The Right Bubble About the Right Technology

25 min read

The question "are we in an AI bubble in 2026?" has an arithmetic answer before it has a philosophical one. David Cahn's analysis at Sequoia Capital, circulating this week in late August, quantifies what until now had been market intuition: roughly $1.5 trillion committed to artificial intelligence infrastructure against approximately $80 billion in combined annual recurring revenue from OpenAI and Anthropic — a gap of $2.9 trillion between what has been built and what is being paid for [1]. The short, quotable answer is this: yes, there is a bubble, but it is not the bubble the word suggests. It is not a bubble of useless technology, like tulip mania; it is a bubble of duration, like the railways — capital with the maturity of quarters financing adoption with the maturity of decades.

The natural reflex when faced with a number carrying twelve zeros is to ask when it will burst. That is the wrong question, or at least the less interesting one. The right question is what this gap measures — and the answer is not found in supply-side balance sheets, where everyone is looking, but in demand-side org charts, where no one is looking. Because the most revealing data point this month comes not from San Francisco or Wall Street: it comes, without coordination and without citing one another, from three management publications — the Harvard Business Review, twice, and MIT Sloan Management Review — describing AI-buying companies held back not by the technology, but by themselves [5][6][7]. Supply has issued commitments against revenue that the buyer does not yet have the internal structure to generate.

There are, then, two clocks running on different time zones here. One is measured in quarters: compute obligations in the hundreds of billions of dollars, credit instruments designed to ease balance sheets, power plants reopened to feed data centers [4][2][3]. The other is measured in decades: the speed at which a real organisation — with its operating model, its risk culture, its scattered pilots — can convert a general-purpose technology into accountable value [8][5]. Cahn's $2.9 trillion gap is the distance between those two clocks, expressed in dollars. And economic history suggests that when that distance is closed by force, it is not the technology that pays — it is the capital structure that financed it.

The pattern is not new; indeed, it is among the best-documented in the history of capitalism. British railway mania of the 1840s ruined a generation of investors and left standing a network that served the country for a century; the fibre-optic overbuild of the 1990s vaporised entire market capitalisations and left in place the capacity on which the commercial internet would eventually run. Carlota Perez formalised this cycle in her distinction between the installation phase — when financial capital overbuilds, speculatively — and the deployment phase, when the real economy slowly absorbs what has been built. In both historical cases, investors were right about the technology and wrong about the timeline. This year's report on the frontier of work explicitly embraces that genealogy: the transformation under way, in the image of the Industrial Revolution and the internet era, will take decades to fulfil its promise, because it involves technological, social and economic change that cannot be compressed by financial decree [8]. Decades — while the capital commitments signed this year demand revenue within years.

It is against this backdrop that the collision between the two dominant readings of the moment becomes instructive, precisely because both start from the same facts. The Economist's Schumpeter column from January describes the innovations in energy and credit surrounding data-centre construction — instruments designed to ease strain on the power grid and on tech companies' balance sheets — as additional air pumped into the AI bubble [2]. That same issue of the Economist frames the moment within a broader deformation: a global economy in which capitalism has become an instrument of power ("Gunboat capitalism", per its editorial title), where political interference reshapes the world's largest companies and compresses their profits, and where even central banks are under attack [2]. In an economy so politicised, the argument suggests, AI asset prices have ceased to be a reliable signal — they are, in part, artefacts of financial engineering and state sponsorship.

The Wall Street Journal Weekend of 10 January documents exactly the same phenomena and reads them with the opposite sign [3]. Three nuclear plants that had been slated for closure are being reopened; demand for gas-fired plants has surged; grid-connection queues are lengthening. In this reading, none of this is air: it is load. No one revives a nuclear plant on short-term speculation — the cost of capital and the regulatory horizon of a reactor only make sense against real, measurable, growing electricity demand. Where the Economist sees instruments stretching a euphoria, the WSJ sees the physical economy responding rationally to actual consumption. The collision is pure: the facts coincide, the verdicts oppose each other. And that is precisely why the dispute cannot be resolved by looking at supply — it is resolved by calling in an arbiter that neither side summoned.

Before turning to the arbiter, though, it is worth gauging the strain on the side of those who built. The Wall Street Journal of 17 January shows OpenAI committed to hundreds of billions of dollars in computing capacity while, simultaneously, turning to advertising — a business model its own leadership had described as a last resort — to diversify revenue that subscriptions do not cover [4]. The gesture is worth more than any projection: a company does not embrace the model it swore to avoid when current revenue is keeping pace with the commitments it has made. Cahn's arithmetic — $80 billion in combined annual recurring revenue from the sector's two leaders against $1.5 trillion in infrastructure [1] — is not a thesis; it is the explanation for the behaviour observed. Supply is assuming, in its capacity contracts, a revenue curve that the buying market has not yet drawn.

And this is where the arbiter enters — the month's most significant convergence, all the more credible for being unintentional. The Harvard Business Review of November–December 2025 recommends that companies stop multiplying AI pilots and concentrate effort on a single area where the technology protects or creates hard-to-copy advantage [5]. The January–February 2026 issue locates the failure in the misalignment between innovation ambitions and operating models, and warns, in the words of Darrell Rigby and Zach First, that "too much change can traumatize your organization" [6]. The MIT Sloan Management Review's Winter issue closes the triangle: AI risk management within organisations is lagging, write Öykü Işık and Ankita Goswami, not because of technical limitation but because of internal cultural and structural problems [7]. Three publications, three independent diagnoses, a single conclusion: the bottleneck on AI value inside companies is organisational, not technological. Demand exists — companies are investing, piloting, reorganising — but it advances at the speed organisations change, which is the same speed it always was.

This is why the artificial-intelligence lens reveals what the financial lens, on its own, cannot. Read as a market phenomenon, the story of 2026 is a duel of multiples and sentiment — bulls against bears, with the same charts. Read as a phenomenon of technology adoption, the story changes nature: the $2.9 trillion gap ceases to be a verdict on AI and becomes a verdict on the maturity of the demand financing it [1][5][7]. None of this month's sources connects the two ends — Sequoia measures the gap from the supply side; HBR and MIT SMR explain, without knowing it, where that gap lives. The juxtaposition implies what no one writes: the gap is not a lack of demand, it is demand with the wrong maturity for the capital structure serving it. For a board of directors, the distinction is everything — because it determines whether the risk to be managed is that of betting on a failed technology, or that of being exposed, as a creditor, customer or competitor, to balance sheets that issued quarterly debt against decade-long revenue. The correction, when it comes, will not select for the companies with the worst AI. It will select for the companies with the worst timeline.

What the AI lens makes visible: four patterns the financial reading misses

The first pattern this lens reveals is the financialisation of compute — the transformation of computing capacity, a technological asset, into a financial asset with maturities, creditors and duration risk. David Cahn's analysis at Sequoia quantifies it with an arithmetic that needs no adjectives: roughly $1.5 trillion committed to AI infrastructure against approximately $80 billion in combined annual recurring revenue from OpenAI and Anthropic, leaving a gap of $2.9 trillion between what has been built and what is paid for [1]. Without the AI lens, this number reads as a valuation ratio — another episode in the long history of market over-optimism. With the lens, it reads as something else: the creation of an entire asset class — data centres, chips, energy contracts, structured-credit instruments — whose ultimate collateral is a curve of enterprise adoption that no one controls and which, as this month's sources demonstrate, moves at the speed of organisations, not the speed of balance sheets. The Economist, in its Schumpeter column of 17 January, denounces precisely the credit and energy innovations designed to ease balance sheets and the power grid as air pumped into the bubble [2]; what the column does not say, and the lens shows, is that these innovations are an attempt to stretch the duration of money without admitting that the original duration was wrong.

The second pattern is the physical materialisation of demand — the moment when a technology thesis leaves the Nasdaq and enters the power grid. The Wall Street Journal Weekend of 10 January documents three nuclear plants that had been condemned to closure and are now reopening, surging demand for natural-gas plants, and long queues for grid connection [3]. These are the very same facts the Economist reads as symptoms of a bubble [2] — and the collision between the two readings is itself this month's most valuable data point. Because a purely financial bubble does not revive nuclear plants: one reopens a plant to serve real, contracted load, with a horizon measured in decades. What the AI lens reveals is that electricity demand is the one link in the chain where adoption is already measurable in megawatts rather than promises — and that, paradoxically, it is also the point where the clocks least coincide, because a nuclear plant amortises over decades while the capital justifying it demands returns within quarters. Where the Economist sees air and the WSJ sees load, the lens sees a duration contract signed by two parties who measure time differently.

The third pattern lives on the supply side, and is called revenue improvisation. The Wall Street Journal of 17 January shows OpenAI committed to hundreds of billions of dollars in computing capacity while, simultaneously, turning to advertising — a business model its own leadership had described as a last resort — to diversify revenue that subscriptions do not cover [4]. Seen through the financial lens, this is prudent diversification; seen through the AI lens, it is a structural confession. When the dominant supplier of a frontier technology embraces the revenue model it had sworn to avoid, the signal is not one of demand strength — it is that contracted revenue is not keeping pace with the commitments made, and that the company is buying time with the one asset it has in abundance: attention. Advertising monetises users; enterprise subscriptions monetise transformation. The migration from the second to the first, even if partial, indicates where the curve is failing — and it is not on the user side.

The fourth pattern is the most important, because it explains the previous three: the organisational bottleneck on demand, now documented in triplicate. The Harvard Business Review of November–December 2025 orders a halt to the proliferation of AI pilots and concentration of effort on a single area where the technology protects or creates hard-to-copy advantage — the diagnosis from Goutam Challagalla, Mahwesh Khan and Fabrice Beaulieu is that companies scatter capital across dozens of experiments that never reach scale [5]. The following issue, penned by Johannes Berndt and co-authors, adds that experimentation with generative AI requires testing at the organisational level, not just at the tool level — and that excessive change, warn Darrell Rigby and Zach First, traumatises organisations [6]. The MIT Sloan Management Review's Winter issue closes the circuit: Öykü Işık and Ankita Goswami show AI risk management lagging due to internal cultural and structural problems, while Chris K. Anderson and Fredrik Ødegaard document lawsuits alleging collusion against providers of pricing algorithms and the companies that use them — some already settled, others dismissed, others active [7]. This last data point deserves the attention of any board: the first algorithm lawsuits at scale are not about the technology failing, but about the technology working too well — coordinating prices without human coordination. AI legal risk is no longer hypothetical; it now has a case number.

Together, these four patterns compose a picture no single source, on its own, offers. Enterprise demand for AI is real — companies are investing, piloting, reorganising — but it is held back by three bottlenecks that no amount of capital can resolve: scattered pilots lacking strategic focus [5], misalignment between innovation ambitions and operating models [6], and a risk culture that has not matured at the pace of the technology [7]. The report on the frontier of work is explicit about the horizon: this transformation, in the image of the Industrial Revolution and the internet era, will take decades to fulfil its promise, because it involves slow technological, social and economic change [8]. Against that clock, capital was raised with quarterly maturity [1][4]. None of these sources contradicts another on the facts; it is the juxtaposition that exposes the mismatch — and it is the mismatch, not the technology, that the correction will punish.

There is an implicit international comparison in this material worth making explicit. Infrastructure is concentrated in the United States — the reopened power plants, the grid queues, the compute commitments are American phenomena [3][4] — but the demand that must pay for it is global, and globally slow. A European company that reads the HBR and concentrates its pilots on a single area of advantage [5] contributes to American supply-side revenue at the pace of its own internal transformation, not at the pace of the debt schedule of whoever built the data centre. The geography of supply and the geography of demand do not coincide — and the AI lens shows that duration risk also has a territorial dimension: the balance sheets that issued the commitment are in one time zone; the organisations that must honour it are in another, in every sense of the word.

For a board of directors, what this lens makes visible — and the financial lens hides — is the exact nature of the exposure. The wrong question is "should we invest in AI before the bubble bursts?"; the right question is "what is the duration of the capital we are exposed to, directly or indirectly, and what is the real maturity of our internal demand?". A company with twelve scattered pilots and no area of focus is not behind on AI — it is financing, with its experimentation budget, the revenue illusion of whoever sells compute [5][1]. A company that concentrates, aligns its operating model and matures its risk governance [6][7] converts that same budget into advantage — and positions itself to buy cheap capacity when the correction repriced infrastructure. In April, this publication wrote that infrastructure speed and regime speed were in compression ("Sam Altman vs Mustafa Suleyman: The Compression Between Infrastructure and Regime"); what was then a qualitative tension now has a price: $2.9 trillion is the cost, measured by Sequoia, of two clocks that never synchronised [1].

BOX — Key concept: duration mismatch applied to AI

In finance, a duration mismatch occurs when an agent finances long-term assets with short-term liabilities — the classic mechanism of banking crises, in which deposits redeemable at any moment sustain thirty-year loans. Applied to AI, the concept describes a situation in which $1.5 trillion of infrastructure — data centres, chips, energy contracts — has been financed against an expectation of revenue that enterprise adoption, for organisational rather than technological reasons, will only deliver over decades [1][8].

The signature of the mismatch does not lie in asset prices, but in the behaviour of the parties. On the supply side, it manifests as revenue improvisation: OpenAI embracing the advertising it had treated as a last resort, to cover hundreds of billions in computing commitments [4]. On the financial side, it manifests in the credit and energy innovations the Economist denounces — instruments designed to ease balance sheets and the power grid, that is, to stretch the duration of the liability without acknowledging it [2]. On the demand side, it manifests in the convergent diagnosis of three management publications that do not cite one another: too many pilots [5], misaligned operating models [6], immature risk governance [7].

The critical distinction is between a thesis correction and a duration correction. A thesis correction invalidates the technology — that is what happened to much of e-commerce in 2000. A duration correction invalidates the capital structure but preserves the asset: the overbuilt optical fibre of the telecoms bubble changed hands for cents on the dollar and went on to sustain the entire internet. This month's data — nuclear plants reopened to serve real load [3], documented but slow enterprise demand [5][6][7] — points to the second type. What is at risk is not AI; it is whoever financed it on the wrong timeline.

For the decision-maker, the concept translates into a discipline of three questions: first, what is the balance sheet's indirect exposure — as creditor, as a customer with long-term contracts, or as a competitor to leveraged companies — to capital structures with this mismatch; second, whether the organisation's own AI initiative portfolio is generating defensible value or merely feeding supply-side short-term revenue [5]; third, whether the organisation will be positioned to acquire capacity — talent, compute, assets — when the repricing arrives. Duration mismatch is not avoided; it is arbitraged. And whoever understands it before the correction buys, afterwards, what others built.

Implications for decision-makers

For the CEO of an AI-buying company, the correct reading of the $2.9 trillion gap quantified by David Cahn [1] is counter-intuitive: the gap on the supply side is, in part, a mirror of one's own pilot portfolio. When the HBR of November–December recommends stopping the proliferation of scattered experiments and concentrating investment on an area where AI protects or creates hard-to-copy advantage [5], it is describing the aggregate behaviour that keeps OpenAI's and Anthropic's combined annual recurring revenue at roughly $80 billion — real, growing, but two orders of magnitude short of the capital committed [1]. Every pilot with no path to production feeds supply-side short-term revenue without building defensible value on the buyer's side. The first board decision, therefore, is not "how much to invest in AI"; it is where AI connects to the value chain with a compounding effect — and where it is merely rent paid to whoever raised $1.5 trillion [1][5].

For the investor, the operative distinction is between exposure to the technology and exposure to the capital structure financing it. The Economist documents how the credit and energy innovations designed to ease balance sheets are transferring duration risk into increasingly illegible instruments [2]; the WSJ Weekend shows the other side of the same coin — three nuclear plants condemned to closure reopening, surging demand for gas-fired plants, long grid-connection queues [3]. A balance sheet may hold not a single share of AI stock and still be exposed: as a creditor to data-centre financing vehicles, as a counterparty to long-term energy contracts, as a shareholder in utilities whose customer base has concentrated around a handful of hyperscalers. The due-diligence question has changed: it is no longer "does this company benefit from AI?"; it is "what revenue maturity supports this company's liabilities?" [1][2].

For whoever leads transformation within organisations, the convergence between HBR and MIT SMR supplies the map of the bottleneck. The HBR of January–February locates the failure in the misalignment between innovation ambitions and operating models, and warns — in the words of Darrell Rigby and Zach First — that "too much change can traumatize your organization" [6]; MIT SMR, through Öykü Işık and Ankita Goswami, shows AI risk management lagging due to internal cultural and structural problems, not technical limitations [7]. The practical implication is that accelerating adoption does not mean buying more compute — it means redesigning risk governance, the operating model, and the alignment between stack and strategy before scaling. Whoever reverses this order will produce exactly the pattern the three publications diagnose: high investment, intangible value [5][6][7].

There is also a talent implication rarely part of this conversation. Heidrick & Struggles' 2026 High End Independent Talent Report records 151% growth since 2021 in demand for interim C-suite leaders, with interim CFOs accounting for 51% of all interim leadership requests and requests for interim COOs up 250% year on year [9]. Sunny Ackerman writes that independent talent is "moving to the center of how critical work gets done" [9] — and the reading through this lens is precise: organisations are assembling flexible execution capacity for transformations whose duration no one can predict. In a mismatch of clocks, leadership elasticity is a risk hedge as relevant as capital elasticity. The board that treats AI as a technology project hires engineers; the one that treats it as a duration restructuring hires reconfiguration capacity [8][9].

Finally, the coldest implication: preparing the balance sheet to buy after the repricing. If the correction that comes is one of duration rather than thesis — and this month's data points in that direction [3][5][6][7] — then compute, energy contracts, talent and physical assets will change hands at a discount, just as optical fibre did in 2001. Liquidity available at others' wrong moment is the oldest of competitive advantages.

The paradox: real demand as alibi for the wrong capital

The tension this lens reveals is the following: the more tangible physical demand becomes, the more legitimate the capital structure financing it appears — and the more dangerous it is. The reopened nuclear plants and the grid queues the WSJ Weekend documents [3] are the favourite argument of those who dismiss the bubble hypothesis: electrical load is not speculation, it is physics. But the physics of demand says nothing about the timeline of revenue. A plant reopened to serve data centres whose monetisation model is still under construction — OpenAI turning to advertising that its own leadership had described as a last resort [4] — is real demand in the service of a liability with the wrong maturity. The concrete is real; the timeline is fiction.

The second layer of the paradox is more uncomfortable: the prudence of buyers, which protects them individually, is what worsens the gap collectively. When the HBR orders concentration rather than proliferation of pilots [5], and when Rigby and First warn against excessive change [6], they are giving organisationally sound advice — and every company that follows it stretches the interval between the capital invested and the revenue that ought to service it. The buyers' micro-level rationality produces the financiers' macro-level fragility. Neither side is wrong; it is the clocks that fail to coincide [1][6][8].

And there is a third layer: the credit innovations the Economist denounces as air pumped into the bubble [2] are, viewed closely, the system's most honest attempt to solve the real problem — stretching the duration of the liability to bring it closer to the duration of adoption. The paradox is that they do so without admitting it, packaging decades-long risk into instruments priced as though they were quarters. The remedy exists; it is simply being administered in secret, which guarantees no one doses it correctly.

Looking ahead: three repricing scenarios

The most likely scenario, in light of this month's data, is a phased duration correction: the most leveraged compute commitments are restructured, physical assets — plants, data centres, energy contracts [3] — change hands at a discount, and the application layer continues to grow on top of repriced infrastructure, just as the internet grew on top of the cheap fibre of 2002. In this scenario, the $2.9 trillion gap [1] does not disappear; it transfers from the original financiers to buyers with patient balance sheets. The winners will be the companies that, following the HBR's prescription, concentrated adoption in a defensible area [5] and arrive at the repricing with operational discipline and liquidity.

The second scenario is forced monetisation: supply, unable to wait for buyers' organisational maturity, accelerates consumer-facing revenue models — OpenAI's advertising is the first signal [4]. This path shortens the duration mismatch, but at the price of transforming frontier labs into media companies, with everything that implies for incentives, product and trust. It is the scenario in which capital wins and the thesis changes nature.

The third scenario — the least discussed — is successful slow absorption: buying organisations, backed by flexible leadership whose demand is already growing at double-digit rates [9] and by matured risk governance [7], convert pilots into production at the pace the frontier-of-work report anticipates for transformations of this scale — decades, not quarters, in the image of the Industrial Revolution [8]. In this scenario, the technology fulfils its promise in full; it simply fulfils it too late for the generation of capital that financed it. It is the scenario in which everyone is right and someone still loses money.

In any of the three, one constant holds: the correction, when it comes, will no longer correct only on the Nasdaq. The convergence between the Economist and the WSJ Weekend has established that the AI bubble has become a phenomenon of physical and financial markets alike — power grid, nuclear, gas, credit [2][3]. The repricing will be felt by balance sheets that never bought a single token.

Conclusion

In April, this publication wrote that infrastructure speed and regime speed did not coincide — that was, then, a qualitative tension. August brought the number: $2.9 trillion is the price, in David Cahn's arithmetic, of that compression [1]. What this month's sources add, read together, is the anatomy of the gap: it does not live in the technology, which works; nor in demand, which exists and is already reopening nuclear plants [3]; it lives in the interval between capital raised with the maturity of quarters and an adoption process that the management publications themselves, without citing one another, describe as structurally slow [5][6][7][8].

Financial history knows this pattern well: short-term debt was issued against long-term revenue, and it was called vision. When the difference comes due, it will say little about artificial intelligence and much about financial intelligence. The AI bubble, if it bursts in 2026, will not burst from an excess of future — it will burst from a lack of present tense in the calendar of whoever paid for it.

Frequently Asked Questions

Is there really an AI bubble in 2026? It depends on what is being measured. David Cahn's analysis (Sequoia) quantifies a gap of $2.9 trillion between the $1.5 trillion invested in infrastructure and roughly $80 billion in combined annual recurring revenue from OpenAI and Anthropic [1]. Physical-demand data — reopened nuclear plants, grid queues [3] — and the organisational diagnosis from HBR and MIT SMR [5][7] suggest the imbalance is one of capital duration, not technology validity.

What does Sequoia's $2.9 trillion gap mean? It measures the difference between capital committed to AI infrastructure and the revenue that infrastructure generates today [1]. It does not measure a lack of demand: it measures demand with the wrong maturity — buying companies are held back by scattered pilots, misaligned operating models and immature risk governance [5][6][7], so revenue arrives at the pace of decades while the liability matures at the pace of quarters.

If the bubble bursts, does AI fail? Probably not. A duration correction invalidates the capital structure and preserves the asset — like the overbuilt optical fibre of 2000, which changed hands at a discount and went on to sustain the entire internet. The correction would punish whoever financed it on the wrong timeline, not the technology [1][8].

How should companies prepare? Concentrate AI adoption in a defensible area instead of multiplying pilots [5], align operating model and risk governance before scaling [6][7], map indirect balance-sheet exposure to leveraged capital structures [2], and prepare liquidity and flexible leadership [9] to acquire capacity when the repricing arrives.

Related reading: "Sam Altman vs Mustafa Suleyman: The Compression Between Infrastructure and Regime" · "The Invisible Infrastructure of Decision-Making"