Gundlach Sees the Most Unhinged Treasury Market of His 40-Year Career, With AI Borrowers as the Hollow Branch in a Rotting Tree
DoubleLine's Jeffrey Gundlach and Rosenberg Research's David Rosenberg discuss bonds, the Warsh Fed and credit risk at a fireside event, October 18, 2026
Jeffrey Gundlach, founder of DoubleLine Capital, opened with an analogy from his house in Buffalo. A century-old silver maple that looked healthy and full of leaves shed a limb roughly 100 feet long. The family then learned the trunk was "completely hollow." Gundlach argued the market is in the same condition. Cap-weighted indices look sound, but David Rosenberg noted that 75% of S&P 1500 stocks are in a correction, with the median stock 10% to 15% below its 52-week high. Gundlach added that the large majority of S&P 500 members are down more than 21% from their highs, even with the index near a record. "You've got to wait for the branch to fall off to figure out that the market's hollow," he said. He sees the same dynamic in private markets.
AI Borrowers Are Where the Rot Shows First
The most concrete evidence Gundlach offered came from splitting credit indices into AI and non-AI components. Investment grade spreads excluding AI have not widened at all. AI-related investment grade spreads have widened about 75 basis points over the past year, and that figure was a few weeks old. In high yield, non-AI spreads sit close to their tights, while AI-related spreads are out more than 150 basis points, and Gundlach guessed the figure is now closer to 200. "Underneath the surface, it's the silver maple again," he said.
Rosenberg framed the same issue from the supply side. The average high yield yield is about 8%, and he noted that three of the four hyperscalers are now free cash flow negative. In his words, they are "running the same policy that the government is," financing deficits on their balance sheets. He asked at what point corporate borrowing costs break the back of that demand for capital, and he floated the possibility that a pullback in hyperscaler issuance could be what finally brings long rates down.
Gundlach was less optimistic about the timing. He expects the hyperscalers to keep borrowing "longer than most people think" because they are price insensitive, and that issuance will then go "to zero" when lenders stop. He called the crossover point "challenging to navigate." Rosenberg added that regulatory and public backlash against data centers, and a likely Democratic sweep on November 3, could accelerate a slowdown in AI-related borrowing.
Why Gundlach Thinks Long Yields Keep Rising
Gundlach said long-end yields are being pushed up by three forces: a war-driven surge in diesel and Brent crude that "is always about to end" but never does, a "monumental" volume of bond issuance across the developed world that AI financing now compounds, and an interest expense problem with no political response. He put interest expense at roughly $1 trillion and the annual deficit at over $2 trillion, with maturing debt being refinanced at higher rates. Treasury Secretary Scott Bessent's argument that the US can grow its way out struck him as unsupported, since debt has outgrown GDP for decades.
One data point stood out. The Fed cut 175 basis points between September 2024 and January 2026, and every maturity on the Treasury curve ended higher than before the first cut, by about 100 basis points across the curve. "That's never happened before," Gundlach said. He argued it corroborates his view that the playbook built during the secular decline in rates from 1980 to 2020 no longer applies. He put the 30-year at about 5.62%, a 24.5-year high, and said that is not yet high enough to prompt the government action he anticipates.
He identified two possible resolutions: printing money, or restructuring the debt through long-end buybacks funded with T-bill issuance and a cooperative Fed holding down bill rates, an Operation Twist at scale. Both would be inflationary, he said, and he called a "default"-style restructuring unlikely in the near term. Rosenberg countered that anything resembling a default would collapse the dollar and speed the loss of reserve currency status. Gundlach said he never expected to say it, but now believes the dollar could lose that status in his lifetime. He said that is one reason he still owns gold. Gundlach noted he has called this "the most unhinged" bond market he has seen in 40 years, and he compared the setup to Canada's 1994 near-failed 10-year auction, which Rosenberg recalled as the point where Canada lost control of its curve.
The Debate Over Whether Kevin Warsh Is the Catalyst
The two diverged on causation. Rosenberg argued that deficits and debt were already known in February, when the 10-year was near 3.97%, so something else changed. His answer is Fed Chair Kevin Warsh, whom he accused of telling markets he will "raise rates into a supply shock" while focusing on headline inflation rather than core. He pointed to 5-year, 5-year forward breakevens of 2.34%, roughly where they were five years ago, and said the move has come entirely from real yields and risk premium. The market has gone from pricing one cut to pricing a hike and then four more, and the 2s10s curve has flattened to about 20 basis points. "I find the guy dangerous," Rosenberg said. "I never said that about any other central banker."
Gundlach agreed that Warsh has introduced a new and confusing regime, with no dot plot, no forecast and little forward guidance. He compared him to Al Gore's shifting debate personas, and said that at the latest press conference he wrote "no answer" next to five or six questions in a row. Still, he maintained that structural forces explain why yields rise even on weak data, since a softer economy widens deficits and the old safe-haven bid is not appearing. The two also agreed that the Fed's dual mandate is now in direct conflict with the interest expense problem. Raising rates to fight inflation increases interest costs, and cutting rates worsens inflation. "It's going to have to break one way or the other," Gundlach said.
Rosenberg argued the inflation case is overstated. New home prices are down 8% year over year, builders hold about 8.5 months of supply, and rents are deflating. He said fuel is "a tax" rather than sustained inflation because the labor market is not cooperating. He cited falling quits in the JOLTS data and the Conference Board's drop from 89 to 82, although its 12-month inflation expectations rose to 5.1%. Gundlach noted the Atlanta Fed's third-quarter GDP tracker is above 5%, partly on inventory building and pull-forward buying that may reflect inflation anticipation.
Positioning: Dry Powder, Seasoned Mortgages and CLOs
Gundlach recommended investors hold at least 20% in dry powder, which he described as short duration, high quality assets, and said the valuation of the equity market is "completely untouchable." For this bucket he favors AAA CLOs at SOFR plus about 130 basis points, or roughly 5.30% with SOFR near 4%. He also likes seasoned agency-style residential mortgage securities originated five or six years ago, which he called effectively risk free on credit given home price appreciation and loan-to-value ratios well below the original levels. They yield about 140 basis points over Treasuries, or nearly 7%. A portfolio yielding 6% with a duration of 2 has what he calls a Sherman ratio of 3, meaning rates must rise 300 basis points before the return goes to zero.
For the roughly 30% he would put in bonds, he suggested a barbell. One half would be a low-risk, US dollar-only intermediate strategy weighted toward mortgage and government securities. The other half would be actively managed local currency emerging market bonds, yielding about 7%, with emerging Asia favored for its lack of fiscal strain. He warned against index funds that include "serial defaulters." For equities he would use equal weight rather than cap weight.
On gold, he said central banks have moved from about 20% to around 30% of reserves in gold, and sees no reason it cannot reach 50%. He said he would "buy with both hands" below $3,800. He also warned against swapping long nominal Treasuries for 30-year TIPS, which he said are "perfectly correlated" with nominals, so investors would take the same loss. He said he would want a 6% yield on the 30-year, a real yield of about 300 basis points against inflation he puts at 3% to 3.5%, before buying.
A model he shared put fair value for the 10-year at about 4.95%, based on the average of the 7-year moving average of US nominal GDP and the German 10-year yield, which has an R-squared of 0.93 over 40 years. The actual yield is about 5.25%, a divergence of about 30 basis points. Gundlach said that is not enormous, but the model suggests yields should be slightly lower. Rosenberg said he would buy bonds on any recession signal. Gundlach said he would too, but as a trade, not a hold.
Insurers and Private Credit Are the Next Flashpoint
When asked about principal-protected notes and leveraged ETFs, Gundlach said the latter are signposts of a mania. He then turned to insurers owned by private equity firms that hold private credit, which he said are being used to absorb poor-quality debt. He likened the situation to the CDO boom of 2005 and 2006, when startups were created as "dumping grounds" for poor collateral. He recalled telling a senior Merrill Lynch executive about the risk, only to see the executive collect a $250 million bonus and leave. Gundlach advised checking who stands behind any annuity or life policy, and said mutual insurers avoid the problem. "It's already begun," he said.
Treasury Refunding as a Possible Relief Valve
Rosenberg raised the November 4 Treasury refunding announcement, the day after the midterms. He pointed to Treasury Secretary Janet Yellen's November 2023 shift toward bills, which coincided with a 100 basis point fall in the 10-year over two months. Gundlach said the Treasury could cut long-end issuance, but a fuller operation would need the Fed to buy at the long end and lower bill rates, and the Fed is moving the other way. Rosenberg also expects gridlock after November 3, which would end fiscal stimulus talk. Rosenberg said that would make an inverted yield curve a more reliable recession signal than in 2022, when stimulus checks blunted it.