What if the figure everyone keeps calling the villain of the machine age is not the one who slams the brakes? I keep coming back to that question whenever the conversation turns to catastrophe, regulation, and a single global pause. Most people I talk to outside rich coastal cities are not losing sleep over a hypothetical superintelligence. They are worrying about power cuts, port delays, credit, tax collection, and a young workforce that needs jobs now. In that world the offer that travels farthest is not a moratorium. It is a complete stack that is cheap, financed, hosted, and already sitting on the phones, the docks, and the grid.
The Growth Offer Hidden Inside The AI Contest
I am not writing a sermon. I am trying to underwrite a map. The popular story says the contest is a beauty pageant of models. One side builds the most impressive systems. The other copies. A third writes rules. Capital then piles into closed labs, chipmakers, and giant clouds. The scoreboard is a leaderboard. That story is not fake. It is incomplete in a way that can misprice power.
It treats the richest buyers as the only buyers that matter. It treats evaluations as destiny. It treats national strength as a software demo. I disagree. The race that will linger is who becomes the default operating layer for the economies that still have the most growth left. Those places will not adopt machine intelligence as a lifestyle accessory. They will adopt it as a growth tool. They will take the package that shows up, attaches to what they already run, and does not blow the budget.
The more probable figure is not the one who promises safety. It is the one who promises growth.
Why The Comforting Market Story Feels Too Narrow
Closed labs can still look dazzling. Benchmarks can still stay Western. Safety papers can still multiply. None of that automatically decides who owns the installed base. If an emerging state can raise output, collect revenue, move goods, and train clerks with a cheaper foreign stack, prestige becomes a side show. I’ve found that markets love visible contests and underprice quiet lock-in. Switching costs do not announce themselves on launch day. They show up three years later when a ministry cannot afford an outage.
Think of it as rails, not fireworks. Ports and freight corridors were an earlier set of rails. Models, clouds, handsets, and financing can become the next set. Once logistics, schools, credit scoring, and tax systems sit on one stack, ripping it out stops being a procurement choice. It becomes a sovereignty choice. Sovereignty choices are not made on price alone.
Where The Mass Of Activity Actually Sits
Start with the denominator. On purchasing power terms, large emerging economies already represent an enormous volume of physical and administrative work. That is the work you can automate. Nominal dollars still flatter the United States when the question is who can buy foreign assets. Purchasing power tells you how much activity is there to wire. For an adoption thesis, the second number is the one that matters.
China is huge on both measures, but the gap looks different depending on the yardstick. India is already a giant on purchasing power. Indonesia and Brazil clear the five trillion range on that basis. Turkey, Mexico, Saudi Arabia, Egypt, and Nigeria are not footnotes. They are places with young populations, urgent growth targets, and thin administrative capacity. That combination selects speed.
| Lens | What It Answers | Why It Matters For AI |
| Nominal output | Who can buy foreign assets | Capital firepower and import capacity |
| Purchasing power | How much real activity exists | How much administration and logistics can be automated |
| Growth urgency | Who needs results this decade | Who takes the stack that arrives first and cheap |
Some observers describe a hierarchical order in which leaders travel, acknowledge relative strength, and trade access for stability. I take the framing seriously without treating it as scripture. The practical point is simpler. If American security promises look less automatic under stress, commercial rails become more valuable. Models can ride those rails.
What Download Data Already Shows
This part is no longer a thought experiment. Over a stretch of years, the American share of public model downloads on a major open repository fell from a dominant position into the mid teens. A Chinese model became the most liked release in that platform’s history. The top of the list is no longer majority American. Large Chinese firms went from almost no public releases to a flood in a single year. Among newly created models under a year old, downloads for Chinese weights surpassed any other country.
Western teams now routinely fine tune those base models because they are often the largest open weights available. That is not a morality play. It is a cost and availability story. A lab head in Jakarta put it bluntly: developers pick the cheapest one. A founder in Malaysia wanted the biggest model and said there was no Western open offering at that size. Chinese cloud footprints in parts of Southeast Asia were already dense, with more availability zones across more local regions than the Western field in some counts.
None of that shows up cleanly in a frontier benchmark table. All of it shows up in habits. Habits become defaults. Defaults become switching costs.
- Price and hostability beat prestige for many builders.
- Open weights travel faster than closed demos.
- Cloud presence near the customer shortens the sales cycle.
- Fine tuning on the largest available base becomes the path of least resistance.
The Institutional Layer Behind The Price Tag
Beijing is not leaving the volume layer to discounts alone. A new cooperation body stood up in Shanghai with dozens of founding members spanning Eurasia, Africa, and Latin America. The five year commitments attached to that kind of body are unglamorous, which is exactly why they matter. Training placements. Joint application centers. A weather early warning system pushed into many countries.
Read that list the way an underwriter would. Training placements create the administrators who write the next procurement. Application centers create reference deployments. Weather systems create dependency inside a ministry that cannot tolerate downtime. Standards get set that way. Not with a prettier chat window. With a bureaucracy that has already learned one stack.
Standards get set with a bureaucracy that has already learned one system, not with a better demo.
Washington Is Running Two Clocks At Once
The United States understands the problem on paper. An export push was designed to sell full packages abroad: chips, models, applications, cybersecurity, cloud, and data centers as a bundle. The contested markets that should sit at the center of that effort look a lot like the map in this piece: Brazil, Egypt, Indonesia, Nigeria, Thailand, the Philippines, Malaysia, Vietnam, Bangladesh. That is the right geography.
The tension is obvious. The same government running an export promotion program is running an export control program, and the second one often moves faster than the first. Partner status can be upgraded late and still come with scoped chip access. Meanwhile commercial vendors from the other side shop parts into the Gulf, Southeast Asia, and North Africa. A self sufficiency ambition on chips has been published on a near term horizon, with plans to expand domestic output.
Set aside whether the controls are correct on the merits. Underwrite the second order effect. Capability still leaks outward through commercial relationships and distillation. American open efforts can start constrained at home. The result can look like a one way street into the volume layer of the world economy, at the exact moment when that volume layer is where the standard gets set.
Western discourse spends a lot of attention on alignment theater, synthetic media, and white collar displacement in rich cities. Those are real problems. They are also rich country problems. The quieter failure is dependency. Once an emerging state’s logistics and revenue collection run on foreign models, leaving is no longer a software swap.
A Country Screen That Looks Like A Company Screen
When you screen a country the way you screen a firm, you are not asking which model it admires. You are asking what it has already installed, who financed the installation, and what it would cost to rip out. The map is wide and uneven.
- Asia: India, Indonesia, Vietnam, Malaysia, Thailand, the Philippines, Bangladesh, Pakistan, Kazakhstan, Cambodia, Laos, Sri Lanka.
- Middle East and Gulf: Saudi Arabia, the UAE, Egypt, Turkey, Iran.
- Africa: Nigeria, Ethiopia, Kenya, South Africa, Angola, Ghana.
- Latin America: Brazil, Mexico, Argentina, Chile, Colombia.
These are not equivalent cases. India can build more of its own stack. The Gulf can simply buy, and it is buying from both sides. Vietnam and Indonesia industrialize and will take whatever shortens the climb. Nigeria and Ethiopia need administration and power far more than they need chat interfaces. Brazil and Mexico live between Western finance and Chinese trade and will hedge. Pakistan and Kazakhstan sit on corridors that were financed once already.
The common variable is growth urgency. Growth urgency selects the stack that shows up. It selects it quickly. In my experience, that is the variable strategy memos still treat as a footnote.
Compute Is The Throat Of The Cycle
Compute is the oil of this cycle, and the supply chain has a throat. One foundry still dominates the pure play market, and its grip at the leading nodes is even tighter than the headline share suggests. You do not need an invasion scenario to price the leverage. You need governments that come to believe only one counterpart can reliably keep the chips, the cloud, the handsets, and the financing flowing. That belief is cheaper to create than a fab and harder to reverse than a tariff.
Perhaps the most interesting aspect is how ordinary this can look while it is happening. No dramatic speech. No cartoon villain. Just a procurement officer who cannot risk downtime, a finance ministry that wants GDP prints, and a vendor who already financed the last crane on the dock.
Two Futures Markets Are Not Pricing The Same Way
In the first future, America wins the cathedral. Benchmarks stay American. Safety papers multiply. Closed models remain impressive and expensive. Emerging economies still buy the stack attached to Chinese devices, Chinese capital, and turnkey infrastructure. Global token volume follows global output, which is to say it follows the parish. Prestige stays in one place. The installed base goes elsewhere.
In the second future, America treats emerging output as the actual battlefield. Competitive open weights exist and can be hosted by states that want an alternative without becoming a tenant of a single lab. Energy, chips, and cloud are treated as national goods rather than line items. Public companies that can deploy intelligence into durable operations are valued above the ones that can only demonstrate it.
Markets, to my eye, are priced closer to the first future than the evidence supports. That gap is the part I care about. It is also the part that is easy to miss if you live inside the cathedral.
The Same Bottleneck, One Order Of Magnitude Wider
Inside listed companies the pattern is already familiar. The technology works. Capital is available. Absorption is the constraint. A thin slice of giant firms captured a wild share of index gains after the first consumer wave, while hundreds of others lagged. Plenty of smaller firms claim a strategy. Almost none describe implementation as mature. That is a governance problem inside public companies.
Widen the aperture and it is the same problem at the level of the state. Intelligence is being manufactured at declining cost. Absorption is the bottleneck. Ask whether an organization can metabolize what it has already purchased. Now ask that question of a country with weak administrative capacity and urgent growth targets. The answer gets worse. That is precisely why the party that clears the bottleneck earns something more durable than a product cycle. It writes the rails under the next order.
Underwriting checklist I keep on one page: What is already installed? Who financed the install? Who trains the operators? What fails if the vendor leaves? How urgent is growth this decade?
What I Would Actually Underwrite
If you underwrite this field as a feature race between rich country labs, you can be right about the models and wrong about the century. Underwrite instead who owns the rails that India, Indonesia, Brazil, Mexico, Saudi Arabia, the UAE, Vietnam, Nigeria, Egypt, and Turkey will actually run. Underwrite who captures adoption where governance is thin and growth is urgent. Underwrite closed labs and national champions as though their real competitor is not the next chat interface, but a hierarchy that intends to make itself expensive to leave.
The contest that matters is not who builds the smartest model in the richest city. It is who becomes the operating system for the economies that still have the most growth left in them. That is not a trade ticket. It is a map. I would rather be early and slightly messy on the map than perfectly current on the leaderboard.
Will some countries hedge forever? Yes. Will some build domestic alternatives? Also yes. Hedging does not cancel lock-in. It only slows the moment when a system becomes the default. Defaults compound. That is the unromantic core of the story, and it is why I keep returning to ports, power, payroll, and tax offices rather than to keynote stages.
A Practical Way To Read The Next Three Years
Watch training programs more than speeches. Watch which weather, customs, and credit tools get deployed into ministries. Watch whether open weights large enough to matter remain usable outside a handful of clouds. Watch chip access that is scoped to approved entities while rival vendors keep shopping the same corridors. Watch energy and data center permits in places that still industrialize.
- Follow installed base, not admiration.
- Follow financing terms, not press releases.
- Follow operator training, not benchmark charts.
- Follow outage risk inside the state, not consumer apps.
- Follow whether leaving would break revenue collection.
I do not claim this is the only lens. It is the lens that keeps surviving contact with how states actually buy tools. Fancy systems lose to present systems when the median citizen is twenty four and the lights flicker. That sentence is not poetry. It is procurement.
If the last cycle taught investors to obsess over model quality, this cycle should teach them to obsess over metabolism. Who can absorb the tool, attach it to ports and payroll, and keep it running when the vendor’s politics get messy? That question is less glamorous than a demo. It is also closer to how power settles.
So here is where I land. Do not ignore the cathedral. Just stop mistaking it for the parish. The parish is where volume lives. Volume is where defaults form. Defaults are what the next order runs on. If that sounds too plain for a subject this large, good. Plain is how rails get laid.
