# Christensen Clayton Christensen identified the mechanism that kills incumbents. Not bad management. Not laziness. Rational decision-making aimed at the wrong target. ## The Core Mechanism Incumbents optimize for their best customers. They pursue sustaining innovations — better margins, better features, better service for the accounts that pay the most. This works until it doesn't. A new entrant appears at the bottom of the market or in an adjacent niche, offering a product that looks inferior on every dimension the incumbent measures. The incumbent ignores it. Rationally. The entrant improves. The product climbs from "not good enough" to "good enough" for the next tier of customers. Then the next. By the time the incumbent notices, the economics have shifted. The entrant owns the volume, the cost curve, and the trajectory. The incumbent owns a shrinking premium segment. Christensen called this disruptive innovation. The word "disruptive" matters less than the asymmetry it describes: the incumbent cannot respond without cannibalizing its own margins. ## Two Disruption Paths **New-market disruption** creates customers who didn't exist. Personal computers didn't compete with mainframes. They served people who had no computer at all. The incumbent's framework had no category for them. **Low-end disruption** targets overserved customers. Discount retailers didn't offer better selection than department stores. They offered adequate selection at lower cost. The department stores retreated upmarket — and kept retreating until the viable market disappeared. Both paths share the same structural feature: the entrant competes on a dimension the incumbent dismisses. ## Where This Shows Up Now [[The Age of Vertical Models]] is Christensen disruption running its course in AI. Frontier labs optimize for generality — the sustaining innovation path. Vertical builders optimize for domain-specific correctness and win where operational trust matters. The frontier labs cannot respond without abandoning the generality that defines their product. [[Incumbent Bundling Risk]] describes the counterattack. Platform vendors ship "good enough" AI at zero incremental cost. Vertical builders survive only by going deeper than bundled features can reach. The moat is domain expertise encoded in evaluation data and training corpora, not model architecture. [[The Deep Tech Growth Cycle is different]] because atoms resist compression in ways bits don't. Deep tech disruption follows the same Christensen pattern but on longer timescales. The entrant must prove the technology works before it can prove the business works. Two innovation paths — tech-led and problem-led — must converge for commercial success. [[Consultancy-to-Platform Transition]] captures the most common failure mode for deep tech disruptors: delivering genuine value through services instead of software. Revenue grows linearly with headcount. The disruption stalls because the entrant cannot scale. [[AI era Defensibility]] maps the sequencing problem. Disruption gets you in the door. Staying power keeps you there. Win distribution fast (the bailey), then build depth (the motte). Speed without defensibility creates a flash followed by collapse. ## The Innovator's Dilemma, Stated Plainly Good managers, responding to good data, making good decisions, will systematically miss disruptive threats. The dilemma is not incompetence. It is the structure of rational optimization within an existing value network. The solution Christensen proposed: create an autonomous unit with its own P&L, free to pursue the disruptive opportunity without the parent's margin expectations. Separate the incentives and you separate the constraints. [[Execution x Evolution x Disruption]] gets at the operational version of this — you cannot change too many things too fast, but you also cannot leave out the thing that differentiates your product because it seems too disruptive. [[Key Insights from 80 Essays]] synthesizes the broader pattern. [[Wright's Law]] compounds the advantage — every doubling of cumulative experience compresses cost. The disruptor rides the learning curve. The incumbent watches from a position that looked strong and turned out to be a trap. [[Bottleneck Business]] identifies where disruption creates the most value: wherever the constraint shifts. The bottleneck was compute. Then it was data. Now it is domain expertise. Disruptors win by solving the current bottleneck, not the previous one. --- Related: [[The Age of Vertical Models]], [[Incumbent Bundling Risk]], [[AI era Defensibility]], [[The Deep Tech Growth Cycle is different]], [[Consultancy-to-Platform Transition]], [[Wright's Law]], [[Bottleneck Business]], [[Execution x Evolution x Disruption]], [[Key Insights from 80 Essays]] Tags: #firstprinciple #mentalmodels #kp #deeptech