AI Bubble Meter Hits 83/100 as Piper Cuts CoreWeave’s Target

Yardeni Research said in a note on 21 October 2026 that an AI bubble exists in the stock market but is not about to burst, according to Finwire’s account of the note. Its new AI Bubble Meter puts bubble conditions at 83 out of 100 and crash triggers at 34. A day earlier, Piper Sandler cut its price target for CoreWeave to $124 from $153 while keeping a Buy rating, according to TipRanks.

In plain terms, the meter has two dials. The first, Conditions, asks whether today’s market looks like the setups that came before the 1999–2000 tech bust and the 2006–2007 housing bust. The second, Triggers, asks whether the forces that ended those booms, a credit boom and a sustained rise in the cost of capital, are under way. Below, we set Yardeni’s reading next to the other views published this week, each with its own measure.

Business Pill · FINANCIAL CONDITIONS

A short Business Engineer explainer on financial conditions: several prices, from yields to credit spreads, pulling at once. It teaches the general idea only and says nothing about any company or forecast.

What the Bubble Meter Says

“The meter says there’s a bubble, but it’s not about to burst,” Yardeni wrote, as Finwire reported. The firm said one trigger has kicked in: the two-year Treasury yield is up 115 basis points from its December low. Credit, it said, is not stretched outside AI-related areas of the economy.

“The preconditions of a bubble are in place and have been for a while, but the forces that ended past booms are only partially present,” the firm said. Since 1965, according to Yardeni, quarters with high Conditions and low Triggers readings were followed by an S&P 500 bear market within two years only 9% of the time, and the index’s median inflation-adjusted gain over the next two years was 18.3%. When both dials read high, the odds of a bear market rose to 42%, with a median loss of 12.5%.

“But the Triggers dial is the one to watch: If it starts rising toward High, so do the risks,” Yardeni said. The firm said it will watch whether the two-year yield moves toward the 200 to 350 basis point rises that preceded the 2000 and 2006 bubbles, and look for signs of idle capacity in hyperscalers’ third-quarter results. Finwire reported that the firm built the meter using Anthropic’s Claude Fable 5.1.

What Was Posted on X

Walter Bloomberg, a wire-headline account, posted the Yardeni Research reading on 21 October: bubble conditions at 83 out of 100 and crash triggers at 34, with similar conditions preceding an S&P 500 bear market only 9% of the time within two years. This piece takes the meter’s details from Finwire’s account of the note.

Bloomberg posted on 21 October that Deutsche Bank’s chief executive does not see a bubble in AI, after Ken Griffin warned of looming setbacks to investors. Christian Sewing’s quotes in this piece come from Bloomberg’s report as syndicated by The Edge.

The Scorecard: Who Says What

Each voice below uses its own measure, so the rows are not one scale. We list the view as each person or group put it, with the source we read.

Griffin, speaking at the Future of Global Markets Conference in New York on 20 October, argued that radical transformations of the economy tend to be tied up with bubbles, investingLive reported. He compared the moment to the dot-com era and said he wished he had held on to his Amazon shares.

Sewing, interviewed at the Berlin Global Dialogue, added: “Do we see the one or the other setback, also with individual companies? That can happen.” He also said “The fear that we are in front of a bubble is not a justified one,” according to Bloomberg’s report as syndicated by The Edge. Fink and Sewing both pointed to fiscal deficits as a more immediate concern.

Chiavarone, Federated Hermes’s chief investment officer for global equities, told Bloomberg that markets are being discerning about which companies receive capital at what rates, the opposite of bubble behaviour as he describes it.

Bill Dudley, a former president of the Federal Reserve Bank of New York who writes an opinion column for Bloomberg, made a rates argument rather than a bubble call. “The AI spending spree is putting the Fed on a path of raising rates followed by a sharp reversal,” Bloomberg said in its post on X summarising the column. The readable opening of the page says the boom is pushing interest rates up in two ways: investment spending is boosting growth and straining the nation’s financing capacity, and a buoyant stock market is lifting household wealth and consumer spending. It is an opinion piece.

What the Market Data Shows

• Chip exports. South Korea’s chip exports rose 222.5 percent to $27.5 billion in the first 20 days of October, Yonhap reported, citing the Korea Customs Service; chips made up 44.7 percent of total exports.
• CoreWeave targets. Piper Sandler kept a Buy rating and cut its target to $124 from $153; Barclays kept a Hold rating with a $105 target, both on 20 October, according to TipRanks.
• SpaceX targets. JPMorgan raised its SpaceX price target to $500 from $240 on 21 October, according to Rallies and posts on X by Walter Bloomberg and Shay Boloor.
• Funding. Axios reported on 21 October that investors are “pulling back funding and demanding more evidence that their investments will actually pay off,” citing higher borrowing costs for CoreWeave, Elice’s fall of more than 30% in its trading debut and Firmus’s shelved IPO.
• Long-run returns. JPMorgan raised its forecast for annual returns on U.S. large-cap stocks to 7.5% over the next 10 to 15 years, up from 6.7% last year, Axios reported.

“They started by financing with free cash flow. That kind of ran out,” David Lebovitz, head of alternative research and strategy at JPMorgan Asset Management, said of the hyperscalers, according to Axios. “Now, they’re moving into the credit markets. Well, credit markets are having a little bit of an indigestion period.”

Update, 21 October: Brookings Paper Puts the Payback Bar at $3.7 Trillion a Year

A Brookings Papers on Economic Activity paper by Columbia Business School’s Stijn Van Nieuwerburgh, “Financing the AI Buildout” (conference draft dated 4 September 2026, presented at the 24–25 September conference), puts the revenue the US AI buildout would need at “approximately $3.725 trillion by 2032”, which it says “represents 9.2% of 2032 GDP of $40.481 trillion.” That central case assumes a 10 percent required unlevered return and a 50 percent operating cash-flow margin on the 182.7 GW of capacity the paper counts as completed between 2025 and 2032.

The paper gives a range: the estimate runs “from $2.29 trillion under a 5 percent required return and a 67 percent margin to $6.72 trillion under a 15 percent required return and a 33 percent margin.” It adds that, given “current estimates of annual combined revenues of OpenAI and Anthropic of around $100 billion”, revenues “would need to grow at roughly 80% per year until 2032”. It also argues the requirement “does not appear implausible relative to current pricing for frontier compute”, equating it to about $5.5 per installed GPU-hour at full utilization.

A post that circulated on X today put the figure at $3.5 trillion a year and 8.8% of GDP, compared with what Americans spend on food. The draft we read gives $3.725 trillion and 9.2% of 2032 GDP and does not make a food comparison. Our earlier coverage carries the paper’s $10.3 trillion investment estimate.

Connecting the Coverage

Our Chamath piece reported on 21 October that Chamath Palihapitiya wrote on X that “roughly, $0.70-75 of every dollar of AI spending right now comes from AI companies spending VC investment dollars,” in a post quoting a summary of Griffin’s remarks; he gave no source for the figures behind his estimate.

Our credit-insurance piece added an update on 21 October on the Open Markets Institute and Americans for Financial Reform Education Fund report, which says “Speculative data center debt is being repackaged and sold to pension and retirement funds, university endowments, and other institutional investors in a manner eerily reminiscent of the subprime bubble.”

Our Merlin piece carries the Piper Sandler and Barclays CoreWeave targets, and our JPMorgan SpaceX piece carries the bank’s $500 target. Our Goldman piece covered the bank’s 19 October report, which estimates roughly $7.6 trillion of capital spending from 2026 to 2031 across computing, data centres and power, according to Axios.

What is new here is a two-number gauge from one research firm, set beside the week’s other readings: as we read it, the debate is less about whether AI spending is large than about whether the financing behind it holds.

The Structural Read

As we read it, the views that say there is a bubble and the views that say it is not bursting are measuring different things: Griffin speaks of valuations, Yardeni of two dials, Sewing of a bubble “which is bursting” and Chiavarone of who gets financed at what rate.

As we read it, the market data this week points both ways: Korea’s chip exports rose 222.5 percent in the first 20 days of October, Yonhap reported, while Piper Sandler cut its CoreWeave target to $124 from $153, according to TipRanks.

As we read it, the financing question is where the voices meet: Axios quotes JPMorgan Asset Management’s David Lebovitz on hyperscalers moving into the credit markets, and the Open Markets report likens data centre debt sales to subprime.

Three Implications

FOR EXECUTIVES PLANNING AI BUDGETS Yardeni says the triggers to watch are a credit boom and a sustained rise in the cost of capital, so the cost of borrowing is the number it tracks.

FOR INVESTORS IN AI INFRASTRUCTURE Chiavarone told Bloomberg markets are being discerning about which companies receive capital at what rates; this week Piper cut its CoreWeave target and JPMorgan raised its SpaceX target.

FOR READERS TRACKING THE DEBATE Each voice uses its own measure, from Yardeni’s two dials to Palihapitiya’s estimate of VC-funded spending, so the scorecard is not one scale.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Subsidized AGI Economy.

The framework argues: “The bubble framing sees -1,650% margin and reads collapse. The structural read sees -1,650% margin and reads purchase order: the labs are spending negative margin today to acquire the three assets that determine who wins when the deflation curve crosses zero.”

As we read it, the same split runs through this week’s views: the bubble readings look at prices and debt, the structural read at what the spending buys. The framework was published in June 2026 and does not discuss Yardeni’s meter.

What Is Not Established

We read Yardeni’s note through Finwire’s account and a post by Walter Bloomberg, not the note itself, and we could not check how the dials are calculated. Bloomberg’s Chiavarone article was readable only in its opening paragraphs, and Griffin’s remarks reached us through investingLive and Bloomberg’s account as syndicated. Bloomberg’s report on Asian chip exports was not readable to us, so the chip figure above is Korea’s customs data only. None of the measures here gives a date for a market turn.

The Bottom Line

Yardeni Research’s new meter reads 83 out of 100 for bubble conditions and 34 for crash triggers, and the firm says the bubble is not about to burst. Griffin says investors are almost certainly in a bubble; Sewing, Fink and Chiavarone say it is not bursting or not yet a worry. As we read it, the voices agree that the spending is large and differ on how sound the financing behind it is.