Venture capital investors are returning to Ridley because AI is doing exactly what he predicted
There is a 2010 book circulating again in the most active venture capital funds in Silicon Valley. It is not an artificial intelligence manual, it is not a study on language models, it has no chapter on GPUs or transformer architectures. It is a book of economic history written by a British biologist who argued, with data stretching back to the Stone Age, that human prosperity is a direct consequence of the exchange of ideas among specialized individuals. That when ideas mix at scale, the result is nonlinear. That no linear projection of resources or limitations has survived history because technology always found a substitution that the models had not calculated.
The book is The Rational Optimist, by Matt Ridley. And the fact that investors with concentrated positions in artificial intelligence infrastructure are rereading it now is not an anecdote about bibliographies. It is a signal about how they are formulating the underlying thesis that justifies capital deployment under conditions of high uncertainty.
Alexis Ohanian, co-founder of Reddit and manager of the Seven Seven Six fund, recently posted that he was listening to it in audiobook format at double speed and that he could not shake the impression that humanity is close to an inflection point. The post generated agreement from other investors. What started as a note about personal reading became a broader conversation about the intellectual framework that is organizing capital allocation decisions in the current artificial intelligence cycle.
Ridley's argument as an architecture of investment
Ridley's central thesis is not complicated, but it has far-reaching consequences when applied to the present. His argument is that prosperity is not generated by harder work, or natural resources, or central planning. It is generated by exchange: when a person specialized in one thing trades with another specialized in something different, both come out better than if they had tried to produce everything on their own. When that mechanism operates at sufficient scale, ideas combine in ways that neither party could have anticipated, and the result is a productivity curve that systematically denies any forecast of collapse or stagnation.
Ridley documents this with data spanning centuries. The price of one hour of artificial light fell from six hours of labor in 1800 to a fraction of a second in the present. Global real incomes multiplied by nine since that same year while the population only multiplied by six. Every Malthusian prediction that population growth would outpace productive capacity was invalidated by an innovation that the models had not incorporated because it did not exist when the models were built.
What investors are reading into that history is a recognizable pattern. Large language models are not adding one more productivity point to mature sectors. They are operating as amplifiers of the mechanism Ridley described: they give every knowledge worker access to a synthesis of global experience, in real time, without institutional intermediaries. If the size and density of the network of ideas determines the pace of innovation, then a technology that expands that network massively should generate returns at fund scale across nearly every sector simultaneously. That is the structural bet. It is not optimism of temperament. It is a historical reading with portfolio implications.
Global venture capital investment in artificial intelligence companies reached $131 billion in 2024, according to the NVCA PitchBook Venture Monitor, representing approximately 38% of all venture dollars deployed globally. The point of comparison that optimistic investors use is the dot-com bubble of 2000: on that occasion there was also capital concentration, but the underlying infrastructure — from broadband penetration to mobile hardware — took nearly a decade to catch up to the investment thesis. The difference they argue now is that the infrastructure gap is closing in months, not years. GPU clusters, access via programming interfaces, and edge deployment are scaling at a speed that has no direct precedent in previous cycles.
Labor displacement as a dividend of specialization
One of the most frequent arguments against current optimism in artificial intelligence is job displacement. Estimates such as McKinsey's, which projects that generative AI could automate 30% of hours worked by 2030, circulate as a warning of a mass destruction effect. Investors who read Ridley arrive at a different conclusion from the same data point.
Ridley's framework on specialization says that new tools do not eliminate work. They reallocate it toward higher-value tasks while collapsing the cost of previous bottlenecks. That pattern repeated itself with agricultural mechanization, with the spreadsheet, with search engines. In each case, the initial alarm was about disappearing jobs. History recorded that what followed was a reconfiguration toward activities that the previous system could not address because the cost of coordination was too high.
Applied to artificial intelligence, the argument is that automating 30% of current hours does not destroy 30% of employment. It frees up human capacity for tasks that were until now inaccessible because they required too much time for preparation, synthesis, or coordination. An analyst who previously spent half of their week consolidating information can invest that same week in interpreting and deciding. A physician who spent hours reviewing clinical literature can dedicate that time to patient interaction. The argument is not that the change is painless — it is that the historical pattern shows that specialization enabled by new tools tends to create more value than it displaces, measured in terms of income, well-being, and the complexity of the resulting human activities.
What that argument does not resolve — and here the analysis must be honest — is the temporal distribution of adjustment. History vindicates the optimists over horizons of decades. Displaced workers operate on horizons of years. That tension does not disappear by reading Ridley, and the investors who apply his framework at fund scale are not necessarily equipped to resolve it at the social scale.
The condition that optimism needs in order to be fulfilled
Ridley is not an unconditional optimist. His book has a central counterexample that investors in the current cycle are citing with the same frequency as his main thesis: the Ming dynasty. China in the fifteenth century had a technological advantage in navigation, metallurgy, and agricultural production. It had the largest network of ideas in the world at that time. And then it deliberately dismantled that advantage by restricting maritime trade, closing borders to external exchange, and consolidating central control over the production of knowledge. The result was that Europe, with smaller but more open exchange networks, ended up capturing the next century of growth.
The analogy does not require much work to become contemporary. The regulatory fragmentation of artificial intelligence between the European Union and the United States, national mandates for the procurement of artificial intelligence technology, closed model ecosystems operating as proprietary silos — all of these are mechanisms that reduce the effective size of the network of ideas at the exact moment when it is supposed to be expanding.
For investors using Ridley's framework, this is the most serious systemic risk — not the bubble in valuations or the competition between models. The underlying bet on artificial intelligence returns depends on exchange remaining sufficiently open. If regulation fragments markets or if dominant models become closed infrastructure with restricted access, the mechanics that justify the optimism deteriorate. Not for reasons of economic cycles, but due to a structural contraction of the network density that Ridley identified as the determining variable.
That threshold — the point at which regulatory policy begins to act on the architecture of idea exchange in the same way that the Ming court acted on its commercial network — is where the optimistic thesis has its most serious breaking point. And it is also the threshold about which there is still not enough evidence to know how it will be resolved.
What rational optimism cannot do for founders
For founders who are reading the moment through the same investors recommending Ridley, there is one strategically useful data point and one that can induce error.
The useful one is that investors aligned with this reading are looking for companies that accelerate the combination of ideas across domains that until now operated in silos. They are not looking for products that automate a singular task with greater efficiency. They are looking for companies that act as network densification nodes: biology and computation, logistics and language models, financial analysis and autonomous agents. The market sizing question those investors apply is not what a given product can capture, but what fraction of the potential GDP growth attributable to artificial intelligence can materialize within a fund horizon. Goldman Sachs projected in 2023 that generative artificial intelligence could lift global GDP by $13 trillion. The investors who find Ridley's historical arc persuasive are implicitly responding that that number — or something close to it — is achievable.
The data point that can induce error is the confusion between the intellectual framework and operational execution. Ridley documents that the mechanism of prosperity is real and robust over long historical horizons. That says nothing about which specific companies capture value, on what timeline, under what margin structure, or whether the current artificial intelligence infrastructure has the unit economics necessary to sustain current valuations. The optimistic narrative is compatible with short-term capital destruction cycles. The great technological waves that Ridley cites were not linear for the investors who were inside them in real time.
What the pattern does point to is that the companies that built exchange infrastructure — not those that built content on top of that infrastructure — captured most of the value in each previous cycle. If that analogy holds, the concentration of capital in foundational models and agent platforms has more structural coherence than bets on vertical applications without network differentiation.
The displacement that this moment reveals is not that of one sector replacing another. It is that of an idea-combination mechanism operating at a speed that no existing institutional structure was designed to absorb, with investors using a book of economic history to justify why that should produce prosperity, and with a condition of openness that no single fund can guarantee on its own.
