Every great technological revolution begins with the same illusion: that we are watching a single story unfold. The railway boom of the early 1870s looked like a story about transportation; it turned into one about land values, capital markets, and the economic development of the United States. The internet looked like a story about the growing influence of telcos and online retail; it became one about the destruction of newspapers, the reinvention of advertising, and the rise of platform monopolies. AI, in 2026, looks like a story about productivity and automation, but will ultimately be one about shifting value pools—and who captures them as AI rewrites competitive advantage.
For C-suite executives and boards, the strategic questions this raises are simple to state but hard to answer. Who will capture the value AI creates? Which advantages built over decades will survive, and which will quietly hollow out? And which AI-native competitor is being built to extract your margin? Companies that avoid these questions will find the market answering them instead, and by then, it will be too late to respond.
The Capital Reallocation Has No Modern Parallel
The acceleration of the past 12 months has been unmistakable. Frontier models have moved from confident assistants to genuinely autonomous agents—booking, coding, researching, and transacting with diminishing supervision. More than 65% of S&P 500 companies now reference AI on earnings calls. Hyperscalers' capital expenditures are on track for $750 billion in 2026 alone. Measured cumulatively from 2026 through 2031, the AI build-out will likely become one of the most concentrated reallocations of capital in modern economic history: in terms of capital expenditure as a percentage of GDP, it is several times the intensity of the telecom industry’s build-out at the turn of the millennium. (See Exhibit 1.)
So far, the spoils of the AI boom have been narrowly distributed. The biggest winners are the companies supplying or controlling AI’s physical foundation: chip designers, memory and equipment makers, and gas-turbine and grid suppliers with order books stretching beyond 2029, as well as utilities and landlords with access to scarce grid-connected power.
But equity markets have already begun voting on the next chapter. The derating of legacy enterprise software has wiped some $2 trillion in market value off companies whose subscription moats once looked impregnable. At the same time, values of IT services, media, and outsourcing firms have compressed on margin fears that have not yet shown up in reported earnings. Markets are not waiting for AI to disrupt these sectors; they are pricing the disruption in advance. To know whether that verdict applies to your company, you first need to understand where the value from AI actually flows—and to whom.
The derating of legacy enterprise software has wiped some $2 trillion in market value off companies whose subscription moats once looked impregnable.
Who Will Capture the AI Dividend?
AI is creating real value, but the distribution of that value will be unequal. Some sectors will expand their profit pools; others will see them compress. Within a sector, value will shift among incumbents, attackers, and the wider value chain. And a new actor—the compute layer—now takes a share of every AI-enabled transaction in a way that did not exist a decade ago.
The arrival of compute as a structural beneficiary is the genuinely new feature of this cycle. The closest analog is electricity at the onset of the 20th century, when utilities became a permanent intermediary in every industrial process. AI's compute layer plays the same role: a claimant on the output of every sector that adopts it, and for most companies, a permanent new line in the P&L.
As for the value within a sector, five structural factors determine who captures it and how quickly it moves. Three govern direction, determining who gets the value: task substitutability, market concentration, and data defensibility. Two govern speed and leakage, determining how fast the transfer happens and how much escapes the sector altogether: compute intensity and regulatory friction. (See Exhibit 2.)
This framework produces strikingly different outcomes across sectors. (See Exhibit 3.) In financial services, oligopoly and regulation has let shareholders retain most of the surplus, at least until the regulatory moat narrows. In retail, hypercompetition forces almost all of the AI dividend through to consumers within a handful of years. In health care, value flows to consumers, while in pharmaceuticals, it also goes to shareholders as AI reduces discovery cost. Across sectors, there is a broader redistribution of value: compute and AI providers benefit in most markets, employees often face pressure, and AI can either expand or shrink the market itself. The same technology produces opposite outcomes depending on the rules of the game.
In retail, hypercompetition forces almost all of the AI dividend through to consumers within a handful of years. In health care, value flows to consumers, while in pharmaceuticals, it also goes to shareholders as AI reduces discovery cost.
Which Advantages Will Survive?
The changes occurring within each sector are sharper and faster than most strategy reviews are capturing. Because the same foundation models and inference economics are broadly available, the underlying technology converges to parity within two to three years. As a result, durable advantage migrates downward, to the assets and structures beneath the technology.
This enables incumbents with true scale, flow data, trusted relationships, or physical-asset advantages to pull ahead. AI-native attackers, unburdened by legacy, are rising beneath them. Midscale incumbents—large enough to carry legacy cost but too small to own a data flywheel or network scale—are compressed from both sides. Two companies in the same sector, facing the same macro scenario, can end up on opposite sides of the same profit-pool shift.
The conventional sources of competitive advantage can be mapped across two dimensions: the depth of the protective moat and its durability under AI. (See Exhibit 4.) The most precarious quadrant is the bottom right, with advantages that were previously deep enough to feel permanent but are now fragile: knowledge and expertise, stock data, and operational excellence. These are the advantages most likely to look intact from the inside while eroding from the outside. The most valuable quadrant is the top right, with deep moats that AI is making more important rather than eroding, such as physical assets, flow data, network scale, and trusted relationships.
In the left-hand quadrants, headcount scale, once a proxy for capability, is now fragile and shallow; capital remains durable but provides less protection. Regulatory position sits at the center of the matrix—the advantage, while meaningful today, is narrowing.
Each of the durable-deep advantages in the top right quadrant shares a common logic. It compounds with use, deepens with scale, and becomes more valuable as AI raises the cost of replicating the advantage from scratch:
- Flow data—continuously generated by a product, feeding an AI that improves it and pulls in more users—compounds in a way that stock data cannot: Visa's VisaNet processes more than 257 billion transactions a year, informing fraud and authorization models that improve with each one.
- Trusted relationships become more valuable as AI commoditizes the analysis delivered through them. For example, in regulated sectors, verifiable AI governance, such as Mayo Clinic Platform's independent validation of clinical AI models, earns pricing that mere compliance cannot.
- System-level integration turns single-process tools that every competitor can buy into a proprietary workflow. For example, BlackRock's Aladdin platform was a moat before AI, and embedding AI deepens it.
- Network scale and scarce physical assets are reinforced by AI. For example, Amazon's million-robot fulfilment network, or Caterpillar's autonomous-mining fleet running on its own software, are moats that AI makes more valuable.
The most dangerous dynamic of all is the one that does not show up in the headline numbers: hollowing. That is, revenue, market share, and customer retention can all look intact while the economic substance—the margin and the pricing power—drains away. The most common cause is an AI-native attacker that extracts the high-margin layer and leaves the rest, though commoditization and the compute layer's rising claim produce the same effect. Relationships may endure, but the advisory margin attached to them is unbundled by an AI-native platform. Similarly, a dataset may still look large, but its defensibility has already been breached. The moat looks intact from the inside; from the outside, it has been crossed.
Revenue, market share, and customer retention can all look intact while the economic substance—the margin and the pricing power—drains away. The most common cause is an AI-native attacker that extracts the high-margin layer and leaves the rest, though commoditization and the compute layer's rising claim produce the same effect.
Which AI-Native Competitor Is Targeting Your Margin?
The structural advantage AI-native attackers hold is organizational, not technological. A company built from scratch around AI-era economics carries no embedded headcount, committed real estate, or inherited workflow. This produces a cost base in knowledge-intensive functions far below that of an incumbent and reduces the minimum viable scale at which a segment becomes profitable. AI-native attackers also have an edge in attracting talent fluent in these new ways of working—people who can move quickly from idea to deployment and participate more directly in the upside they create. The gap is structural: bolting AI onto a legacy operation closes only part of it, because architecture and talent, rather than tools, drive the advantage.
The standard incumbent playbook assumes an entrant will use that advantage the obvious way—doing what the incumbent does but at lower cost, to take market share. Instead, the most dangerous attackers focus on the highest-margin component of the value chain, deliver it as a standalone product, and leave incumbents with the low-margin remainder. Because they are competing so narrowly, share is the wrong unit of measure for assessing the threat.
The digital-era precedents make the pattern clear. In banking, Chime owns the customer relationship and the interchange economics, while the chartered banks beneath it hold the deposits, balance sheet, and payment infrastructure. Retail media networks (such as Amazon’s) allow the platform that controls what the customer sees to extract advertising margins of 70% to 90%, while the retailer fulfills the actual product at low single-digit margins. Neither is AI-driven, but both illustrate the mechanism: capture the high-margin layer without taking the revenue base.
AI now runs the same play. In industrials, platforms like Augury optimize uptime on equipment they do not build, positioning themselves to capture the after-market service margin the OEM has historically owned. In customer service, AI agents like Sierra and Decagon resolve interactions autonomously and set prices based on the outcome: per successful resolution rather than per seat. This allows them to compete for the margin in the resolution, leaving the incumbent with the residual of hard, escalated cases that still require human intervention. In each case, the attacker competes for margin, not revenue—invisibly because the competitor set is defined by share.
Think Like a Founder, Not a Steward
The diagnosis points to a clear conclusion: AI demands an entrepreneurial response, not a managerial one. Leaders cannot treat the task as optimizing the existing business. They need to think like founders: rethink the strategy and business model from first principles, make asymmetric bets where the analysis points, secure scarce resources before rivals do, and move the organization at a pace incumbents rarely sustain.
AI demands an entrepreneurial response, not a managerial one. Leaders cannot treat the task as optimizing the existing business.
Five imperatives should set the agenda:
Audit your competitive advantages and tailor your response. Consider whether each of your advantages is being strengthened, eroded, or hollowed by AI. Invest behind what is amplified; shift weight away from what is eroded; and act urgently to protect margins in what is hollowing, before the damage shows up in the P&L.
Defend against AI-native businesses seeking to extract margin. Pinpoint each attacker's model, the segment it would target, and the highest-margin slice of your value chain it could extract and sell on its own. Then build defensive postures that preserve revenue share, such as tighter customer integration, stronger data access, smarter bundling and pricing, or partnerships that make the attack harder to scale. The most dangerous attacker is usually the one your competitor dashboard was never built to show.
Make bets with your eyes open. Pressure-test your strategy against the uncertainties that could force a rewrite: compute that proves more expensive than your business case assumes; equity-market repricing that’s sharper than your capital plan allows; or access to frontier models that become conditional on geopolitical factors outside your control. A bet that pays off in only one scenario is a gamble, not a strategy. The next article in this series delves into the uncertainties a robust strategy must both withstand and exploit.
Redesign the organization for AI-first execution. The bottleneck to executing an AI strategy is rarely the technology. More often, it is talent, workflows, and culture: the people who can build AI-native operations rather than supervise the old ones; processes redesigned around what AI now does rather than bolted onto it; and a culture that rewards the asymmetric bet over the avoidance of error. The incumbents that struggle will not be the ones that lack the technology; they will be the ones whose organization cannot metabolize it.
Articulate the narrative for shareholders. Identify the AI thesis already embedded in your share price and ask whether it is the one you would write. If your multiple has compressed relative to your sector, the market sees you as a loser in the redistribution; if it has expanded, it sees you as a winner. Either way, name the narrative that the buy-side is using and decide to reinforce or change it. In the absence of a clear signal from you, the market will write the story itself.
Technological revolutions are often misread by the people living through them—not because they ignore the technology, but because they misidentify the threat. In the AI era, the illusion is not that transformation is coming; most leaders have accepted that. The illusion is that the threat will be visible in time: on a competitor dashboard, in reported margins, or in the form of an obvious new attacker. Companies that see through that illusion will build the moats that matter and capture the returns AI makes possible. Those that do not may discover too late that they were strengthening defenses around a business model that the battle had already moved beyond.
