2026-09-04

Emad Mostaque's Answer to AI Sovereignty: Locally-Owned Companies Modeled on TSMC's Founding

AIBusinessPolicy🌍 Global

Intelligent Internet, the venture Stability AI founder Emad Mostaque started after leaving that company, unveiled "the Champion" on September 4: a proposed local company, founded and majority-owned by the people and institutions of the place it serves, designed to own what Mostaque calls "the last mile of intelligence." His framing: "The cost of intelligence is going to zero. The value of human cognition will go negative. The economy will change forever. We need new institutions to navigate this transition."

The diagnosis, and where it comes from

Mostaque traces the idea through a self-published trilogy from Intelligent Internet: The Last Economy (on what happens once hiring more GPUs and robots becomes cheaper than hiring people), Intelligent Economics (rebuilding economic theory for a world where machines, not just people, are capable economic agents), and Common Wealth (on what people must legally and constitutionally retain — personhood, inspection rights, the ability to contest a decision). A fourth analysis, Machines, is cited for a specific claim: that roughly 80% of the commercial opportunity in robotics lies downstream of manufacturing, in integration and deployment rather than building the hardware itself. All four are Intelligent Internet's own publications — the entire intellectual foundation for the Champion pitch is self-sourced, not independently reviewed economic research, which doesn't make the underlying diagnosis wrong but does mean it hasn't been tested outside the company making the pitch.

How a Champion is actually structured

The mechanics borrow directly from TSMC's 1987 founding: Taiwan's government and local companies subscribed the founding capital and held a local majority, with Philips subscribing the remainder in exchange for technology. A Champion follows the same shape, adapted for AI. Its founding round is reserved for local people and institutions at a nominal valuation of one dollar (or local-currency equivalent), with ownership proportional to capital contributed. Everyone under 21 at the point of setup receives a combined 10% of the equity, vesting on their 21st birthday; every child subsequently born in the state receives a share of a further 0.5%, issued annually. Later, strategic investors — cloud providers, multilateral institutions — can buy in at market rate. The plan then calls for "an accelerated IPO" giving international investors "a ticker that tracks the intelligence penetration of the state" — a phrase that names an ambition without specifying a mechanism; how a public listing would actually track something as diffuse as "intelligence penetration" isn't laid out in the announcement.

What a Champion is meant to do, day to day

Three lines of business: an "agent for everyone," a personal AI assistant that can reach frontier models while keeping memory and permissions rooted in the local Champion rather than a foreign provider; "integration engineers," local teams paid to make AI actually work inside schools, hospitals, and public services; and "humanoids as a service" — local financing, deployment, and maintenance of physical robots, the piece of the robotics market the Machines analysis claims is the larger opportunity. All three run on what Mostaque describes as Intelligent Internet's "state-of-the-art, open-source stack," customized locally — with an explicit design requirement that a Champion must be able to swap out II itself, or any model or cloud supplier, without losing its own records, keys, contracts, or ability to operate.

The argument that lands squarely in this week's sovereignty debate

Mostaque's central claim is one this blog has been tracking from the government side all week: "Governments are right to secure compute, cloud, models and data, but a system can be hosted locally while its memory, relationships and value are governed elsewhere. Sovereignty at the top of the stack does not guarantee ownership at the last mile." That's structurally the same worry France's economy minister raised about relying on Mistral alone, the same tension Alex Karp pointed at when he called some sovereignty moves "un cinq à sept sans préservatif" — deploying a domestic company's software that still runs on someone else's closed model underneath — and adjacent to the EU's own funding push toward AI-cyber compliance and SME adoption. Mostaque's answer is a private-sector, ownership-based mechanism rather than a government procurement or funding one: instead of a state or a national champion company owning the stack, distribute ownership of the customer-facing layer directly to the local population, one Champion at a time.

What doesn't exist yet, and who's proposing it

As of this announcement, no Champion has actually launched. The company says it is "forming the first Champions now," each starting with a Founding Council before opening to the broader population it's meant to serve — an invitation, not a working example. It's also worth knowing who's making the pitch: Mostaque resigned as Stability AI's CEO in March 2024 amid a wave of executive departures, mounting financial losses reported at roughly $8 million a month, and board pressure from investors including Coatue and Lightspeed, both of which had pushed for his exit. None of that disqualifies a new institutional idea on its own merits, but a founder proposing an ambitious ownership-and-governance structure for how AI enters people's daily lives is worth weighing against how his last company's own governance and finances actually played out.

What to expect next

  • Watch for an actual first Champion to launch, and where — the difference between a Founding Council forming and a functioning local company with real revenue is the whole test of whether this works as described.
  • Watch for the "accelerated IPO" mechanism to get specific. "A ticker that tracks the intelligence penetration of the state" is evocative but currently undefined; a real public listing structure would need to say what it's actually pricing.
  • Watch whether any government treats this as a genuine sovereignty alternative rather than a private venture pitch — the argument directly competes with the state-led approaches this blog has covered from France and the EU this same week.