The Computation Revolution

Almost 250 years ago, in March of 1776, The Wealth of Nations was published. Written by Adam Smith, it was an early exploration of what we have come to call “the industrial revolution” in which wealth would be measured by a nation’s production and labor rather than its gold reserves. A critical component of the book was its political dimension and, though it was an early hypothesis, much of his thinking was directionally correct on how economic growth would shift power from clergy and royalty to a new kind of state where increasingly capital and labor would have a greater voice.

Looking at 2026 it seems to me that we have a similar shift occurring, what we may come to call “the computation revolution,” in which wealth will be measured by a nation’s, a company’s, or an individual’s computation rather than its production or labor. Much as the industrial revolution realigned politics, society, and economics the shift to valuing computation will also generate forces for realignment. Some of the current market behaviors which seem puzzling (the market cap of Nvidia or the compute commitments of OpenAI for example) make sense when viewed as bets on chokepoints and should be evaluated in the context of this coming realignment and not in the context of the past industrial age systems and structures.

This realignment doesn’t decouple the systems of value, just as they were not decoupled in the shift from a mercantile to an industrial economy. In the development of the industrial economy finance was still critical to a company’s and a nation’s success. Concentration of value creation in capital, which is to say the owners of productive machinery, did however change the importance and role of practices which had previously been considered normal. Wealth of Nations is in part an argument for free markets, and Smith specifically attacked national monopolies such as the East India Company, entrenched local goods producers who were protected by tariffs, and the broader “mercantilism” of the period which prioritized the accumulation of precious metals and trade controls, all of which largely accrued power to government.

The industrial age saw the emergence of a new power source, global corporations, and an uneasy truce developed between the capitalists behind those organizations and the labor required to operate their productive machinery. New forms of governing (democracy, communism, fascism) all emerged in part due to this realignment of power. All of what we take as “normal” today in the functioning of society, markets, investing, and the very basis of fiat currency, emerged from this transition to an industrial economy. Now we should expect to see another wave of change.

In the computation economy while production and labor (and access to money) will continue to play a role, advantage will flow to those who can reliably turn energy, data, and algorithms into useful cognition at scale. Machine intelligence will power physical robots which will run the factories, warehouses, transportation, and even stores and restaurants. Human labor will be considered a luxury good, much as hand-made goods are today considered a luxury vs. industrially produced goods. Machine intelligence will also transform information work, displacing much of what people do today in these roles. While it is comforting to envision a new economic stability in which machines and human are complementary, transit from our current reality and any such new state will be bumpy as our institutions grapple with the speed and scale of this disruption.

Governments, companies, and investors should be considering the “chokepoints” of this new economy. The valuations and investments in these areas indicate that some already are.

Manufacturing (fabs & tools): Advanced logic manufacturing is concentrated particularly at TSMC for leading nodes and the single-supplier reality of EUV lithography (ASML) could give states leverage through export policy and service restrictions. We’ve already seen licenses pulled and service updates controlled for China-bound tools, a live demonstration of policy power over access to advanced manufacturing capacity.
Accelerator dominance & software moats: The warfighting and intelligence edge of modern AI rests on GPUs/AI accelerators plus full-stack software. Market leadership by a few vendors (e.g., NVIDIA) creates a potential “compute cartel” dynamic. Who gets priority shipments when supply is tight, at what price, and on what terms. NVIDIA will continue as the dominant vendor even as ASIC/alt-GPU options grow.
Cloud concentration: A handful of (all US-based) hyperscalers deliver the majority of global cloud compute. It’s a single-point-of-failure risk for states that depend on them for critical services. Already across every region questions are being raised and investments made to create sovereign compute infrastructure. The U.S. has already made one attempt to monitor and potentially restrict foreign access to U.S. cloud compute for training large AI models. In a crisis, that becomes a rapid coercive tool: deny capacity, delay deployments, or force verified-use regimes.
Energy and interconnect chokepoints: Compute runs on power and fiber. AI-driven data-center demand is projected to more than double by 2030, with tight clustering in a few hubs. This will create grid stress and offer tempting targets for disruption. Subsea cable ownership is also shifting toward hyperscalers, raising questions about resilience and state leverage over physical routes.

Naturally given the calculation of a rapidly increasing value of cognition and a set of clear chokepoints for control of both the existing and prospective computation which supplies that cognition, the assets which sit at these chokepoints are going to be both highly valued and anticipated to continue to grow in value. A set of technical questions depending on how they are answered can bend these valuations but will not derail them (open weight models, increased chip manufacturing capacity, new chip architectures, etc.). Thus any prediction that assets are at peak value or that the levels of investment are unsustainable has to first wrestle with the question of the ultimate value of the output of these assets with respect to the productivity of the entire global economy. If OpenAI is able to reach $100 billion in revenue in 2028, it will be because their algorithms and data centers are supplying machine intelligence which creates $1 trillion in value for their customers.

Google DeepMind’s mission, as Demis Hassabis puts it, is “to solve the problem of intelligence and then use intelligence to solve everything else." The industrial era measured prosperity by tons of steel and hours of work. The computation era will measure it by accessible, trustworthy, low-latency compute, an abundance of intelligence which will power robotics as the new engine of production, provide the tools of creation for media and entertainment, and will generate innovations in biology, chemistry, and physics. A new world order will emerge from this transition to a computation economy.