AI Will Not Make You Rich

Jerry Neumann argues that generative AI is not the next microprocessor revolution but the next shipping containerization — a transformative, society-reshaping technology whose economic value will flow almost entirely to customers and consumers, not to the founders and investors building it. The piece is the best single-article framing of AI investment strategy I've read: historically grounded, analytically sharp, and willing to say explicitly what most AI commentary hedges.

Key Quotes

"Anyone who invests in the new new thing must answer two questions: First, how much value will this innovation create? And second, who will capture it?"

Neumann's analytical frame in one sentence. Most AI discourse obsesses over the first question and assumes the second answers itself. The entire article is an argument that they decouple — and that right now they've decoupled hard. #concept

"The lessons learned from investing in tech over the last 50 years are not the right ones to apply now."

This is the gut-punch. Every VC pattern-matching from the ICT era — bet on founders, bet on technical moats, bet early and hold — was calibrated on a particular phase of a particular wave. We are not in that phase. #pattern

"Success, in other words, will mean defeat — always a bad thesis."

On AI application companies: if Perplexity or Writer or Abridge becomes genuinely valuable, the model companies will extract those profits through discriminatory pricing or vertical integration. You're building a business whose success triggers its own destruction. Neumann is unusually blunt about how structurally broken this position is. #concept

"For decades, the way to make money was to bet on what the new thing was. Now, you have to bet on the opportunities it opens up."

The prescription: invest downstream, not upstream. Not in the companies building AI, but in the knowledge-work companies — professional services, healthcare, education, financial services, creative services — that will use AI to cut costs and expand. Same as containerization made IKEA and Walmart rich, not SeaLand. #pattern

"Big companies hate surprises. That's why uncertainty makes a perfect moat for startups."

Why the PC revolution created wealth and AI won't: in the 1970s, nobody was watching. IBM didn't care. The NYT mentioned PCs four times in all of 1976. That distributed, permissionless, no-one-is-looking experimentation is the essential precondition for an irruption phase. AI has had the opposite: every incumbent had a strategic plan on day one. #concept

Neumann's "uncertainty as a moat" gets a name in Moats — 80 Strategies for Competitive Advantage: Ango calls it "obscurity" (survival by remaining unknown or undetected) and "unpredictability" (surprise, confusion, variability) — both edges that work precisely because the incumbent isn't watching, and both of which AI's hype cycle foreclosed.

Key Themes

Carlota Perez's Wave Theory as Investment Framework

Neumann structures the entire argument around Perez's four-phase model: irruption (uncertainty, experimentation, hard to invest), frenzy (speculation rewarded, culminates in bubble), synergy (patient capital, applications, real businesses), and maturity (predictable, captured by incumbents, near-impossible to make new money). ICT went through all four; containerization was a maturity-phase innovation. AI is maturity-phase ICT. #concept

Containerization as the Cautionary Tale

The article's central case study. Containerization changed the world — enabled globalization, collapsed shipping costs, reshaped manufacturing geography. But exactly two people got rich from it: Malcom McLean (sold SeaLand to R.J. Reynolds in 1969 at the top) and Daniel Ludwig (8.5x return, sold same year). Everyone else — the shipbuilders, the port operators, the shipping lines, R.J. Reynolds itself — either lost money or made commodity returns. The value went to IKEA, Walmart, Costco, and consumers. #comparison

Rodney Brooks independently lands on containerization as his canonical example of Time Scale 4: technologies that reshape the economy invariably take 50+ years of continuous at-scale deployment, a human lifetime. Neumann's investment thesis and Brooks' time-scale framework are convergent — both argue the economic value of a transformative technology arrives on a schedule that no amount of venture capital can accelerate. Four Time Scales for Technology Development and Deployment

The PC Revolution as the Success Case

Neumann walks through the 1970s PC story in detail: Intel didn't see it coming, Wozniak built the Apple on a $25 chip, IBM ignored it until 1981, and the distributed tinkering of thousands of hobbyists discovered what the technology was for. This is the template for how a technological revolution should start — but it required surprise, permissionless experimentation, and time to build moats. AI has none of these. #comparison

The Seven-Point Rhyme Between AI and Containerization

Neumann maps seven structural parallels — the first four already underway, the rest his prediction: 1. Revolutionary technology, obvious upsides 2. Everyone sees it coming, immediate strategic response 3. Rapid competition, squeezed profits 4. Capex becomes enormous, consolidation begins 5. Model companies consolidate into oligopoly (his prediction) 6. Value flows to customers, not builders 7. Customers are knowledge-work industries

This is the most testable part of the article. If AI model companies keep raising at higher valuations and nobody forces consolidation, Neumann is wrong. If the consolidation happens, the rest of the argument snaps into place. #pattern

Where to Actually Invest

Neumann goes layer by layer through the AI stack and explains why each is a bad bet — except one: - Model companies: Too late, unless you're already in. Will consolidate into a few, owned by Big Tech. - App companies: Structural trap. Success invites extraction or acquisition. - Intermediation companies (Hugging Face, Glean): Will stay mid-sized; model companies won't let them get strategic leverage. - Infrastructure (Nvidia, data centers, cloud): Already priced for s-curve. If AI capex peaks and declines — as shipbuilding did after 1973 — these get crushed. - Downstream knowledge-work companies: The only good bet. Companies that use AI to lower costs in professional services, healthcare, education, financial services. But even here, the returns are modest: ~2% annual tailwind at best. #tool

Consumers Win

The cleanest takeaway: most of the $1-7 trillion in projected AI GDP gains will be captured by consumers through cheaper knowledge-intensive services (healthcare, education, advice). This parallels Baumol's cost disease in reverse — AI lets knowledge-work productivity rise, lowering prices in sectors that have only gotten more expensive for decades. #concept

Critical Analysis

This is a genuinely important article, and I mean that in the non-inflationary sense. It belongs on the short shelf of "things you should read before investing a dollar in AI."

Neumann's strength is that he's a retired VC — he's not selling anything, not raising a fund, not justifying his portfolio. He's just applying economic history to the present moment with the clarity that comes from having no stake in the answer. The containerization analogy is devastating because it's so precise: revolutionary technology, obvious benefits, no surprise, immediate competition, value capture by customers. Every beat maps.

The weakness is that the argument is almost entirely analogical. Neumann doesn't provide a mechanism for why AI must follow the containerization path rather than the PC path — he just shows that it rhymes so far and bets the rhyme continues. The analogy could break. If open-weight models become genuinely competitive with frontier models (as GLM-5.2 Is the Step Change for Open Agents suggests is happening), the "permissionless experimentation" precondition for a new irruption phase might suddenly be met. Neumann acknowledges this possibility — "when any would-be innovator can build and train an LLM on their laptop" — but treats it as unlikely. Given the speed of open model improvement, I think he underweights it.

The other gap: Neumann doesn't engage with the possibility that AI could be different from containerization because it's recursive. Containerization made shipping cheaper; AI makes everything cheaper, including AI research itself. That recursion could accelerate the consolidation he predicts, or it could make it impossible — if the cost of building competitive models keeps collapsing, the oligopoly never forms. He can't have it both ways, and he doesn't pick.

A third gap sits one level up from Neumann's frame. He asks who will capture the value AI creates; How to Get Rich in America documents that incumbent asset holders are already getting rich off AI before that question is settled — through the stock-market wealth effect, not through building or buying the technology. Distribution is happening now; capture is still unresolved. The two theses coexist, but only if you notice the distinction, and neither author does.

That said, the investment advice is hard to argue with. "Don't invest in model companies, don't invest in app companies, infrastructure is already priced in, go downstream to knowledge-work industries and expect modest returns" — this is exactly the advice every non-salesperson in AI gives privately. Neumann just has the historical framework to make it persuasive rather than cynical.

The most useful sentence in the piece might be: "For decades, the way to make money was to bet on what the new thing was. Now, you have to bet on the opportunities it opens up." That's the mental model shift. Whether you accept his containerization thesis or not, the shift is real.

Sources: AI Will Not Make You Rich Last updated: 2026-07-05