AI Fund Economics

make this a short paragraph in my voice: Here's my take on whether this makes economic sense long-term: **The basic math:** - $7 trillion fund × 5% annual return = $350 billion/year - $350B / 340 million Americans ≈ $1,030 per person/year That *sounds* clean. But the assumptions hiding underneath are aggressive. **The good:** - **Sovereign wealth funds actually work.** Norway's fund (~$3 trillion) finances ~20% of its national budget. Alaska's Permanent Fund has been paying dividends since 1982. The model isn't theoretical. - **AI is genuinely concentrated.** If the top 20 companies control half the market, a 50% ownership stake captures real economic power, not just diluted shares. - **A 5% return is conservative.** The S&P 500 has returned ~10% nominally over the long term. If this fund captures AI's growth trajectory, 5% is achievable. **The problems:** 1. **$7 trillion is a lot.** The entire U.S. AI market was ~$174 billion in revenue in 2025, projected to reach ~$976 billion by 2035. For the fund to be worth $7 trillion, the AI sector would need to grow to roughly $14+ trillion in market cap. That's plausible *if* AI becomes as big as the entire financial services industry, but it's not guaranteed. 2. **5% of $7 trillion = $350 billion/year.** That's roughly the entire current U.S. AI market revenue. The fund would be extracting more in dividends than the entire industry generates today. Even by 2035, that's ~36% of projected market revenue going to dividends alone (before any corporate reinvestment needs). 3. **Concentration risk.** Norway's fund owns 1.5% of the *entire global stock market* — thousands of companies. This fund would own half of the AI sector. If AI regulation shifts, if a new paradigm (open source, edge AI, AGI disruption) changes the landscape, the fund's returns swing wildly. 4. **The 50% tax creates a valuation shock.** If the government takes half of every qualifying company, the remaining shareholders' stakes are worth 50% less overnight. This could reduce venture investment in AI, slow R&D, and push talent to un-taxed sectors. 5. **The $200M revenue threshold catches a lot.** That's not just OpenAI, Anthropic, and Google. It catches Palantir, Snowflake, ServiceNow, UiPath, Datadog, and dozens more. The tax base is broader than headlines suggest, which means more companies face dilution. **Bottom line:** It makes sense if you believe AI will become a ~$10+ trillion sector over the next decade and the fund earns a market-rate return. It breaks down if AI growth plateaus, if the sector gets disrupted, or if companies react to the tax by shifting operations offshore. The $1,000/year check is politically compelling. The $350B/year extraction rate is where the economic tension lives. I'd want to see how the fund's governance commission balances dividend payouts against reinvestment in the sector — if they distribute 100% of returns, the underlying companies get starved of capital for growth.

Here’s how I’d frame it in my take: The basic idea looks clean on paper: a $7T sovereign-style AI fund earning ~5% spits out about $350B a year, which works out to roughly $1k per person annually in the US. The issue is that this only works if the assumptions are doing a lot of heavy lifting. Sovereign wealth funds can work—Norway and Alaska prove that—but they’re diversified across huge markets, not concentrated in a single fast-moving sector like AI. The scale problem is real: $350B in payouts is roughly equal to today’s entire AI market revenue, so you’re effectively extracting as much as the industry currently produces, before it even reinvests. On top of that, owning ~50% of the sector introduces serious concentration and policy risk, and the 50% extraction itself would likely hit valuations and slow investment or push activity offshore. So it really comes down to whether AI becomes a $10T+ durable sector that can sustain that kind of capital extraction without stalling itself out.