If a Stone Age analyst had modelled the market for fire, they might have started with an estimate of demand based on the human population multiplied by the number of cold nights. Logical, but limited. Early fire use was about warmth and warding off predators—but that was only the spark. Once harnessed, fire became a platform: lighting the night for storytelling, hardening tools, and cooking food—fuelling bigger brains, longer lives, and population growth. Fire was not just a source of heat; it was the original general-purpose technology.
General-purpose technologies are rare. In AI, we may have found another. Like fire, its value lies in second-order effects that no spreadsheet can predict. Even seasoned investors who navigated earlier such shifts—for instance, Uber surpassing global taxi revenues—are in unfamiliar territory. As Eric Schmidt said at TED recently, “the AI revolution is underhyped”—just as fire likely was in the Stone Age.
The physics of distribution matter. Technological progress is not linear; it is a stack of S-curves. S-curves depict a process that starts slowly, accelerates to rapid growth in the middle, then levels off toward the end. The 1960s saw semiconductors; the ’70s and ’80s, software—Apple, Microsoft, Adobe. The ’90s brought office networking and email. The 2000s saw the rise of the mobile internet, enabling Amazon, YouTube, and Facebook. Culminating in the 2010s, smartphones and cloud computing changed how technology is distributed. Uber and Airbnb were not just apps—they were inevitable outcomes of decades of compounded change in distribution.
When ChatGPT launched, no new hardware was needed. You just needed to download the app. In a world where capability scales instantly and distribution is frictionless, AI spreads exponentially. A child in an Indonesian village can access the same AI as one in Hampstead, UK. We do not yet know the implications of such developments—just as the printing press transformed the West but did not change the East.
When DeepMind’s AlphaGo beat the world champion at Go—a strategy board game—it did so with the now-famous move 37, unknown in 4,000 years of Go history. AI has since invented new molecules and enzymes never before seen in nature. We do not need artificial superintelligence for AI to transform the world—in its present state, it is already creating knowledge.
What is an investor to do? If AI is a general-purpose technology like fire, born into an era of near-infinite distribution and already performing tasks once thought impossible, should we simply invest in Microsoft (OpenAI), Google (DeepMind), and Meta (open-source AI)? While these giants will likely dominate, UK investors have three additional frameworks for profit.
Number one: picks and shovels. Like selling firewood in the fire revolution, building AI requires proprietary data and physical infrastructure. The UK hosts firms serving both.
Number two: vertical specialists. The savviest investor cannot become a neurosurgeon overnight. Domain knowledge and undocumented workflows power ‘vertical software’ firms. No matter how advanced large language models (LLMs) become, these vertical tools remain valuable until ‘tech-up’ LLMs meet ‘workflow-down’ software. That is why many UK assets in this space are M&A targets.
Number three: things that will not change. AI cannot yet read a room, or time a stock pick in a noisy market. Even in a world where it can, it will still need to pay or be paid. Fintech that reduces regulatory friction in payments is a fertile area—AI is additive here, not disruptive.
Damindu Jayaweera,
Technology Research Analyst
