AI can transform the economy while investors overpay for it.
Cheaper intelligence could benefit users while squeezing the companies building it.
If a bust comes, watch who can afford to buy the infrastructure.
The aftermath could concentrate power further.
Follow the incentives.
China today proposed export controls on 23 AI technologies, while leaving access through APIs and apps exempt.
The strategic distinction matters: selling access can preserve control.
For countries building essential services around foreign AI, the question goes beyond capability.
Who sets the price?
Who controls updates?
What happens if access changes?
A country can become more productive while becoming more dependent.
AI sovereignty should mean having credible alternatives, the ability to switch providers, and control over critical systems.
Follow the incentives.
AI may become most powerful when we stop noticing it.
Today we measure AI adoption by things we can easily count: ChatGPT subscriptions, Claude users, Gemini downloads, enterprise licenses.
But that may eventually become the wrong way to think about AI.
Consider the internet.
We don’t wake up and decide to “use the internet.” We open a banking app. Stream a movie. Order dinner. Check a map. Send a message.
The internet disappeared into the infrastructure of everyday life.
AI may follow the same path — except potentially much faster.
It will increasingly sit behind search engines, phones, financial systems, healthcare tools, customer service, software development, logistics, education and government services.
Eventually, millions of people may interact with AI dozens or hundreds of times a day without consciously deciding to “use AI.”
And that changes the important questions.
The question stops being:
How many people subscribe to AI?
And becomes:
How many decisions are being influenced or made by AI?
Which loan gets approved?
Which job candidate gets interviewed?
Which information reaches you?
Which medical condition gets flagged?
Which investment gets made?
Which piece of software controls critical infrastructure?
This is where the conversation about AI becomes bigger than technology.
Because when a technology becomes infrastructure, control of that infrastructure becomes power.
And if increasingly important decisions flow through systems built by a relatively small number of companies, the question isn’t simply whether AI makes society more productive.
We also need to understand where that productivity, information and decision-making power accumulate.
AI doesn’t have to “take over the world” for the world to change profoundly.
It may simply become part of everything.
And eventually we may stop noticing that it’s there.
That’s when understanding who controls it matters most.
The remarkable number here isn’t 98%. It’s 2.2%.
AI is already driving enormous investment in chips, data centers and electricity while direct paid household adoption remains tiny.
But there’s another hidden variable: paying for AI and using AI are increasingly different things.
AI is being embedded into search, phones, productivity software, customer service and eventually almost every digital product. Most people may never consciously decide to “subscribe to AI.”
They may simply wake up one day and realize it has become part of everything they use.
The biggest technology shifts often stop looking like technology once they become infrastructure.
The remarkable number here isn’t 98%. It’s 2.2%.
AI is already driving enormous investment in chips, data centers and electricity while direct paid household adoption remains tiny.
But there’s another hidden variable: paying for AI and using AI are increasingly different things.
AI is being embedded into search, phones, productivity software, customer service and eventually almost every digital product. Most people may never consciously decide to “subscribe to AI.”
They may simply wake up one day and realize it has become part of everything they use.
The biggest technology shifts often stop looking like technology once they become infrastructure.
@AndrewYang AI doesn’t need to eliminate entire professions to transform employment. If one worker with AI can produce what three workers produced before, the job may still exist — but fewer people may be needed to do it. That distinction matters.
@SteveRattner Companies don’t need AI to replace every worker. They only need it to make each remaining worker more productive. If output rises while headcount falls, the economic consequences are very real. Follow the incentives.
@PierceLilholt The bigger question isn’t whether “jobs disappear.” It’s who captures the productivity gains. If AI dramatically increases output while ownership remains concentrated, workers could face disruption even as the economy becomes wealthier.
@kevinnbass “AI won’t take jobs” may be asking the wrong question. If AI lets 5 people do the work that once required 20, the job still exists. The other 15 positions don’t.
@joeroganhq Democracy distributes political power. AI may concentrate technological and economic power. What happens when those two forces begin moving in opposite directions?
AI can create enormous value. But a bigger industry is not a safety test.
Homodeus examines the difference between optimism, capability, and safeguards—and the incentives shaping the race.
Follow the incentives.
#AI#AISafety#HomodeusIntel