In 1986, Richard Foster of McKinsey proved that every technology follows a logistic S-curve toward a physical performance limit — and that the very existence of that limit is the structural opening for the next technology. The ceiling is always built in. The seed is always germinating below.
In 1955, RCA's vacuum tube division was the undisputed foundation of American electronics. Performance improved every year. Engineers were producing better, faster, cheaper tubes. Returns on R&D investment were solid. The company had invested hundreds of millions refining the physics of thermionic emission — and it was paying off.
The conventional strategic wisdom was equally clear: keep investing in what works. Technology, in this view, improves continuously with R&D effort. More investment, more improvement. Always. There is no ceiling.
This view was about to be falsified by the shape of a curve.
When you plot performance against cumulative R&D investment — not just for vacuum tubes, but for every technology ever studied — the same shape emerges: the S-curve, or logistic function.
The early phase is slow: fundamental principles are poorly understood, experiments are inefficient. Then comes a steep growth phase — returns on R&D are maximized, performance climbs fast. Then, inevitably, the curve flattens. Progress slows. The technology approaches a ceiling it cannot cross.
After analyzing sixty-one companies in seven industries over twenty-five years, Foster concluded this shape was not coincidence. It was a structural property of all technological progress. The flattening was not a management failure. It was a message from physics.
Every technology exploits a physical, chemical, or biological principle. That principle has a maximum performance limit no engineering effort can exceed.
The vacuum tube's limit was thermal: higher frequencies required smaller electrode gaps, but smaller gaps meant more heat, more fragility, harder manufacturing — a wall in the physics of thermionic emission. Other limits: the Carnot bound for heat engines (~44% peak efficiency). The Abbe diffraction limit for optical lithography. The grain-physics tradeoff in silver halide film. Each is a ceiling built into the underlying science.
As a technology approaches its ceiling, each incremental performance gain requires exponentially more R&D investment. The engineers push harder. The returns keep shrinking. The ceiling feels like difficulty. It is actually completion.
Here is the signature of a mature technology: engineers among the best in the world, working harder than ever, producing less and less improvement per dollar.
Foster tracked R&D investment against performance output across decades of data. His finding: when R&D efficiency — performance gain per dollar — declines consistently for three or more periods, the technology is almost certainly on the upper portion of its S-curve, approaching its physical limit.
The bars on the chart grow taller (more R&D). The curve barely moves (less gain). This is the trap. The incumbent is not failing — it is exhausting the potential of its physical principle. The ceiling is close. And somewhere else, on a different physical principle entirely, something new is starting from zero.
In 1947, Bell Labs announced the transistor. On virtually every metric vacuum tube customers cared about — power output, high-frequency reliability, cost, manufacturing consistency — transistors were worse. The first commercial transistors cost $16 each. They drifted. They broke. They couldn't handle the frequencies radio engineers needed.
RCA's engineers evaluated them and concluded, correctly by the evidence, that transistors posed no near-term threat. This assessment was accurate. It was also strategically catastrophic. The transistor wasn't starting where the vacuum tube was. It was starting a completely new S-curve — one built on solid-state physics, with no thermionic emission limits. Its performance ceiling was orders of magnitude higher. Its entire steep growth phase was still ahead.
The seed was planted. The gap was 91 units. The trajectories were on different curves entirely.
Through the 1950s, transistors improved fast. Costs fell from $16 to $2. Reliability improved dramatically. Texas Instruments introduced the transistor radio in 1954 — a new market the vacuum tube manufacturers' customers didn't serve. Hearing aids, portable devices, early digital computers: applications where small size and low power mattered more than raw frequency performance.
The old technology was still better for incumbent applications. The new technology was building a foothold in new markets while climbing its own S-curve. RCA looked at the transistor radio and saw a toy. Texas Instruments looked at it and saw the bottom of a growth phase.
The gap between the two curves was closing — not because the vacuum tube was getting worse, but because the transistor's improvement rate was steeper, from a lower base, with a higher ceiling.
Foster called this moment the discontinuity: the point at which the new technology's S-curve crosses the old technology's performance. For transistors and vacuum tubes, the crossing occurred progressively through the early-to-mid 1960s, with different application areas crossing at different times.
The crossing is not a gentle handoff. Once the new technology reaches the performance threshold for mainstream applications, the market can shift rapidly. Customers who were waiting have a product that meets their needs at lower cost and better reliability. The incumbent's entire market is suddenly in play.
Eighteen years from transistor invention to vacuum tube displacement. For most of that time, vacuum tubes were objectively better. Then, in a few years, they were irrelevant. The discontinuity is a structural event, not a sudden one — but when it arrives, it looks sudden.
Foster's central finding: the companies that dominated the old technology are systematically worse positioned to dominate the new one. Not because they are incompetent — but because their organizational capabilities, customer relationships, manufacturing infrastructure, and management values are all built around the old S-curve.
RCA invented foundational transistor patents. RCA had the resources and the engineers. RCA lost the transistor era to Texas Instruments and Fairchild Semiconductor — companies with no stake in vacuum tubes, no customers demanding vacuum tube improvements, no manufacturing lines to protect.
The attacker's advantage is structural: the challenger commits fully to the new S-curve. The defender must serve existing customers while exploring the new technology — a split commitment that optimizes for neither. The new ceiling is 75 points above the old one. The headroom belongs to whoever bets the new curve.
Sailing ships to steam (50-year transition). Vacuum tubes to transistors (18 years). Film photography to digital (25 years — Kodak invented the digital camera in 1975 and filed for bankruptcy in 2012). Mechanical clocks to quartz to atomic. 14-inch disk drives to 8-inch to 5.25-inch to 3.5-inch to solid-state. Each generation: inferior entry, steep improvement, crossing, displacement.
In disk drives alone, Foster documented the same pattern across five successive form-factor generations. Each started below the incumbent, each carried a higher performance ceiling, each eventually displaced the previous one. Each old technology's limit was the new technology's opening.
This is Foster's lasting insight: the ceiling is always the seed. The physical limit built into every technology at birth is not where progress ends. It is where the next S-curve begins. Three generations. Always the same structure.
This sentence — spoken with complete accuracy by the engineers of every displaced incumbent — also appears in the post-mortem of every major technological transition. RCA's vacuum tubes were still improving in 1962. Kodak's film emulsions were still improving in 1998. The 14-inch hard disk manufacturers were still improving in 1975, the year 8-inch drives entered the market. Each statement was factually true. Each was strategically irrelevant.
Foster's model explains why: the relevant question is not "Is our technology improving?" but "How much performance headroom does our physical principle have remaining, and what is the improvement trajectory of the alternative?" A technology can be improving and still be three years from irrelevance. An incumbent can be executing flawlessly and still be walking toward a cliff its instruments cannot detect.
Performance gain per unit of R&D investment as a function of position on the S-curve. As a technology matures toward its physical limit, each dollar of R&D yields progressively less improvement — the derivative of the logistic function approaches zero near the ceiling. The challenger technology, starting its own S-curve from the beginning, has its entire growth phase ahead, with R&D efficiency that peaks far above the incumbent's current level.
// performance gain per R&D dollar · logistic derivative · old tech approaching limit vs. new tech in growth phase //
Every technology carries the architecture of its own displacement from the moment it is invented. The physical principle that enables the technology also sets its maximum performance. That maximum is not precisely known at birth — it reveals itself only as engineers approach it, through the compounding experience of diminishing returns. But it is always there.
This reframes the relationship between success and vulnerability. A technology that has been refined and optimized through decades of engineering effort — that has been driven very close to its physical limits — is a technology that has nearly exhausted its S-curve. Its very excellence is its fragility. The more completely a technology has been optimized, the less improvement headroom remains, and the greater the structural opening for a challenger with a fresh curve.
Duration from a challenger technology's market entry — when it was demonstrably inferior on incumbent metrics — to the point of majority displacement. Each transition follows the same structure: an extended period where the new technology is genuinely worse, a relatively short period of rapid displacement once the crossing occurs. The duration varies widely; the structure does not.
// years from entry to majority displacement · approximate · sources: Foster (1986), Utterback (1994), Christensen (1997) //
This is Foster's deepest paradox. An industry that has optimized brilliantly for its current physical principle has, in doing so, built the most elaborate possible obstacles to switching to a new one. Specialized workforce, specialized equipment, specialized supply chains, specialized customers. The ecosystem that makes a technology economically viable is the same ecosystem that makes it difficult to abandon.
Foster's prescription was radical for its time: the role of management at a mature technology company is not to optimize the existing S-curve. It is to identify the next S-curve and invest in it — typically before the existing S-curve has fully plateaued, which means before the urgency is obvious, and while the new technology still looks inferior. The window for action closes precisely when the case for action becomes undeniable.
Select a historical scenario or configure custom parameters to explore S-curve dynamics. Adjust the old technology's remaining headroom and the new technology's growth rate to find the crossover point and assess the attacker's structural advantage.