Every major technological revolution — from the railways to the internet — has followed the same four-phase script: excitement, mania, crash, then a genuine golden age. Understanding where we are in that cycle is one of the most powerful edges a long-term investor can have.
Carlota Perez is a Venezuelan-British economist at Cambridge University. In her 2002 book Technological Revolutions and Financial Capital, she studied 250 years of industrial history and found something remarkable: every major technological revolution follows nearly the same pattern, from the British canal boom of the 1790s to the dot-com bubble of the 1990s.
Her framework is not a trading signal or a market-timing tool. It is a mental map for understanding where we are in the long arc of technological change — and why the companies that win in the early excitement phase are often not the ones that capture the most value in the end.
Perez identified five major surges in the modern era, each powered by a new set of technologies that transformed how economies operate — and each following the same four-phase cycle.
Water-powered mechanisation, cotton textiles, iron foundries. Launched in Britain with Arkwright's mill.
Steam engines, railways, iron production at scale. Britain's railway mania of the 1840s is the first great speculative bubble on record.
Steel, electricity, heavy chemicals, civil engineering. Centred in Britain, Germany and the United States.
Fordism, oil, petrochemicals, highways. The 1920s boom and 1929 crash fit the Perez pattern almost perfectly.
Microprocessors, internet, mobile, software. The dot-com bubble (1995–2000) was the frenzy phase; the 2000s–2010s were the deployment golden age.
Deep learning breakthroughs, large language models, autonomous systems, physical AI. We appear to be inside this revolution right now.
Each revolution unfolds across two main periods — Installation and Deployment — each containing two phases. Understanding which phase you are in tells you a great deal about how assets are priced and where the real value is being created.
A new cluster of technologies emerges and disrupts existing industries. A small group of pioneers invest heavily. The opportunity is real but not yet visible to most people.
Financial capital floods in. Valuations decouple from fundamentals. Everyone wants exposure. The technology is real, but the prices reflect fantasies.
After the crash, production capital takes over from financial capital. The technology spreads throughout the whole economy. Real productivity gains materialise. This is the golden age — and the best time for long-term investors.
The technology saturates the economy. Returns on new investment decline. Financial capital searches for the next revolution. The seeds of the next great surge are planted.
A financial crisis, market crash, or regulatory intervention separates the frenzy from the golden age. It is painful in the short term — but it clears away the speculation and resets conditions for genuine, sustainable growth. The turning point is not the end of the story. It is the beginning of the best chapter.
One of Perez's deepest insights is the distinction between two types of capital and which one dominates each phase of the cycle.
The fifth technological revolution is recent enough to study in detail. It is also the most important template for thinking about AI.
Mosaic browser, early websites, Amazon and Yahoo launch. A small group of technologists and investors understood what was coming. Valuations were still modest.
The Nasdaq rose from 1,500 to 5,000. Any company with ".com" in its name attracted capital. Cisco, at its peak, had a market cap implying it would one day be worth more than the entire US economy. Pets.com and Webvan burned billions with no viable business model.
The Nasdaq fell 80%. Cisco dropped from $80 to $11 and never recovered to its peak (it is still below the 2000 high today). Trillions in paper wealth evaporated. But the internet itself — the fibre cables, the server infrastructure — remained intact.
Google went public in 2004. Facebook launched in 2004. iPhone in 2007. Amazon Web Services in 2006. Netflix pivoted to streaming. These companies built enormous businesses on the infrastructure laid during the frenzy — at rational valuations. Investors who entered here captured extraordinary returns.
Cisco provided the infrastructure for the internet revolution — and its stock fell 86% and never recovered. Google, Amazon, and Facebook built applications on top of that infrastructure — and created some of the greatest wealth in history. The companies that win in the installation phase are often not the ones that win in the deployment phase. The picks-and-shovels seller doesn't always get rich. Sometimes the gold miner does.
The AI revolution — anchored by the deep learning breakthrough around 2012 and the public launch of ChatGPT in late 2022 — appears to follow the Perez pattern closely. The question that matters most for investors is: which phase are we in?
AlphaGo, GPT-3, early transformer models. Academic breakthroughs visible to specialists. NVIDIA already profitable from gaming, AI a secondary use case.
ChatGPT moment, NVIDIA stock up 10×, every company claiming to be "AI-first", IPO boom in AI startups, valuations decoupled from fundamentals. Classic frenzy characteristics.
Not yet happened. A correction — whether driven by a macro shock, regulatory action, or simply disappointing earnings from AI-inflated companies — is a likely precursor to the golden age.
AI becomes infrastructure. Healthcare, legal, education, manufacturing are fundamentally transformed. The big winners emerge — and they may not be today's most-discussed AI companies.
The Perez framework helps distinguish between companies operating in the installation layer (building the infrastructure of the revolution) and those in the deployment layer (using that infrastructure to transform real industries).
| Company | Perez Layer | Analogy | Key Consideration |
|---|---|---|---|
| NVIDIA | Installation | Cisco of the internet (networking hardware) | Built the infrastructure; CUDA moat is real but history shows infrastructure leaders don't always become the deployment era winners |
| Anthropic / OpenAI | Installation | Oracle / SAP of the internet (foundational software) | Selling the "picks and shovels" of AI (API access); value is captured by the companies using the API, not necessarily Anthropic itself |
| Microsoft | Both | IBM — survived multiple transitions | Azure is infrastructure; Copilot in Office 365 is genuine deployment with measurable productivity gains — arguably the clearest proof of AI ROI today |
| Meta | Deployment | Google of advertising — used infrastructure to transform a market | AI ad-targeting improvement is directly visible in revenue. Most financially verifiable AI ROI among large-caps |
| Tesla | Both | Toyota meets Google — hardware meets AI software | FSD is a genuine deployment application (physical AI); Optimus could become Layer 6. But promises have repeatedly preceded delivery |
| Unknown future companies | Deployment | Amazon, Google, Facebook of the AI era | The companies that capture the most value in the golden age may be ones that don't yet exist, or aren't yet public, or are operating in industries we don't associate with AI today |
The Perez framework doesn't tell you what to buy. But it provides a powerful lens for understanding what you are actually paying for at each stage of the cycle.
Asset prices reflect the maximum possible optimism about the technology's future. The infrastructure companies (like NVIDIA) attract the bulk of capital. Valuations price in years of perfect execution. Returns are possible but rely heavily on the sentiment continuing — not on fundamentals. Value investors like Buffett, Munger, and Li Lu typically sit out the frenzy or remain very selective.
Infrastructure is now cheap or commoditised. Application-layer companies — those that use AI to transform real industries — can be valued on fundamentals: gross margin, NRR, free cash flow. This is where value investors find "great companies at reasonable prices." The Turning Point is emotionally difficult (it follows a crash) but historically the most rewarding entry point for patient, long-term investors.
Before investing in any company positioned as an "AI play," the Perez framework suggests three diagnostic questions:
Infrastructure companies (chips, cloud, foundation models) are essential to the revolution — but the historical pattern suggests they are not always the biggest long-term winners. Ask whether the company is selling shovels, or actually mining for gold.
Genuine deployment-phase companies show it in the numbers — improving gross margins, NRR above 120%, accelerating revenue growth driven by real productivity gains. If the AI value proposition is purely a story about the future, you may be paying a frenzy-era premium for deployment-era outcomes.
History suggests a correction is a normal — and ultimately healthy — part of the cycle. Companies with high gross margins, strong free cash flow, and genuine product-market fit tend to survive and thrive after the turning point. Companies that depend on continued sentiment expansion are most vulnerable.
Every technological revolution produces a frenzy that enriches financial capital and a golden age that enriches patient investors who wait for the deployment era — because the companies that transform industries are often not the ones that looked most exciting during the bubble.
The ideas in this article are drawn from Carlota Perez's original research, combined with investment frameworks developed by value investors including Warren Buffett, Charlie Munger, and Li Lu. If you'd like to apply structured investment frameworks to analyse specific companies, the Five-Framework Stock Analyser tool on this site walks through KK, Darwin, Magic Formula, Buffett/Munger, and Peter Lynch scoring in a systematic way.