The shadow-lending debate: whether the boom in private credit is a durable asset class or the place the next blowup starts.
The count, last 30 days: 5 bullish, 8 bearish. The tape leans bearish. A count, not a verdict.
The tape
Jul 06
Jul 20
Jul 27
Aug 10
Aug 17
Sep 07
bullishbearishclaims filed per week
What the people who move markets are saying about this, in their own
words. These are views on the record, not scored bets: when a call here grows a number
and a deadline, it moves to the scoreboard. Newest first.
Neutralleaning · 2026-09-10
Jack Caffrey tells the anchor a corporate credit reckoning is deferred in the short term because treasurers termed out debt at very low 2021-22 rates, though shorter rates are starting to bite.
From the shorter term, I think it becomes a case that the piper is going to be deferred.
Miriam Wheeler (Goldman Sachs) tells Bloomberg the heavy post-summer leveraged-finance pipeline (~$40bn Europe, similar US) will be absorbed efficiently given sidelined cash.
we think that it will be absorbed efficiently both here and in the United States.
Wheeler says ~$80bn of US non-IG AI issuance is done year to date and roughly $60bn more could come by end-2026, the market's biggest wild card.
if you think about U.S. AI issuance in the non-investment grade market year to date we've seen almost 80 billion dollars. When we look at the pipeline of AI issuance between now and the end of the year in the U.S. that could be another 60 or so to go
Wheeler says credit spreads sit within a few bp of multi-year tights but flags unknowns: software, global rates and AI supply volume.
they have been incredibly resilient. We have spreads within a couple basis points of their multi-year tights just speaking to the resilience of the market but there are a number of unknowns out there.
Wheeler says most 2026 AI non-IG deals now trade below issue price as heavy supply and rates weigh, though she thinks supply will still be absorbed.
we have seen some spread widening particularly in the A.I. space. Most of the deals that have been originated this year are now trading below their new issue price.
Wheeler says European AI-related non-IG issuance is muted near term but $5-10bn could print by end-2026 if mega projects get signed.
over the next couple of months in Europe it's probably somewhat muted. But I think that you could see you know five ten billion dollars happen maybe between now and the end of the year if we start to see some of these mega projects get inked.
Wheeler says AI-related debt issuance in Europe, including hyperscaler non-US IG deals, is mostly a 2027 story, likely if the capex build-out continues.
So I think it's really more of a 27 story for Europe. But we do think that it's likely coming if you see this capex build out in Europe continue.
Rep. Vince Fong tells Fox that California high-speed rail will never be finished and will waste $1bn a year until then.
The state of California is going to waste $1 billion every single year now until this thing is, quote, unquote, completed. But it's not going to be finished in our lifetime. I don't think it ever will be.
Scott Berg of Needham tells Bloomberg the biggest risk to a Workday buyout is whether private credit can fund the ~$20-24bn of debt required.
can they raise what's ultimately probably 20 to maybe twenty four billion dollars worth of debt required to get this leveraged bio done. I'm not saying they can't but that's I think probably the biggest concern that the market would have
Yung-Yu Ma (PNC) tells CNBC that corporate debt issuance is at record levels, far surpassing last year's record, adding to Treasury supply hitting capital markets.
Corporate debt is at record-breaking levels. Last year was a record level. This year is far surpassing last year's record levels.
Mitchell Green says the scary difference from the telecom bubble is the explicit and implicit leverage (debt, circular Nvidia/Oracle IOUs) financing AI capex.
What's a little scary this time is the amount of leverage being put on this system and the amount of implicit leverage.
Jay Goldberg of Seaport tells Prof G Markets he strongly suspects Nvidia is backstopping some portion of the $500bn, making it effectively Nvidia debt.
But my strong suspicion is that Nvidia is backstopping, providing some form of guarantee, not for the whole $500 billion, but some portion of it.
Jay Goldberg of Seaport tells Prof G Markets that lenders will not accept GPUs as collateral as Jensen Huang pitches, so Nvidia will end up taking on obligations.
I personally don't think that's going to land with investors. I think they're still going to want to see some other forms of guarantees. And that's ultimately why NVIDIA is going to take on some form of obligation here.
Bill Campbell (DoubleLine) tells Bloomberg credit spreads are at historic tights across markets, so his firm stays up in quality and finds only idiosyncratic pockets of value.
credit spreads are at pretty much historic tights no matter where you look in the markets.
Karen Fang (Bank of America) tells Bloomberg the US needs trillions of dollars of critical infrastructure investment, far more than BofA's $250 billion 18-month initiative.
Well, I think we need trillions of dollars for our nation's critical infrastructure development
Karen Fang (Bank of America) tells Bloomberg that about 90% of new electricity added to the US grid has been renewables and storage because they are cheap and fast to build, while gas and firm nuclear/geothermal are still needed.
we actually have seen 90 percent of the new electrons added to the United States power grid being renewable energy and storage.
Karen Fang (Bank of America) tells Bloomberg she is pretty confident BofA will deploy its $250 billion infrastructure commitment within the 18-month window, framed as a stretch goal to be revised upward.
So in a year when we reveal our 18-month results, we're pretty confident we can accomplish this
Thomas Hayes (Great Hill Capital) tells David Lin that lending against 2026-vintage Nvidia GPUs is not secured lending because the chips will be worth less in seven years, and he expects the securitization story to unravel around five years out.
You cannot make the case. If I lend on a 2026 vintage Nvidia GPU, that in seven years, it's going to be worth more than it's worth today. And I'm going to get my money back.
if AI spend pauses, levered smaller suppliers face debt distress
“the Nvidia is in the Microsoft world are going to be just fine based upon their core businesses these other companies might have a really hard time servicing their debt”
hyperscaler financing moved from equity to debt. STANCE corrected bear->neutral: description, and he leaves open whether spread widening is worry or supply
“to now being on the debt side of the balance sheet”
Explaining Goldman's overweight equities, neutral bonds, underweight credit allocation. He argues 'sovereign bond yields, in the medium term, have to find an equilibrium that's higher' @30:59 and sees 'certain late cycle tendencies. There's re-leveraging, re-leveraging that's linked to the AI ecosystem. There's restructuring, M&A, all of that tends to lead to worsening credit quality.' @31:07. Bonds' portfolio role is shifting toward 'generating yield and to some extent less protecting the portfolio' @30:43.
“But credit spreads are ultra tight right now. So we just don't see the fixed income complex currently as very attractive.”
Discussing a UBS-proposed / Partners Group-executed structure that wraps private-credit fund stakes in Nationwide insurance to lift ratings to A2 (from BBB/B), Carpenter draws GFC parallels; he hedges that 'there's a lot less of it going on than before' but says it's 'worth paying attention to.'
“There's so much money sloshing around that says to me there's going to be a lot of deals that don't deserve that top rating of A2.”
Pitching the new $15B fund (50% larger than predecessor): capital concentrating in the AI buildout 'leaves some other parts of the infrastructure universe' underfunded (@60:35). Notably declines to endorse the 'froth' framing on large-cap AI infra when asked (@62:21); cites NY data-center moratorium as demand evidence for its 1GW dispatchable power asset.
“Overall about one and a half trillion of investment needs in in the coming decades really ripe for uh for private capital investing”
Responding to the host citing Morgan Stanley estimates of hyperscaler debt issuance rising from about $100 billion last year to $130 to $150 billion, and total AI-linked issuance possibly hitting $570 billion this year: 'You're a hundred percent right' @3:37, since capex once funded by internal cashflow now taps equity and bond markets; 'we have to be very mindful of how concentrated parts of our portfolios in certain segments are becoming to this AI trade' @4:03.
“This is simply exacerbating the issue with respect to concentration risk. If you take a look at issuance in the bond market over the course of this year, a meaningful portion of that is related to AI.”