AI as Infrastructure

What the UAE taught me about power

Jan 31, 2026

Dear readers and friends,

Welcome to the Human Override. Twice a month (or less), I share short reflections about AI and society, ending with the “Human Override” (strategies for leveraging AI, helpful tips, or ways to preserve humanity when everything turns synthetic). Paid subscribers get access to special essays and a monthly AMA session where you can ask me anything in a live webinar.

STORY

In the early 20th century, the US and China adopted electricity at roughly similar technical levels. The difference was spread. In America, electrification ran through factories, homes, and cities within decades. Output jumped. In China, the spread lagged. Over time, the gap compounded.

That analogy came from the Chief Strategy and Safety Officer at MGX, a $40 billion AI fund placing the UAE at the centre of the current industrial shift. I heard it late one morning in Abu Dhabi, the city turning white-ish yellow against the Gulf, inside a tower at ADGM on Al Marya Island.

Their working thesis was blunt: AI reshapes economies and power structures the way electricity did. Winners adopt early, push it everywhere, and build the capital, compute, and rules that make it diffusable.

I spent two weeks in a country I respect and feel grateful toward. I lived in Abu Dhabi for almost six years, which is long enough to see what wisely used wealth, long-term horizons, and state intent can actually build. Perhaps it’s the advantage of not betting on fast wins that wins consensus, or makes a CEO popular at the next quarterly meeting. That is the advantage of investing in systems meant to last.

If you’re not familiar with the UAE’s approach to artificial intelligence, one fact matters more than most commentary admits: in 2017, it became the first country to create a Ministry of AI. Around the same time, Australia cut major funding for AI research. Timing has a sense of humour.

I went back recently. An unplanned trip that quickly turned into a wonderful event. I lectured at NYU Abu Dhabi on the political, economic, and social power of AI. I visited MBZUAI, the first university fully focused on AI, Khazna data centres (part of the Stargate buildout), the Ministry of AI, and MGX itself.

I came back with a bit more excitement about the opportunity of AI, a more hopeful view, but mostly, with a cleaner way to think about power.

The real centre of gravity

Most public debate tracks the next system or the next personality. People argue about agents, home setups with many Mac minis, and screenshots going viral (most made by AI… so, hard to tell what’s true and what’s not). After a few hours with MGX’s leadership, that frame felt small.

The contest that shapes AI’s future is economic, geopolitical, and infrastructural. The real centre of gravity sits where three forces meet: long-horizon capital, national infrastructure with policy permission, and deep use across institutions. That intersection matters more than model releases, celebrity founders, or viral tools.

MGX doesn’t talk about AI as a product cycle. They treat it as a general-purpose technology in the economic sense. Like electricity, the steam engine, the internet.

And the UAE’s position becomes clear when you see the combination:

The state wants early and deep use, backed by money and policy.
Capital is patient. Decades matter more than quarters.
Rules are clear enough that projects don’t stall for years.

You don’t see that combination often. Silicon Valley builds systems. Places like this build the conditions that let those systems spread. By the end of the meeting, it was obvious this wasn’t for show; rather, it was a sound plan.

THE HUMAN OVERRIDE

If AI is changing how your industry works, five things matter.

First, treat AI as infrastructure.

If you see it as another tool or feature, you’re missing the point. It sits underneath how work gets done, how decisions are made, and what kinds of output are possible. Ask where it belongs at the core (operations, decision flow, products). Look for areas built on repeated mental labour and experiment with them systematically. It takes a lot of trial and error to realise the potential

Second, depth beats being first. History doesn’t reward early moves without follow-through. Map where AI actually sits in your organisation: core processes, support functions, or optional add-ons. Push it from the edges into the centre within a year.

Third, match your timeframes to the work. Short cycles don’t build infrastructure. If you’re funding AI internally or externally, judge it over two to three years. Measure how much it changes workflows and decisions, not just ROI in a few months.

Fourth, clear permission matters. Progress slows to a crawl when rules are vague, or overly cumbersome. Know what’s required where you operate. Start pilots in grey zones with strong internal controls, and push for standards that remove doubt.

Fifth, choose partners for reach. Convenience or your schoolmate as selection criteria doesn’t cut it. You’ll pay the price later. Capital at this scale pulls ecosystems together. Work with partners who give you access to compute, talent, and long-term alignment, and structure a deal with balance. The right relationship spreads use. The wrong one just adds features, or will colonise you.

SPARK - The frame matters

If AI becomes the base layer of this century, those who push it deepest into their institutions will shape what comes next. Using AI is not the same as being reshaped by it.

If your leaders are still thinking of AI as a way to boost productivity or replace jobs, they have the wrong frame. I wish fewer people focused on how to do the same things faster, and instead chose to explore and relentlessly experiment with what else becomes possible. Hours saved in inefficient meetings, communications, and decks can go into speaking with customers, creating new products, solving new problems, and entering new markets.

If AI is successful, the addressable market expands. If you’re blind to the opportunity, it won’t.

This is the first issue in a series on global AI strategy, economic adoption, job impacts, and institutional responses, including when regulation helps and when it doesn’t. This one is open. The next issues are for subscribers.

I’m also opening up the AMA session in February for everyone (usually reserved for paid subscribers only), so you can get a sense of what you're missing out. Sign up here.

Consider supporting Human Override:

If this was useful, reply and tell me what you want the next piece to answer.

Discussion about this post

Ready for more?