Why Investing in AI Mirrors the Container Shipping Boom—and Why Most Investors Lose Money

Clarifying a Key Point: Technological Revolutions Aren’t Synonymous With Wealth Creation

History shows two kinds of tech revolutions. The first creates massive new wealth—railroads, Bessemer steel, electricity, the internal‑combustion engine, microprocessors—sparking decades of follow‑on innovation and new fortunes. The second reshapes the world without generating much new wealth, merely reinforcing existing structures; container shipping exemplifies this.

Jerry Neumann asks whether generative AI belongs to the first or second type.

Case Study 1: Microprocessors and the Personal Computer

In 1971 Intel invented the microprocessor, initially seeing it only as a way to simplify calculator‑chip design. Hobbyists, without any licensing or guidance, began building computers from the chip, creating a “permission‑free invention” that fueled the wave.

Early costs were high—Intel’s 8080 sold for $360 (≈$2,300 today). MITS sold the Altair kit at $75 per chip with almost no profit. The turning point came in 1975 when MOS Technologies released the 6502 for $25 (≈$150 today), enabling Steve Wozniak to afford parts and assemble the first Apple prototype in his garage.

Jovanovic & Rousseau’s data show that electricity and personal computers each took about 30 years to reach half of U.S. households, underscoring the slow diffusion of new tech.

In 1979 Apple’s ads asked consumers, “What will you do with it?” because even the company was unsure. Large incumbents like IBM, HP, and DEC ignored the market, believing no demand existed. The New York Times only began covering PCs in 1981 after IBM’s entry. In 1976 the Times mentioned personal computers only four times.

These “window‑of‑ignorance” periods gave small firms a moat. Apple raised $100 million in its 1980 IPO, becoming the only surviving major competitor before IBM entered.

Asymco data illustrate the rapid market reshuffling once IBM arrived. The ecosystem of software, memory, floppy drives, and modems reinforced each other, creating a self‑strengthening loop. Capital chased the bottleneck: when PC memory grew, software complexity rose, but external storage lagged, prompting venture capitalist Dave Marquardt to back Seagate in 1980, earning a 40‑fold return after its 1981 IPO. Within three years, $270 million poured into the hard‑disk sector, a pattern that later fed the 1990s internet bubble.

After the bubble burst, public sentiment turned against over‑expansion in tech, leading to tighter regulation and a shift from speculative to sustainable growth.

Case Study 2: Container Shipping

In 1956 truck driver‑turned‑entrepreneur Malcolm McLean repurposed surplus WWII ships to launch the first container cargo vessel, Ideal‑X. He realized that trucks, railroads, and ships all performed the same task—moving goods—so standardizing containers would eliminate repeated loading and unloading, saving time, labor, and money.

McLean faced universal resistance: dockworkers, unions, politicians, port authorities, railroads, and major shippers all opposed the disruption. He navigated these obstacles by using surplus ships to cut costs, routing short coastal trips, basing operations in Newark (avoiding Manhattan congestion), and striking a deal with the New York dockworkers’ union, which saw him as a non‑threat.

McKinsey and Levinson data show a rapid rise in container‑ship numbers in the 1960s, outpacing demand. However, standardization arrived quickly—U.S. maritime authorities began standardizing containers in 1958, just two years after Ideal‑X’s maiden voyage—allowing any player to enter.

By 1965‑66 container shipping showed clear benefits; large shipping firms flooded in. By 1968 container cargo was still under 1 % of global trade, yet container‑ship capacity had already exceeded demand, triggering price wars, profit compression, and industry consolidation into cartels. The capital‑intensive nature of ever‑larger ships and ports made the sector highly capital‑dense.

McLean sold his company to RJ Reynolds in January 1969, cashing out—perhaps the only founder to exit cleanly.

OECD data reveal that the true economic payoff of containerization only materialized about 25 years after the technology emerged. CEPII trade data confirm its long‑term contribution to global trade growth.

Who profited? McLean himself, investor Daniel Ludwig (bought at $8.5 /share in 1965, sold at $50 /share in 1969), shipbuilders (≈$10 billion spent globally 1967‑72, ≈$800 billion today), port‑construction contractors, and later shipping giants like Maersk and Evergreen. Yet “almost no one got rich by investing directly in container shipping.”

Downstream users captured the real gains: IKEA leveraged cheap global logistics to become the world’s largest furniture retailer; Walmart used containers to build a just‑in‑time supply chain, slashing inventory costs; meanwhile, traditional retailers like Sears and Woolworths declined.

Understanding the Underlying Logic

Economist Carlota Perez divides each tech wave into four phases. Microprocessors were in the explosive phase of the 1970s, when few noticed the opportunity, giving startups a moat. Container shipping sits in the mature phase of the previous wave—everyone understood it, leaving no surprise window.

Schumpeter’s “creative destruction” claim—that new entrants temporarily suspend perfect competition—fails in the mature stage, as container shipping demonstrates.

Neumann places AI at the final chapter of the ICT wave (microprocessor → PC → internet → mobile internet → AI), not the start of a new wave. Evidence: no surprise, decades of algorithmic, hardware, data, and infrastructure progress have already set the stage.

Where Is AI in the Cycle?

Neumann’s assessment: AI is the ICT wave’s closing act, lacking a surprise factor. The experimental power now resides in a handful of large model companies (OpenAI, Anthropic). Early investors like Sam Altman may reap outsized returns, but model‑company capital expenditures are massive, competition fierce, and the eventual landscape will be dominated by a few well‑funded players.

Vertical‑focus model firms (e.g., Cursor for programmers, Harvey for lawyers) may be the most valuable because they earn trust in niche markets, yet fine‑tuning costs are low, making them easy to copy and likely acquisition targets (e.g., Google’s 2010 Invite Media purchase).

Application‑layer companies (Perplexity, Writer, Abridge) are numerous but face a fundamental problem: once they scale, model owners can price‑discriminate or launch competing products, eroding profits. Most will be acquired or fail; only those with strong customer bases or elite teams might survive.

Middle‑layer platforms (Hugging Face, Glean) help enterprises manage model interfaces and protect data, but such firms did not exist in the container‑shipping era and are unlikely to grow large under current dynamics.

Upstream suppliers (NVIDIA, Scale AI, Lambda, SambaNova) face a risk similar to 1960s shipbuilders: if AI firms consolidate or cut spending, demand for specialized chips and data services could collapse, leaving suppliers with inflexible capacity commitments.

CNA shipbuilding data show a demand curve that mirrors today’s concerns about NVIDIA’s S‑curve pricing; a post‑peak slowdown would make current pricing models untenable.

Data companies (e.g., Bloomberg) that sell specialized, real‑time, hard‑to‑replicate data may thrive, but generic data is a commodity with no pricing power.

Where Is the Money?

Neumann’s core advice: don’t chase upstream AI infrastructure; instead, “fish downstream.” Knowledge‑service sectors—professional services, healthcare, education, finance, creative services—account for one‑third to half of global GDP and have seen little productivity gain from automation. AI can dramatically cut costs in these areas.

The key isn’t merely cost reduction; it’s how companies redeploy the saved money. Using savings to boost profit is a loser’s play. Winners will reinvest to lower prices, increase volume, and open new markets—mirroring IKEA’s high‑volume, low‑price model enabled by container shipping.

Companies with an existing “high‑volume, low‑price” DNA (e.g., IKEA, Walmart) are naturally positioned to benefit from AI‑driven efficiencies. New entrants may arise, akin to Costco’s early‑80s redesign of the retail model.

These downstream opportunities require little private capital; IKEA never took VC money, and Costco raised a single round before IPO. Hence, the primary arena for returns is the public equity market, not venture capital, and stock selection must be extremely disciplined.

Neumann estimates that even under optimistic scenarios AI will add 7 % to global GDP over the next decade, with roughly a third accruing to knowledge‑service firms—translating to about 2 % annual growth for those companies, a modest tailwind rather than a windfall.

The real value shift will flow from firms that ignore AI strategy to those that embrace it, just as Walmart captured market share from Sears.

The Ultimate Beneficiary: Consumers

Neumann argues that most of AI’s new value will be captured by consumers, not investors. Historical mechanization lowered clothing and food prices, but service‑sector wages rose to match productivity gains elsewhere—a phenomenon known as Baumol’s Cost Disease. AI will likely make knowledge‑intensive services cheaper while face‑to‑face services become more expensive, creating divergent opportunities.

Key Takeaway

Past five‑decade tech‑investment playbooks no longer apply. Previously, investors bet on “what the new thing is.” Today, the bet should be on “what opportunities the new thing unlocks.” Early‑stage uncertainty provides a moat for startups; once the uncertainty disappears, competition becomes perfect and moats vanish. AI is the final chapter of an existing wave, so wealth will flow to users, not builders.