some point, if you are a government with already very high debt and you keep your deficit at wartime levels, you have to let go of one of the three sides of this triangle."
His base case is currency debasement first, credit stress later. Letting long-term yields rise risks a credit crunch, especially given "the credit and the AI borrowing that's happened"; letting the currency fall hits voters already squeezed by inflation. "Ultimately, something has to give, and we think that initially it will be the currency"
2. Japan is the real risk
Crombie pressed on the velocity of the rate move — whether 10-years hit 5% and 30-years hit 6% soon.
Net interest costs are on a clear path higher even without a fresh shock. At current rates, federal net interest outlays climb from about 3% of GDP toward "over 4%" over the next few years
His real worry is Japan, which has anchored global yields for decades. Japanese inflation is climbing back toward 2%, and Gallo doubts the Bank of Japan can keep hiking "very slowly" indefinitely. "Eventually, one central bank, and I'm worried about the BOJ in particular, might have to hike a little bit faster"
If Japanese savers stop buying Bunds and Treasuries in favor of JGBs, he said that would put even more pressure on both
He framed it as a global, pro-cyclical spending race, not a U.S. story alone. Germany is spending after 23 years of restraint, Japan is spending amid its own inflation, and the U.S. keeps spending — competition for commodities and technology across governments simultaneously
3. One hike, sticky inflation
Crombie asked directly whether the Fed hikes, and how many times.
Gallo's base case is one hike each from the Fed and the Bank of Japan — less than the situation calls for. "Our base case is that central banks will still be leaning dovish. So they'll hike, but they'll hike less than needed"
He rejects the word 'transitory' a second time, having first pushed back on it during COVID. Inflation has run above target for close to six years, which he called "a global switch to pro-cyclical fiscal policy" that started under the first Trump administration and continued after. "The U.S. government never had a 7% deficit with unemployment, below 4%."
High headline yields don't mean real returns, once hedonic adjustments are stripped out. A 10-year Treasury near 4.8% still leaves investors in "an environment of financial repression" once true inflation — not the BLS's quality-adjusted version — is counted. "You need better returns... that are linked to real assets, not only paper assets"
4. Who's on the sinking boat
Crombie returned to the Titanic image and asked who the passengers actually are.
The passengers, in his telling, are government-bond investors in states that import commodities and are losing the ability to tax. He named the U.K. specifically — a service economy importing 50% of its goods, whose competitive edge relied on low taxes and limited competition that "no longer works" as the rest of Europe wakes up
Western economies have spent decades buying commodities with better technology rather than building an industrial base, and China has taken up to 30% of global goods production as a result. "Unfortunately, China is winning slowly," owning both commodities and manufacturing share
Governments are choosing stimulus over reform. "It goes a lot into financing consumption, into helping demand, into making the economy kick the can until the next election," rather than into productivity
He reached for the late Roman Empire. Debasing currency to pay soldiers eventually cost Rome its legionaries' trust — "there is a slow erosion of credibility," which he linked to the dollar's declining share of global reserves and said the Iran war "might have accelerated" it
5. The Jenga tower of debt
Crombie asked what all this means for corporate credit markets, where issuance is running at record levels and spreads are very tight.
He sketched the market as a Jenga tower: roughly $100 trillion in government bonds at the base, then emerging-market and investment-grade debt, then high yield, then a new block on top — $3 trillion in private debt
Private credit has returned about 8% net of fees over the last four years, he said, using data from Apollo and other firms — comparable to public high yield. "So you're locking your capital over a five-year period without the ability to get out"
Two forces pushed capital into that top block simultaneously. Private equity firms moved into credit to boost the IRR on their LBOs — three-quarters of private credit holdings are LBO firms — while software and AI companies needed capital for a capex boom; IMF data puts over 40% of private credit exposure in tech and software, much of it negative free cash flow
The AI build-out is now the second-largest sustained expense in U.S. history, behind only the Louisiana Purchase and ahead of the 19th-century railroad build-out, at 2.8% of GDP a year. Hyperscalers now account for about 10% of investment-grade issuance, second only to banks
He called the setup "heads I win, tails you lose" for credit investors lending at 200-300 basis points to capture roughly 8% returns in what has so far been the best possible environment — four years of low rates, tightening spreads and no defaults
6. The math on breaking even
To break even on AI capex, Gallo said hyperscalers will need to roughly double their revenue. Cumulative capex over 10 years comes to about $10 trillion once refinanced at 5-7% yields, against current hyperscaler revenue of "below 2 trillion in revenue, 1.9 trillion in revenue." "I'm sure they can do it, but I'm not sure they can do it in the next two, three years"
He does not think the trigger is a worse model from a rival — that is just one of several risks. The bigger one is credit investors realizing that a portfolio of ten bonds paying 3% each loses three to four years of carry the moment just one defaults, a payoff shape with none of equity's upside optionality
10% of private credit is now "pay in kind," which Gallo called the "Schrödinger's default" — companies without enough cash flow to pay a cash coupon that stay technically current only because lenders keep extending. "There's already some losses that are hidden in the system"
He said this cycle's investors learned 2008's lesson and hid the risk better, not less. Life insurers — the largest buyers of private credit, some of them owned by private-equity shops such as Athene and Global Atlantic — hold assets with no mark-to-market, and a paper by University of Texas law professor Andrew Granato found some of those losses could ultimately land on taxpayers through state-level policyholder guarantees
7. Short everything
James Crombie asked directly where Andromeda sees the best risk-adjusted opportunity in AI-linked credit, and whether the fund avoids the space entirely.
"We're short everything," Gallo said — the managers, the insurers, the data centers. "Not all of them, but the ones that we like the least."
He questioned why hyperscalers let off-balance-sheet vehicles like CoreWeave borrow at 500 basis points instead of funding data centers themselves at 50. Keeping leverage off their own balance sheets makes it easier to restructure an external vehicle than risk a downgrade on their own debt
The short case grew out of watching companies quietly shift bonds from public to private markets — often ones with weak disclosure, high leverage and wide spreads that couldn't fund publicly, a pattern Gallo said flagged where excess private-credit capital was landing
The trade uses equity, bonds and CDS together, built to be "positively convex" rather than a simple bearish bet. U.S. investment-grade and high-yield positioning, per DTCC data on the CDX indexes, sits at its longest in the 15 years of the dataset — the trade "has worked a bit in Q1" but "hasn't worked after April" as central banks kept rate volatility low
8. Where the relative value is
Stefan Kovacev and Crombie asked where Andromeda actually wants to be long.
Europe over the U.S., despite the U.S.'s stronger backdrop if growth holds. European credit spreads are wider than U.S. spreads "across every sector," which Gallo attributes partly to home bias — 80% of the top 100 asset managers are U.S.-headquartered
He favors companies with pricing power on real assets — steelmakers, energy producers, miners — outside the U.S., and said the same home-bias premium shows up in emerging markets: Ukraine's sovereign spread trades three to four times wider than Ecuador's or Argentina's despite comparable risk
Andromeda has held Ukrainian sovereign and corporate debt since the teens, betting on eventual EU accession and reconstruction demand tied to the country's 15% share of global corn and wheat production and roughly 10% of European steel and titanium output
His broader long book leans on real assets: convertibles, infrastructure, defense, banks, energy, mining — sectors that preserve value if currencies depreciate, plus gold, while staying short "paper assets" and short private credit
9. Defaults, Warsh, and gold
Crombie closed by asking about default rates into 2027 and what worries Gallo most.
He expects a gradual rise in corporate defaults from roughly 3-4% today to 5-6%, mostly restructurings rather than a spike. "Are we seeing a recession? Are we seeing a spike in defaults? No, not right now" — liquidity and fiscal spending are still there
He compared Fed Chair Kevin Warsh to Bank of Canada's Mark Carney: polished, competent, more talk than action so far. "There is more talk than action on the hawkish side," and the unemployment mandate, in Gallo's reading, still comes first
His real worry isn't a credit crunch — it's reform of the monetary system itself. He pointed to a New York Fed paper on revaluing gold reserves, proposals to sell 100-year bonds to allies, countries repatriating gold from the New York Fed, and talk of stablecoins and tokenization as early signs of an attempt to "re-anchor the dollar standard"
His one indicator to watch is Japanese inflation, alongside any new policy tool from the Treasury — "new initiatives, any new tools, any new toys." He recalled the Fed buying credit ETFs directly during COVID — what Andromeda called 'QE infinity' at the time — as evidence that "it's really hard to get out of stimulus"
Bonus Insights
He named Milton Friedman on the durability of stimulus. "Milton Friedman said that there's nothing more permanent than a temporary government program," Gallo said, adding that it is true of both fiscal and monetary stimulus alike
Andromeda's portfolio is built as a barbell, not a single bet. Citing Nassim Taleb's Antifragile, Gallo said a mix of cash and bonds carries the same yield as a fully invested book but more optionality — "when you have vol, you can buy more"
His read on where systemic risk is concentrated has shifted away from banks. "But if things go badly, this time around, the risk is not so much into the bank's balance sheets. It's not so much into Deutsche Bank or Unicredit. It's in the U.S. life insurance sector."
Gallo's bottom line is that debasement, not default, is the trade playing out right now — governments will keep leaning on low real rates to fund AI-era spending, but the private-credit block stacked on top of that debt is already showing cracks his fund is positioned to profit from.
Products, Companies & Tools Mentioned
Andromeda Capital Management (Gallo's hedge fund, whose stated aim is the best risk-adjusted returns in credit markets and which is short managers, insurers and data centers tied to AI credit)
Apollo (One of the private credit firms whose data Gallo cites for the roughly 8% net private-credit return)
CoreWeave (The off-balance-sheet AI data-center lender Gallo cites as a case study — its bonds are "trading 10 points down" after moving from unsecured to secured issuance)
SpaceX (Its bonds are also "trading 10 points down," in Gallo's telling, alongside CoreWeave's)
Athene and Global Atlantic (Life insurers owned by private-equity shops that Gallo names as large private-credit buyers)
Books & Resources Mentioned
Antifragile – Nassim Nicholas Taleb (The source of the barbell-portfolio framing Gallo uses to describe Andromeda's mix of cash, bonds and convex upside)
If this was worth your time, send it to someone who has to have a view on this.
Get the latest market chatter as it happens: