AI and the Looming Competition for Margin
Philipp Carlsson-Szlezak and Paul Swartz (BCG) · article · 2026-07-30 Reader 🖥️
AI Unit EconomicsFrontier Model Convergence / CommoditizationIndustrialization of AIJevons ParadoxMoats & DefensibilityTechnology Diffusion & Transformation Curves
AI productivity gains are inherently deflationary — because a widely diffused technology confers no durable competitive advantage, the industries best at converting AI into productivity growth will see the fiercest price competition and margin erosion, making consumers rather than firms the real winners, so executives must invest anyway (the risk of abstention is asymmetric) while expecting the payoff to be survival, not profit.
practical · mixed — Builds a theoretical economic framework (three pathways to productivity growth, all deflationary) and grounds it in historical case studies (autos, agriculture, internet retail), but the piece exists to tell executives what to do today — invest anyway, expect margin erosion, measure output not inputs — so it is ultimately practical/prescriptive.
Structure
| Part | Role |
|---|---|
| Opening framing: the 'winner's curse' and the deflationary nature of technology | thesis statement |
| Three-pathway framework: same with less, more with same, new replaces old | theoretical mechanism |
| Auto industry and agriculture case studies (70-year and century-long deflation) | historical evidence |
| 'Deflation Will Be Uneven': competition inhibitors and amplifiers | qualification / counterweight |
| 'What This Means for Executives Today': five prescriptions | practical conclusion |
Key terms
- Same with less
One of three productivity-growth pathways: firms use AI to reduce input costs for a given level of real output, gaining a cost advantage that spurs price cuts and price competition. - More with same
Firms use AI to grow real output with steady inputs, creating a volume/supply advantage that drives price competition, especially when demand is inelastic. - New replaces old
Firms use AI to build entirely new business models, offering a temporary escape from the immediate margin battle — until diffusion lets rivals copy the innovation and the battle for share and margin resumes. - Competition inhibitors / Competition amplifiers
Paired vocabulary for the forces that slow (moats, low utilization, regulation, societal skepticism, market structure) or speed (size of the productivity gain, inelastic demand, non-scalable capacity) the translation of AI productivity into price competition. - Winner's curse
Repurposed from auction theory to describe how being best at converting AI into productivity backfires: the firms and industries that succeed most at using AI are the ones that suffer the fiercest margin erosion.
Propositions
- explicitA widely diffused technology like AI can only generate productivity growth through three mechanisms — reducing input costs ('same with less'), growing output with the same inputs ('more with same'), or building new business models ('new replaces old') — and because none of these confers durable competitive advantage once diffused, all three converge on margin erosion rather than margin expansion.conf: highsupportsSimon Wardley's Post — Industrialisation of ML: nine notes on what to expect · p9 — Wardley's prediction that SaaS players who haven't industrialised their offerings will get margin-squeezed is an instance of exactly the 'same with less'/'more with same' squeeze this article frames as a general law of AI-driven productivity growth.
- explicitThe industries best at converting AI into productivity growth are likely to see the fiercest price competition and margin erosion — a 'winner's curse' in which strong productivity growth can mean lower profits for firms even as it makes consumers the real winners.conf: highextendsHow the Escalating AI Wars Benefit You · p4 — Generalizes the AI-lab-specific claim ('pricing and capacity wars among AI labs benefit users') into a structural economic law that applies to any widely-diffused-AI industry, not just the AI makers themselves.
- explicitcontestedAI monopolists with deep, defensible competitive moats are unlikely to emerge at scale, because the underlying AI models are broadly accessible — everyone uses essentially the same models, so advantages built in applications on top of them are hard to defend.conf: highcontradictsChatGPT Just Became A Work Agent · p8 — This article dismisses durable AI moats as 'less likely' because base models are shared; the corpus's frontier-convergence concept argues advantage doesn't disappear but migrates to the product/harness layer — a more precise claim this article's blanket dismissal doesn't engage with.
- explicitA budding price war is already underway among AI model makers (ChatGPT, Claude, Gemini, Llama, Grok) as AI budgets come under strain, and AI labs will likely accept 'token deflation' to fight for market share.conf: highextendsHoward Yu's Post — American AI models went from 70% of usage to 30% in 12 months · p8 — The observed 'good enough and cheap beats best and gated' disruption pattern among AI model makers is a specific instance of this article's general claim that widely diffused technology cannot sustain price/moat advantages.
- explicitcontestedThe 'more with same' pathway erodes margins most severely when the resulting supply shock is durable and demand for the output is inelastic; the article implies that where demand is elastic instead, volume growth could in principle absorb the added supply without a comparable price collapse.conf: mediumcontradictsYou Don't Get to Create Anything (Still Burning: Beck + Randy Shoup on Jevons paradox and distributed systems under AI) · p1 — This article's margin-erosion mechanism for 'more with same' explicitly requires inelastic demand; the corpus's Jevons-paradox claim that cheaper cognition creates MORE cognitive work (elastic demand) implies that, by this article's own logic, cognitive/software industries should be relatively exempt from the margin erosion it treats as broadly applicable.
- explicitAI's deflationary impact on the broader economy will be slower than often cited, for five reasons: a lag between input growth (e.g. code volume) and output/revenue growth, low daily-usage rates despite broad adoption, restrictive regulation, societal skepticism, and market structures (e.g. medical care) that resist price competition.conf: high
- explicitThe bigger AI's productivity impact on an industry, the bigger the resulting deflationary/price-competition force — small efficiency gains won't trigger a price war, but large ones very likely will.conf: high
- explicitcorroboratedSitting out the AI race is not a viable option for executives, because the risk of abstention is asymmetric: if AI proves transformative, non-adoption is existentially risky, while if AI disappoints, the cost of having invested is merely competitively neutral, not damaging.conf: highsupports ←The Questions You Should Be Asking About the AI Bubble · p12 — Same author pair restates, in the bubble-risk context, the asymmetric-abstention-risk argument ('sitting out the AI race is not viable') from their companion HBR piece on AI margin competition.
- explicitOnly relative efficiency gains count competitively: delivering AI-driven efficiency gains is worthless if a firm's gains lag behind the leaders', so firms compete on the differential, not the absolute improvement.conf: high
- implicitBecause the same AI investment that plausibly builds competitive standing is also what drives industry-wide margin erosion, the AI investment decision is not really a choice about pursuing competitive advantage — it is a forced, defensive move where the realistic payoff is survival, not profit, and the whole industry ends up running faster merely to hold position.conf: medium
Arguments
Productivity growth from AI converges on margin erosion regardless of which of the three pathways a firm or industry pursues. strength: medium
Explicit premises:
- AI is widely diffused — everyone can access essentially the same models.
- Each of the three pathways (same-with-less, more-with-same, new-replaces-old) creates a cost, volume, or business-model advantage.
- Once a technology is widely diffused, any advantage it creates can be replicated by competitors.
- When an advantage is replicable, competitors race to copy it, intensifying price competition to capture share.
Implicit premises:
- The durability of a competitive advantage is inversely proportional to how widely diffused its underlying technology is — a claim asserted via examples (autos, agriculture) rather than argued from first principles.
- Firms behave as atomistic price competitors rather than coordinating (tacitly or otherwise) to avoid a price war — the piece never considers oligopolistic restraint, collusion, or regulatory limits on price competition as a check on the model.
Inferences:
- If the argument holds, first-movers who succeed at AI-driven productivity gains earn no durable reward for having moved first — the rational strategy shifts from 'invest to win' to 'invest to avoid losing,' a Red-Queen race the authors partially acknowledge in 'only relative gains count' but don't name as such.
AI's deflationary impact on margins will be slower and more uneven than the paper's own central thesis implies.
Explicit premises:
- A more than tenfold increase in code volume is only producing a 1.3x increase in product releases (the input-to-output translation lags).
- 50% of workers use AI but only 13% use it daily — low utilization caps realized productivity gains.
- Regulators can block deployment outright (NYC declining to renew Waymo licenses).
- 50% of Americans are more concerned than excited about AI (Pew, March 2026) — societal acceptance is not assured.
- Some markets (e.g. medical care) are structurally resistant to price competition regardless of underlying cost changes.
Implicit premises:
- These five frictions are treated as slowing the deflationary dynamic rather than durably blocking it in any given industry — but the piece never states whether or how a friction converts from 'delay' to 'permanent exemption.'
Inferences:
- If some of these frictions are structural rather than transitional (e.g. medical care's market structure, or regulation in transportation), certain industries may never experience the margin battle the article's headline thesis predicts as near-universal — a possibility the piece doesn't rule in or out.
Narratives
- Autos: over ~70 years, technology steadily replaced labor (falling employment share from ~2% to 0.6%), prices fell, and margins that hovered near 30-40% decades ago have been whittled to near-zero today.
→ Empirically anchors the 'same with less' pathway and the claim that widely diffused technology produces long-run margin erosion. - Agriculture: mechanization, fertilizers, and genetic modification cut labor inputs so far that farm employment fell to 1% of the US workforce and food's share of consumer budgets fell from 40%+ to 13%, despite recent grocery-price inflation complaints.
→ A second, independent long-run case for the same 'same with less' deflation dynamic, reinforcing that the pattern is general rather than auto-specific. - The internet's effect on retail: it cut inventory and supply-chain costs, which let scale players use their efficiency and tight margins to force out competitors — and those same scale players then built new, durable revenue streams around advertising and third-party marketplaces.
→ Presented as an illustration of all three pathways operating simultaneously — but on inspection it actually shows surviving scale players (e.g. Amazon-like retailers) ending up MORE profitable and durably advantaged via new moats (ads, marketplace), not converged toward zero margin. See gap g1.
Questions addressed
- ✅ Will AI's productivity gains show up as higher profits for the firms that achieve them, or will they be competed away?
Competed away, via price competition, flowing to consumers as the 'real winners' — except where competition inhibitors (moats, regulation, market structure) block price competition from taking hold. - ✅ Should executives invest in AI given the prospect that success itself may erode their own margins?
Yes — invest regardless, because the downside risk of not investing (existential, if AI proves transformative) is asymmetrically worse than the downside of investing (merely a disappointing, competitively-neutral ROI). - ⭕ Which specific industries or firms will actually escape the deflationary dynamic and sustain durable moats or monopoly profits?
The piece says this outcome is 'less likely' in general and gestures at deep, hard-to-replicate deployments as the exception, but offers no criteria for identifying such cases in advance — and its own retail/internet example arguably shows scale players doing exactly this (see gap g1) without the piece noticing the tension.
Gaps
- illogicalThe piece's own retail/internet case study undercuts its 'no durable advantage' premise: it describes scale players using AI-like efficiency gains to force out competitors and then building new, durable revenue streams (advertising, third-party marketplaces) — i.e., a winner-take-most outcome with growing, not eroding, aggregate profitability for the survivors. That is closer to the corpus's documented pattern of durable advantage migrating to a different layer (see moats, frontier-convergence) than to the universal margin-erosion story the example is used to support.
Read next: Corpus material on moats and frontier-model convergence — e.g. how competitive advantage relocates to the harness/product/distribution layer rather than disappearing when the base technology commoditizes. - incompleteThe 'more with same' pathway explicitly conditions severe margin erosion on demand being inelastic, but the piece never examines which industries have elastic vs. inelastic demand for AI-driven output. The corpus's well-documented Jevons-paradox pattern for cognitive and software work (cheaper cognition/code has historically increased total demand for it, not just made existing output cheaper) suggests demand in exactly the industries most associated with AI (software, cognitive services) may be relatively elastic — which, by the article's own stated logic, should blunt rather than accelerate margin erosion there, a distinction the piece's mostly physical-goods examples (autos, agriculture) don't surface.
Read next: Corpus material applying Jevons Paradox to cognition and software engineering — cheaper cognition increasing total demand for cognitive/engineering work rather than shrinking it. - incompleteThe prescriptive 'What This Means for Executives' section treats the deflationary thesis as urgent enough to mandate immediate, full-scale investment for every executive, without differentiating strategy by how fast or slow deflation is likely to arrive given the very frictions the piece itself lists (regulation, market structure, utilization rates). A firm in a friction-heavy sector (the piece's own medical-care example) gets the identical 'the race is mandatory' prescription as one in a friction-free sector, even though the auto and agriculture case studies show the same dynamic can take seven decades to fully play out.
Read next: None specified by the author — an open strategic question about how urgency should scale with an industry's specific competition-inhibitor profile. - incompleteThe auto-industry chart's claim that margins fell 'to near 0%' doesn't specify which margin metric (gross, operating, net) is plotted or whether it's an industry-aggregate accounting figure vs. actual firm-level profitability — making the empirical anchor for the central thesis difficult to independently verify or compare against known auto-manufacturer profit margins.
Read next: Primary BEA/BLS/BCG Center for Macroeconomics data cited as the chart's source.
Closes gaps elsewhere
- What Makes Chinese Companies So Competitive? · g2 — That gap flags that hyper-competitive excellence (China's dense industrial ecosystem) coexists with chronic overcapacity and margin-eroding price wars, without explaining the mechanism. This article supplies the general economic mechanism: widely diffused productivity-enhancing capability confers no durable advantage, so the very firms best at deploying it face the fiercest price competition and margin erosion — the 'winner's curse' — which is a general theory of why competitive excellence and margin erosion co-occur, though it doesn't address China-specific policy responses like anti-dumping tariffs.
Related documents
- ChatGPT Just Became A Work Agent
- You Don't Get to Create Anything (Still Burning: Beck + Randy Shoup on Jevons paradox and distributed systems under AI)
- How the Escalating AI Wars Benefit You
- Howard Yu's Post — American AI models went from 70% of usage to 30% in 12 months
- Baseten CEO Tuhin Srivastava on the AI Inference Crunch, Custom Models
- Simon Wardley's Post — Industrialisation of ML: nine notes on what to expect
- Andreessen Horowitz's Post
- Presentation: A Few Predicted Talks From QConAI 2030
- Travel Through the Lens of AI with Booking.com CEO Glenn Fogel
- GitHub - musistudio/claude-code-router: Use Claude Code as the foundation for coding infrastructure, allowing you to decide how to interact with the model while enjoying updates from Anthropic.