Speaker · Updated
Bob Elliott: latest market views
Unlimited Funds
Summary
Co-founder, CEO and CIO of Unlimited, which uses machine learning to build ETFs replicating hedge fund and other alternative-investment returns, and a former Bridgewater Associates senior investment executive who led Ray Dalio's research team.
Chorus has 16 views from Bob Elliott across 2 episodes on Excess Returns and CNBC Closing Bell since October 9, 2026. Their latest calls: bearish on Anthropic.
Where they stand
Their latest bullish or bearish call on each company.
- AnthropicBearish ·
What they talk about most
The latest views
Newest first. Every quote is checked against the episode's transcript.
AI infrastructure & data centers
Democratic governors now cover about 80% of expected data-center build-out, up from 30%, a drag on AI.
Governors normally matter little, but they have real authority over data center investment and can restrict data centers by executive order, so the election matters more than usual. Before the election, Democratic governors covered about 30% of the expected increase in data center construction or power over the next five years. After the election that share is about 80%. Elliott also notes stocks rose while video fell after the results. For a market so concentrated in AI, this is another chink in the armor of the AI mania. It does not change the picture alone, but it adds to the drag on the narrative.
“Democratic governors covered about 30% of the perspective, expected increase in data center construction or power over the next, whatever, five years, and now 80% of that is covered by democratic governors.”at 4:52
Excess Returns: Spending Is Up 7%. Paychecks Are Up 3%. What Happens When Stocks Stop Paying the Difference? ·
AI mania has priced in extraordinary earnings growth, setting up disappointment and bad forward equity returns.
Elliott describes the AI complex as a mania, a dynamic in which a seed of reality, here strong recent earnings growth, gets extrapolated into the future. Current pricing implies the most extraordinary earnings growth over the next five years of the entire post-WWII period. That creates two risks: disappointment against very high expectations, and tightening, both market-based (rising spreads and more persistent borrowing) and from the Fed, which the oil complex forces to run tighter policy. Historically that combination is bad for manias on a forward basis. He compares it to 2007 housing, when stocks made new highs even as losses were coming, and says the timing is uncertain.
“what's penciled out is, you know, the most extraordinary earnings growth by a lot in the entirety over the next five years of the entirety of the post war World War two period.”at 8:52
Excess Returns: Spending Is Up 7%. Paychecks Are Up 3%. What Happens When Stocks Stop Paying the Difference? ·
Nominal growth has peaked and is slowing into Q4, while market pricing still expects very high growth.
Elliott distinguishes peak nominal growth from peak pricing of nominal growth. Nominal GDP growth in Q3 was already lower than in Q2 on a GDP basis, and he expects it to slow even more significantly in Q4. Markets have extrapolated strong nominal growth, so when expectations of very high GDP growth meet slowing GDP growth, disappointment follows. He describes this as alligator jaws, with market pricing diverging from macro outcomes.
“nominal growth in the third quarter, but just on a GDP basis was lower than it was in the second quarter... and likely will slow even more significantly in the fourth quarter.”at 11:02
Excess Returns: Spending Is Up 7%. Paychecks Are Up 3%. What Happens When Stocks Stop Paying the Difference? ·
Consumers are in a savings-driven expansion with income growth lagging spending, showing cracks and flat real spending.
Elliott argues the expansion moved from income-driven, when spending and income both grew about 6%, to savings-driven, where nominal spending runs 6-8% to keep up with high prices while income growth has slowed to about 3-3.5%. Households have been stressed since early 2025 and acutely since oil prices rose this year. Real consumer spending was zero in September, and he sees cracks. The model relies on asset prices rising to keep incentivizing savings drawdown, but stocks have been flat for six months, 401(k)s are flat, and housing is flat or down in sensitive areas, removing that support.
“we've transitioned from an income driven expansion where essentially spending was 6% annual growth and and income was 6% annual growth, to DIS savings driven expansion, where spending is six, seven, 8% nominal”at 15:39
Excess Returns: Spending Is Up 7%. Paychecks Are Up 3%. What Happens When Stocks Stop Paying the Difference? ·
Central banks are abdicating with inflation at 3-4%; tightening is market-driven, and the Fed is lazy, not behind.
Elliott notes inflation is around 3-4% across the developed world, yet central banks are doing little beyond speeches. In a past regime with low unemployment and inflation of 3-4%, a central bank would be tightening. Pressure is needed to get even one hike. The Royal Bank of Australia is an exception. He reads this as a shift in the reaction function: the tightening cycle is more market-based than central-bank-based, and policy is less acute than conditions warrant. Asked whether central banks are behind and will catch up or are simply lazy, he bets on lazy. Menon sees it as a debate, not a settled view.
“Inflation is three to 4%, and it's pretty much three to 4% across the developed world, maybe a little higher in certain places. And yet, like, what are central banks doing? Like, nothing. Basically, no.”at 21:53
Excess Returns: Spending Is Up 7%. Paychecks Are Up 3%. What Happens When Stocks Stop Paying the Difference? ·
Bonds look more favorable than stocks on a forward-looking basis, as priced-in growth looks too high.
Elliott frames the stocks-versus-bonds trade as a bet on whether extraordinary nominal growth persists. Over the last few months bond yields rose sharply while stocks barely moved, which only makes sense if markets priced even better earnings just as the main source of earnings growth is being challenged. On days when a major AI company's earnings disappointed, stocks fell and bonds rallied, showing repricing of growth. Pricing does not assume terrible outcomes, but the extraordinary outcomes priced in look much less likely, so bonds look more favorable to stocks forward. Not impossible, but much lower.
“bonds look more favorable to stocks on a forward looking basis from my perspective.”at 27:43
Excess Returns: Spending Is Up 7%. Paychecks Are Up 3%. What Happens When Stocks Stop Paying the Difference? ·
No macro evidence yet of AI-driven productivity; real GDP per worker sits on its pre-ChatGPT trend.
Elliott says there's no evidence of a meaningful macro productivity increase from AI. Real GDP per worker is consistent with the trend that existed just before COVID, the post-COVID trend before ChatGPT, and the trend from the nineties and early 2000s. Many people say AI makes them incrementally more productive, but that is not the test. The real question is whether AI creates an acceleration in productivity, which is what the extraordinary investment and earnings already priced in would require. Nothing he sees at the macro level is meaningfully changing that.
“It is consistent and in line with the trend line that existed just before COVID”at 36:34
Excess Returns: Spending Is Up 7%. Paychecks Are Up 3%. What Happens When Stocks Stop Paying the Difference? ·
The real economy shows no sign of paying for AI chips; tech margin gains reflect capex accounting, not AI.
Margin expansion is concentrated in tech and comes from fixed investment: capex is amortized over about twenty years while the cash cost is spent upfront, which makes margins look better without AI itself producing income. Outside tech, nothing in margins or income per worker shows companies willing to pay for the chips. Productivity gains may exist at the margin, but much of the benefit may be consumer surplus, which doesn't raise incomes. If AI is priced as the greatest technology in history, its macro effects should be visible, so the burden of proof sits with the bulls.
“the answer is, like, nothing in the real economy indicates that they're willing to pay for [Nvidia's] chips, in terms of their margins, in terms of their income per worker, none of it.”at 38:24
Excess Returns: Spending Is Up 7%. Paychecks Are Up 3%. What Happens When Stocks Stop Paying the Difference? ·
AI revenue is far too small to justify the roughly $10T buildout; recent revenue growth has flattened.
About $10T of AI-related investment is expected cumulatively by 2030, and servicing it would need roughly $3-4T of annualized revenue. Current AI revenue is maybe $150-200B. Reaching $3T in a couple of years would require exponential growth, but revenue at the highest-frequency AI adopters has been roughly flat for about three or four months after looking parabolic over the summer. Bob argues the expectations built into markets are far ahead of what the revenue data shows.
“You can't have many flat for four months periods if we're gonna go from 200,000,000,000 to 3,000,000,000,000 in the course of a couple of years. Like, we are so far off the curve in order to get to what is priced into the market right now.”at 41:54
Excess Returns: Spending Is Up 7%. Paychecks Are Up 3%. What Happens When Stocks Stop Paying the Difference? ·
Credit system is too unlevered to produce a systemic credit event; no existential credit risk remains.
Banks have gone from about 30x levered to about 8x. Much credit now sits on low-leverage terms, around 50% loan-to-value, and private credit funds are often only about 2x levered. Because of this, losses can be absorbed without ruin. Commercial real estate losses have run about 20-30 basis points over the last three or four years, and the 50% LTV loans explain why losses were contained after SVB. Credit has been moved from banks to unlevered pension funds and retail investors via private credit, so it now behaves more like equity than credit and has removed the pre-GFC type of existential credit risk.
“we've transformed the credit system to make it look a lot more like equity than it like credit. And as a result, have basically sucked out all of the, the the type of existential credit risk that existed, you know, in the pre GFC period.”at 49:40
Excess Returns: Spending Is Up 7%. Paychecks Are Up 3%. What Happens When Stocks Stop Paying the Difference? ·
AI infrastructure & data centers
About $6T of AI forward commitments far exceed roughly $200B of income, risking a self-reinforcing downturn.
Forward contractual commitments across the AI cycle total about $6T, against about $200B of actual income, which Bob calls pretty levered. Those commitments work like debt: firms engage in economic activity today on the promise of future income. If people doubt AI will generate enough real-economy revenue to deliver on those commitments, they scale back activity. That slows capex, which leads to an earnings downturn, then an asset-price decline and negative consumer spending, a self-reinforcing cycle. Bob stresses this is the dynamic, not a near-term prediction. He calls it a real-economy risk rather than a credit risk.
“we have built up, you know, $6,000,000,000,000 of forward contractual commitments, particularly related to the AI dynamic.”at 51:32
Excess Returns: Spending Is Up 7%. Paychecks Are Up 3%. What Happens When Stocks Stop Paying the Difference? ·
AI capex needs $3T of AI revenue by 2030, but revenue growth has turned linear, so the trade looks unsupported.
Elliott says the capex penciled into Nvidia and Micron numbers requires $3 trillion of revenue going to the AI complex by 2030. He points to OpenAI's new information and Anthropic's flatlining revenue. Over the last three to six months, AI growth has shifted from parabolic to linear. Linear growth will not reach the revenue needed, so the US stock market is one big bet on AI success.
“In order to justify the type of CapEx that's currently penciled out... you need $3,000,000,000,000 of revenue going to the AI complex by 2030.”at 41:17
AnthropicBearish
Anthropic's revenue has flatlined, which signals AI revenue is shifting from parabolic to linear growth.
Elliott cites the flatlining of Anthropic's revenues as a major part of the AI economy's revenue cycle. Combined with new OpenAI information, he says growth has changed from parabolic to linear over the last three to six months, which will not support the needed revenue.
“combined with the flatlining of revenues from Anthropic. Those are big parts of the overall revenue cycle in the AI economy right now”at 41:30
Forward P/E numbers rely on extraordinary earnings growth that may not be realised.
Elliott says the forward P/E ratios being cited are based on forward earnings, which are not real earnings. Those multiples are baking in some of the most extraordinary earnings growth in the postwar era.
“those are forward PE numbers, and forward earnings is not real earnings. ... What they're baking in is they're baking in some of the most extraordinary earnings growth in the postwar era”at 43:21
Consensus expects over 20% earnings growth for six to eight quarters, which is hard to meet.
Elliott says forward P/E ratios are based on forward earnings, which are not real earnings. The consensus expects the most extraordinary earnings growth in the postwar era, above 20% year over year for the next six to eight quarters. That is very hard to meet, and it depends on AI capex continuing, which requires real-economy companies to start spending on AI.
“consensus expectations above 20% earnings growth quarter after, you know, year on a year on year basis for the next six to eight quarters. That's very hard to meet”at 43:32
Banks are more robust and profitable than in past cycles, so any trouble is likely outside the banking system.
Elliott says banks were at the core of problems in previous cycles, but they are now more robust, more profitable and have better net interest margins than in a long time. He argues that if trouble comes, it will be outside the banking system and core financials, in the nooks and crannies where leverage has been building up.
“the banks are more robust, more profitable, better NIMs than we've seen in a long time. And so, if we're going to have a trouble, it's going to be outside the banking system”at 44:59
Shows
- Excess Returns · 1 episode
- CNBC Closing Bell · 1 episode
Ask your AI what Bob Elliott thinks
Chorus gives Claude, ChatGPT and other assistants every view Bob Elliott has shared on the shows it covers, next to everyone else's. Ask where they stand, what changed, and who disagrees.
Chorus reports what speakers said. It isn't investment advice.