The David Lin Report · · 51 min
Collapse Of Tech? Rosenberg &. Bianco Debate AI Bubble, Soaring Bond Yields
With David Rosenberg, Jim Bianco
Summary
Jim Bianco and David Rosenberg debate whether AI is in a bubble and where the US economy and bond market are heading. Rosenberg argues the AI capex wave, with about $10T pledged for 2026-2032, is far larger than the addressable market and will likely see roughly 40% written down, with credit default swaps on AI names tripling as evidence the risk differs from 1997. Bianco places the AI cycle closer to 1997-98, expects demand to follow once enterprise AI restrictions ease, and sees the buildout as a strategic necessity against China. On rates, both see bonds as cheap at roughly 5% yields, though Rosenberg expects growth to slow as fiscal stimulus fades while Bianco expects the Fed to hike in December after hot CPI, and the two also disagree over whether bond volatility reflects curve flattening or a broader signal. The episode also covers inflation as a series of supply shocks versus persistent spending, and the expectation of tax increases after 2028.
Key takeaways
- Jim Bianco on US rates & the Fed: Fed will hike 25bp in December and must hike again; public demands action on inflation.
- David Rosenberg on AI infrastructure & data centers: AI infrastructure capex is overbuilt relative to realistic revenues; expects heavy writedowns, like telecom in 2000.
- David Rosenberg on AI bubble & tech valuations: AI boom looks like the start of 2000, not 1997; a washout and bear market are likely underway.
- Jim Bianco on US Treasuries & yields: Bonds at a 5% yield are a core holding for anybody right now.
- David Rosenberg on US stock market outlook: Equities are expensive: Shiller PE near 40 makes stocks riskier than bonds.
What was said, by company and theme
29 views from 2 speakers. Every quote is checked against the episode's transcript (how).
US Treasuries & yields
Where everyone stands on US Treasuries & yields
Jim Bianco · Bianco Research, L.L.C.
Bonds have value at around 5% yields; limited room for yields to fall further.
Bianco says bonds still have value at current levels. He does not think they are cheap and does not expect yields to fall much unless the economy changes dramatically. Once the 10-year reached about 5%, the risk of buying bonds and taking a capital loss from yields rising further looked low to him. The only scenario he saw for higher yields was $200 oil, and oil is moving the other way. He argues investors can own bonds here and hold the roughly 5% coupon.
“Yes. Still think that bonds now have value. I am not a screaming. I don't think they're cheap. I don't think that they're going to fall a lot in yield unless there's a dramatic change in the economy.”at 2:57
David Rosenberg · Rosenberg Research
Treasuries offer a big cushion and decent value, with 2.8% 10-year real yields.
Rosenberg agrees with Bianco that the risk-reward in bonds is much better than in 2020, when a 10bp rise in the 10-year would put holders underwater. Now there is a large cushion: the 10-year real yield is about 2.8% and the long bond's real yield is over 3%. Bonds are cheap relative to equities on a yield-for-yield basis. He says that, based on where rates trade, they offer decent value. His firm already holds twos and tens and has recently moved into the long end.
“Right now, you have a significant cushion, and 2.8% real yields in the ten year over 3% real yields in the long bond gives you an enormous cushion.”at 4:51
Jim Bianco · Bianco Research, L.L.C.
Bond volatility tracks yield-curve flattening driven by Fed hike bets, not outright yield moves.
The MOVE index, which measures bond volatility, has been rising. Bianco says it corresponds much more closely with the shape of the yield curve than with outright moves in yields. The curve has been flattening because markets are pricing Fed rate hikes in December, and the post-CPI talk of a possible 50bp hike has added to that. So the bond-volatility spike is a reaction to curve flattening, not a separate signal about the direction of yields.
“What the bond volatility measures actually correspond with a lot closer is the yield curve than they do outright movements in yields.”at 5:47
Jim Bianco · Bianco Research, L.L.C.
Bonds at a 5% yield are a core holding for anybody right now.
Bianco says bonds at a 5% yield are attractive relative to equities, whose valuations and risk make their expected return about 6%. He sees bonds as the anchor of a 40–50% allocation. He argues that cash, around 4%, is also a reasonable return compared with the near-zero yields of the last decade. His view is that bonds are an anchor in an allocation that is less dependent on equity-level returns.
“bonds at a 5% yield, I really think that they're a core holding for anybody right now.”at 45:19
David Rosenberg · Rosenberg Research
Bonds are cheap relative to equities; the roughly 2.8% real 10-year yield beats stock earnings yield.
Rosenberg says the Shiller PE near 40 for five months in a row implies a real earnings yield of about 2.5%. That compares with a real yield of about 2.8% on the 10-year Treasury and above 3% on the long bond. He argues that bonds are cheap yield for yield relative to equities, and that bonds offer a better return than equities for the risk. He disputes Jim Bianco's view that bonds are not cheap.
“Bonds are cheap relative to equities, yield for yield.”at 49:49
AI infrastructure & data centers
Where everyone stands on AI infrastructure & data centers
David Rosenberg · Rosenberg Research
Data-center capex growth will slow sharply next year but not turn negative.
Rosenberg notes that data-center construction spending rose 90% this year and real AI capex rose 16%, while the rest of capex was flat. He says next year will not repeat that. AI capex will continue, but its growth rate will fall substantially. The sector will remain a force on the economy but a much smaller one, with the equity wealth effect also weaker. He thinks the economy is in the late innings of the cycle.
“I think that the AI CapEx will be there. But in terms of the growth of the derivative, I think it's going to go down a lot. It's not going to go negative...”at 25:42
Jim Bianco · Bianco Research, L.L.C.
Data-center buildout must continue; moratoriums would let China win, and backlash is resolvable.
Bianco calls AI the most transformative technology since the railroads and says the US must be at the leading edge. If there is a moratorium or political will not to expand AI, he thinks the US will face problems and China will take over. He sees the political backlash against data centers as anger at the AI industry, which has promoted AI with overheated claims such as Bill Gates saying it could kill a billion people. He treats issues like water use, electricity and noise as resolvable. He sees 2027 and 2028 as the period when the US must decide whether to resolve the backlash and keep building or block AI on the demand and supply sides. He expects a resolution.
“We need to do this. We are going forward with this one way or another. If we do a moratorium, if we have a political will not to expand with AI, then I think we're going to run into problems that the Chinese are just gonna take over.”at 34:09
David Rosenberg · Rosenberg Research
AI infrastructure capex is overbuilt relative to realistic revenues; expects heavy writedowns, like telecom in 2000.
About $10T of AI capex is pledged for 2026 to 2032, which is unprecedented as a share of GDP. Rosenberg argues nobody knows how big the AI market will ultimately be, so the buildout risks overshooting demand. He cites a Gartner estimate that investors need about a 12% return, which would require roughly $8.5T of cumulative AI revenue, while Gartner's own estimate is about $4T, so the capex is far too large for the addressable market. He says analysts expect about 40% of this capex to be written down. He compares it to telecom and fiber, where about 30% of capex was written off in the early 2000s and Lucent and Nortel collapsed. He says the race with China is a dangerous reason to overbuild, and that bankruptcies would follow for most players.
“And I don't think anybody has a clue how big is this market gonna be ultimately.”at 37:24
David Rosenberg · Rosenberg Research
AI borrowers are stretched: net interest expense is four times operating income; AI CDS tripled.
Rosenberg says demand for AI is not the issue. The problem is the financing: one heavily spending AI company has net interest expense about four times its operating income, and its capex is about three times revenue (David Lin's point). He asks how that is sustainable. He also says credit default swaps on AI names, apart from Microsoft, have tripled, which he argues shows that AI financing stress is different from the 1997 internet cycle. He sees the leverage and capex build-up as a different risk from past tech booms.
“Their net interest expense is four times bigger than their operating income. How is that going to be sustainable?”at 41:49
Jim Bianco · Bianco Research, L.L.C.
Wall Street will invent new products such as compute futures and SPVs to finance AI capex.
Bianco argues that every new technology has forced financial innovation. The Venetians created the secondary bond market to fund ships, the Dutch East India Company helped create public equity markets, and Lou Ranieri invented mortgage-backed securities for the 30-year mortgage. He argues that AI capex debt will need to come off balance sheets and into new products, including compute futures and special-purpose vehicles that buy the assets. In his view, Wall Street bends to new technology, not the other way round, so this financing shift is a precondition for how far the AI buildout can go.
“Wall Street is gonna have to mutate. It's going to have to get this stuff off the balance sheets. It's gonna have to make new products... new markets, whether or not it's compute futures, ... whether or not it's SPVs that buy these things or something.”at 44:51
AI bubble & tech valuations
Where everyone stands on AI bubble & tech valuations
Jim Bianco · Bianco Research, L.L.C.
Every technology ends in a bubble; AI is probably still closer to 1997–98 than to the peak.
Bianco argues every technology cycle ends in a bubble, so the real question is how close AI is to the peak. He places the AI cycle closer to the 1997–98 stage of the internet cycle than to its final stage. That implies more room before the bubble peaks, though he does not give a timeline or valuation evidence in this passage.
“Every technology ends in a bubble. Now the question is is how close are we to the bubble [peak]? I'm in the camp that we're still probably closer to '97, '98.”at 0:57
David Rosenberg · Rosenberg Research
AI bubble is not 1997; AI-sector credit default swaps have tripled, a risk 1997 lacked.
Rosenberg disputes the 1997 comparison. Credit default swaps across the board have widened, except for Microsoft, and CDS on AI-sector names have tripled. He says that did not happen in 1997. He points to stresses from leveraged capex, which he sees as a different kind of risk from the 1990s internet bubble. The argument rests on credit-market signals showing that AI-related borrowing carries more stress than the earlier tech cycle.
“This is not 1997. By 1997, credit default swaps across the board, save for Microsoft right now, credit default swaps in the AI space have tripled. That was not happening in 1997.”at 1:12
David Rosenberg · Rosenberg Research
AI boom looks like the start of 2000, not 1997; a washout and bear market are likely underway.
Rosenberg says this is not 1997. Netscape IPO'd in 1995 and the internet was only a two-year-old theme by 1997, while ChatGPT launched in late 2022 and AI capex is already far more acute as a share of GDP. Credit default swaps on AI names have tripled, which did not happen in 1997. He notes the median AI stock was down 35% from its peak as of last week, a meaningful washout that he calls a rotation. He says the 2000 pattern of a jagged rotation before the final peak is visible now. He is not sure the flush-out is finished, since bear markets come in waves, and says people only call a bubble after it pops.
“I think this is looking a lot like the beginning of two thousand.”at 42:10
Jim Bianco · Bianco Research, L.L.C.
AI boom is still around 1997–98; overbuild peak is still ahead, with compute constrained.
Bianco agrees every technology ends in a bubble, so the question is how close to the peak we are. He compares late 1999 or early 2000 with 1997 and says he is in the camp that we are still closer to 1997–98. He agrees the market will eventually reach an overbuilt stage, but he does not think it is there yet because compute is still constrained. He expects AI and technology stocks to lose a third or half of their value a couple of times before the bubble peaks, so the path has higher risk and higher expected return.
“I'm in the camp that we're still probably closer to '97, '98, and we still got more to go before we eventually get to that complete overbuilt stage.”at 43:41
US stock market outlook
Where everyone stands on US stock market outlook
David Rosenberg · Rosenberg Research
Equities are in the late innings of the bull market, not the first inning.
Rosenberg says the equity wealth effect supporting high-end consumer spending will be less powerful whatever the stock market does, since the market is already elevated. He does not know where stocks are going, but says it looks like late innings and not the first inning of a bull market. This is his reason to expect the wealth-effect support to fade as growth slows.
“Those things could go away. I don't know what the stock market is going do, but it looks like we're late innings. This is not the first inning of a bull market.”at 25:35
Jim Bianco · Bianco Research, L.L.C.
Equities should return roughly 6–7%, only slightly above bonds' 5%, given valuation risk.
Bianco says that given the valuation of equities and the risk in them, investors should expect equities to return about 6–7%, maybe closer to 6%. That is only a little above bonds, which yield about 5%. Equities have a place in a portfolio, but they carry higher risk. Buying AI or technology stocks offers a higher expected return, but Bianco warns of drawdowns of a third or a half along the way.
“I think you're probably better off thinking about equities is returning you six or 7%, maybe closer to six compared to bonds being five.”at 45:32
David Rosenberg · Rosenberg Research
Equities are expensive: Shiller PE near 40 makes stocks riskier than bonds.
Rosenberg's preferred valuation measure, the Shiller CAPE, has been running around 40 for five months. That implies a real earnings yield near 2.5%, below the 10-year Treasury's real yield of about 2.8%. He says it is odd that the stock market is treating equities as less risky than bonds. He says he is not ready to discard Markowitz's modern portfolio theory on that basis, so he remains cautious on stocks.
“My favorite multiple is the Cape, the Shiller PE, it's running now 40 for five months in a row.”at 49:09
US rates & the Fed
Where everyone stands on US rates & the Fed
Jim Bianco · Bianco Research, L.L.C.
Fed will hike 25bp in December and must hike again; public demands action on inflation.
After a hotter-than-expected CPI print, Bianco expects the Fed to raise rates by 25bp in December, with a 50bp hike debated but unlikely. He expects another hike at the following meeting. His reasoning is political: voters just swept the party in power out of office over affordability and inflation, and he says the Fed must respond. He thinks it cannot keep citing TIPS breakevens and disinflation in airfares and used cars while the public says inflation is out of control. Failing to hike would leave the Fed with a lot of explaining to do.
“I think that they have to raise rates in December, raise rates at their next meeting after that. The potential of 50 is on the table. I'd be surprised if they actually do it, but they're gonna hike rates, I think.”at 7:35
David Rosenberg · Rosenberg Research
Fed is over-tightening by chasing lagging CPI and jobs data; financial conditions equal six hikes.
Rosenberg thinks the Fed is making a mistake by relying on lagging indicators such as CPI and the U3 unemployment rate, and by focusing on headline PCE, which he says ties policy to diesel prices. He says the 2% target is arbitrary and is not a reason to keep tightening. Since February, with only one Fed hike, financial conditions have tightened by about 150bp, which he says is like six hikes. The effect will hit the economy with a lag. The yield curve is only a couple of hikes from inverting. He says it did not give a real recession signal in 2022–23 because of the stimulus checks, a cushion that does not exist now.
“And so since the end of February, with the Fed just actually tightening once, we've had an overall tightening in financial conditions of a 150 basis points.”at 13:55
US inflation
Where everyone stands on US inflation
David Rosenberg · Rosenberg Research
Inflation is shock-driven, not structural; without shocks core and headline would be below 2%.
Rosenberg argues inflation is a sequence of shocks rather than a sinister, embedded problem: COVID and its response, the Russia-Ukraine war, tariffs, the fiscal bill and now the war with Iran, which uses the Strait of Hormuz as a choke point. His team's analysis suggests that without these shocks, core and headline inflation would be below 2%. Progress has stalled because of energy, and now food and transport. Wages have lagged prices for seven months, unlike the 2021–22 wage-price spiral, so this is an imported supply shock squeezing real incomes. That will cut demand next year. He also says the BLS imputes much of the services data, so it is not trustworthy, while commodity prices are falling in real time: diesel and gasoline are both down more than 20% from peak and agricultural prices are down more than 20%.
“We just had a lot of shocks, one by one by one. And without those shocks, core and headline inflation would actually be below 2% right now.”at 12:59
Jim Bianco · Bianco Research, L.L.C.
Public anger stems from real wages lagging inflation since 2020, not from inflation level.
Bianco agrees with Rosenberg that the problem is the cost-of-living level, not the rate of change. He argues that the public's rate-of-change test is real wages: are wages growing faster than inflation? Since 2020 the answer has been no, while from 2010 to 2020 wages consistently outpaced prices. Whatever the inflation rate, the public focuses on wages falling behind, and that gap is the source of the anger behind the election result.
“But the rate of change that the public seems to be focused on is real wages. Are there wages going up faster than the inflation rate? And since 2020, the answer was no.”at 17:23
US growth & recession risk
Where everyone stands on US growth & recession risk
David Rosenberg · Rosenberg Research
Growth is fragile: AI boom added only about 2 points and those supports could fade.
Despite the AI boom, GDP growth is only a 2.1% four-quarter run rate, against 5% during the internet boom of the late 1990s. Rosenberg says the growth this year came from three static supports: data-center construction, the equity-market wealth effect at the top of the income scale, and the one big beautiful bill. Together they add about two percentage points. He argues all three could fade, which means there is more fragility in the economy than the AI headlines suggest. The AI boom has also drawn resources away from the rest of the economy, which he describes as K-shaped.
“You add up the three static effects that adds up to two percentage points. What's GDP growth? 2.1. Those things could go away.”at 25:28
David Rosenberg · Rosenberg Research
Fiscal gridlock next year will cut about 60bp from GDP growth.
Rosenberg says growth in 2026 was boosted by the one big beautiful bill, which accounted for about a third of the economy's growth. Next year that stimulus is gone and the fiscal picture is gridlock. He puts the drag from gridlock alone at 60 basis points off GDP growth. Combined with a drop in aggregate demand from around 2% toward zero to one percent, this creates slack, which pushes unemployment and wage growth down.
“And there's no one big beautiful bill. We just get fiscal gridlock next year. Fiscal gridlock next year alone is gonna drag 60 basis points off of GDP growth just from that.”at 25:59
US deficit & government debt
Where everyone stands on US deficit & government debt
Jim Bianco · Bianco Research, L.L.C.
Government spending and deficits are highly correlated with inflation; falling rates will fuel more spending.
Bianco argues the US keeps spending faster no matter which party is in power. Trump wanted a $5,000 check and Bernie Sanders is pushing Medicare for All. He says government spending and deficits have a high correlation with inflation. He asks how inflation falls when the government keeps spending more, and whether falling interest rates would give the government more room to spend, since lower interest costs free up budget space and could keep inflation higher.
“And that has a high degree of correlation, government spending and deficits to inflation.”at 18:42
US tax & spending policy
Where everyone stands on US tax & spending policy
David Rosenberg · Rosenberg Research
After 2028, Democrats will likely raise taxes on high earners, wealth and corporations.
Rosenberg expects Democrats to raise taxes significantly after 2028. He expects them to reverse the 2017 corporate tax cuts, which could take the corporate rate from 21% to about 28%. He also expects wealth taxes and high-income surcharges. He cites the near failure of California's Proposition 40 as a sign that such measures will keep coming to voters. He says budget scoring by the CBO constrains deficits, so tax increases would be a bigger part of the picture than spending growth.
“So I see a future of wealth taxes. I see a future of high income taxes, tax surcharges.”at 22:26
US elections & politics
Where everyone stands on US elections & politics
David Rosenberg · Rosenberg Research
A party sweeping midterms has about a 75% chance of winning the next presidential election.
Rosenberg cites historical experience: when a party has a sweep like Democrats had in the midterm election, that party has about a 75% chance of winning the next presidential election. He uses this base rate to argue that the post-2028 tax and spending shift he expects is a real possibility, not a remote one.
“the party who does that has a 75% chance of getting all the keys to the kingdom in the next election. So 75% odds.”at 23:33
Enterprise AI adoption & AI software
Where everyone stands on Enterprise AI adoption & AI software
Jim Bianco · Bianco Research, L.L.C.
AI demand, not financing, is the bottleneck; corporate IT blocks will ease within a year.
Bianco argues the real constraint on AI is demand from businesses, not financing or politics. Many companies with more than 10 employees have IT departments that restrict staff from using AI because of agentic control of computers, hacking and data-leak fears. He compares this to 1990s advice never to enter a credit card online, which was later resolved as the internet matured. He expects the same thing to happen with AI within the next year or so, after which workers will use AI to build slide decks and pull data, and demand will follow.
“I do think it has been the demand side for AI that has been the bottleneck more so than the financing side or the political side.”at 32:37
US–China AI & chip race
Where everyone stands on US–China AI & chip race
David Rosenberg · Rosenberg Research
Racing China is no reason to overbuild AI capacity; China's own overbuilding is a deflation risk.
Rosenberg agrees that the race with China matters but says the problem is overbuilding. He argues that the AI capex wave resembles past overbuilds that ended in writedowns and bankruptcies. He points to China's property market, where overcapacity led to a long period of deflation, and says China may repeat that with its own tech buildout. He says China's financial problems will be bigger than the US's, so the US does not need to chase China by overspending.
“I'm not gonna disagree with anything that Jim said about the race with China, but the problem with the race, is you overbuild.”at 37:16
China's economy
Where everyone stands on China's economy
David Rosenberg · Rosenberg Research
China may repeat property-style overbuilding, leading to multiyear excess capacity and deflation.
Rosenberg says China has a history of overbuilding, and its property market shows the result: a massive multiyear excess-capacity problem and deflation. He argues the same route may apply to the AI buildout. He says the country's financial disaster will be bigger than the US's, and that its capacity to overbuild should not be underestimated.
“maybe China is going to have, as they have with their property market, a massive multiyear excess capacity problem and massive deflation out of it.”at 38:56
Topics
- 2028 election odds
- AI bubble comparison to 1997
- AI bubble risk
- AI capex financing
- AI debt and credit risk
- Fed policy lag
- GDP growth drivers
- GDP growth outlook
- bond volatility
- bond yields
- bubble comparison
- bull market stage
- data center buildout
- data center capex
- deficits and inflation
- enterprise AI adoption
- expected returns
- inflation outlook
- overcapacity and deflation
- race with China
- rate hikes
- real wages
- tax increases post-2028
- valuations
More from The David Lin Report
- Markets ‘Spiraling Out Of Control’: When Will Stocks Bottom? | Chris Vermeulen · November 16, 2026
- Worst Housing Demand Ever: Where Prices Fall In 2027 | Reventure CEO · November 16, 2026
- Biggest Economic Disruption Ever: Why AI Revolution Has Barely Begun | Ethan Mollick · November 16, 2026
- Frank Giustra: Bubble 'Unravelling' Started, Entire Market, Economy Next · November 15, 2026
- Oil Plunges 20%: Is $60 Or $125 Next? | Doomberg · November 14, 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