High Entropy Markets, a Closer Look
The insurance market is about 670 years old, dating from marine insurance contracts in 14th century Genoa and Florence. The electric utility market is 144 years old, from Edison’s Pearl Street Station. The payment network market—Visa, Mastercard—is roughly 68 years old. The generic pharmaceutical market, in its modern form, is about 42 years old.
The quantum computing market is 45 years old, if you start from Feynman’s 1981 talk at MIT. The AI agent market is somewhere between 40 years old (if you count distributed AI research) and 3 years old (if you count generative AI agents). The metaverse is either 34 years old (Neal Stephenson’s Snow Crash) or 5 years old (Facebook’s rebrand).
Which raises an interesting question: why are some 40-year-old markets stable and predictable while others remain chaotic?
The answer, I think, lies in what I’ve been calling High Entropy Markets, markets where the fundamental confusion isn’t about execution risk but about the nature of the opportunity itself.
A Working Definition
A High Entropy Market is characterized by fundamental confusion about what is being built, sold, or bought—where participants lack consensus on terminology, value location, and applicable mental models, and where underlying technology shifts faster than stable business categories can form.
The key distinction: High Entropy is not the same as High Risk.
A high-risk market might have clear definitions and known value chains but uncertain outcomes. Will this drug pass FDA trials? Will customers adopt this product? The questions are well-formed even if the answers are unknown.
A high entropy market involves confusion about the questions themselves. What even counts as an “agent”? Who captures value in a “decentralized” system?
The Three Dimensions
High entropy manifests across three dimensions:
| Dimension | Key Question | What to Look For |
|---|---|---|
| Semantic Uncertainty | What are we talking about? | Same terms used with different meanings; different terms for identical concepts; no stable taxonomy |
| Economic Uncertainty | Who captures value, and how? | Unclear where money accrues in the value chain; no established business model patterns |
| Technical Uncertainty | What is actually possible? | Capabilities shift faster than documentation; business models obsolete before maturity |
Five High Entropy Markets
1. AI Agents ($7.84B) — Very High Entropy
Semantic: The term “agent” has been “diluted beyond utility,” as one academic paper put it. OpenAI defines agents both as “automated systems that can independently accomplish tasks” and as “LLMs equipped with instructions and tools”—two incompatible frameworks from the same company. Microsoft distinguishes agents from assistants. Anthropic acknowledges the term can mean several things.
Economic: Where does value accrue? The model providers? The orchestration layer? Agent creators? The shift from SaaS (pay for tools, measure time saved) to “service-as-software” (pay for outcomes, measure whether humans were needed) destabilizes every pricing model.
Technical: In 2024, agents were “small and specialized.” By 2025, they could “plan, call tools, and complete complex tasks.” Now we’re hearing about “agent control planes”—fundamentally different architectures than existed 18 months ago.
2. Web3/Blockchain ($7.62B) — Very High Entropy
Semantic: “Web3 does not have a single specific widely accepted definition.” Some use it narrowly to mean blockchain applications. Others include everything from AI to VR. The term originally meant something entirely different—Tim Berners-Lee’s Semantic Web.
Economic: Will Web3 bypass today’s platforms through decentralization, or will incumbents control it? Will smart contracts enable creators to capture value, or will complexity favor intermediaries?
Technical: The centralized-versus-decentralized architecture question remains unresolved after 17 years. Can blockchains scale? Does “trustlessness” work as designed?
3. Quantum Computing ($3.52B) — Very High Entropy
Semantic: Multiple competing paradigms (superconducting qubits, photonic, trapped ion, quantum annealing) fragment the field. Terms like “logical qubits” and “quantum advantage” shift meaning as technology evolves.
Economic: The commercialization path remains unclear. Current applications are narrow—specific physics, chemistry, or optimization tasks. Where value concentrates in hybrid quantum-classical systems is unknown.
Technical: Capability timelines keep compressing. “2026 could be the fastest-moving year yet for quantum hardware,” which means any strategic bet may be obsolete by summer.
4. Metaverse ($227.9B) — Very High Entropy
Semantic: Is the metaverse VR environments via headsets? Decentralized digital economies with blockchain-based ownership? Or simply the threshold when people value digital assets as highly as physical ones? The definitions point to incompatible futures.
Economic: Business models remain contested. VR requires expensive hardware, limiting addressable market to perhaps 1 billion versus 5 billion internet users.
Technical: The architectural fork of centralized platforms versus interoperable protocols remains unresolved. Whether digital assets can transfer across platforms is still an open question.
5. Synthetic Biology ($20.5B) — High Entropy
Semantic: Unlike the others, synthetic biology has relative semantic clarity. Scientists agree on what CRISPR does. Boundary questions exist but confusion is modest compared to agents or metaverse.
Economic: The supply chain is “uncompetitive and not fit for purpose.” Companies face a Catch-22: they need economies of scale to compete on price, but can’t reach scale with expensive products. Many “bank on charging green premiums of 200%” without validating demand.
Technical: Existing life cycle assessment frameworks don’t work for synthetic biology. The field lacks validated tools to measure the environmental benefits used to justify investment.
Defining Low Entropy Markets
A Low Entropy Market is the opposite: stable definitions, established value chains, predictable technology evolution. Everyone knows what the product is, who pays for it, and how the underlying technology works.
Low entropy doesn’t mean boring or unprofitable. It means the strategic questions are well-formed.
Five Low Entropy Markets
1. Consumer Packaged Goods
Semantic: A box of cereal is a box of cereal. Categories are stable (food, beverage, household, personal care). Brand positioning varies but product definitions don’t.
Economic: Value chains are mapped to the penny. Manufacturers, distributors, retailers each have established margins. Private label economics are understood.
Technical: Manufacturing and distribution technology evolves incrementally. No paradigm shifts in how soap is made or shipped.
2. Utilities (Electric, Gas, Water)
Semantic: Electricity is electricity. A kilowatt-hour has a universal definition. Service categories (residential, commercial, industrial) are standardized.
Economic: Regulated monopolies with rate-of-return economics. Capital investments, operating costs, and allowed returns are transparent.
Technical: Generation, transmission, and distribution technologies are mature. Even renewable integration follows known patterns.
3. Payment Networks
Semantic: A payment is a payment. The four-party model (cardholder, merchant, issuer, acquirer) is universal. Interchange categories are defined.
Economic: Revenue streams are established: interchange fees, assessment fees, processing fees. Margins at each layer are known.
Technical: The card rails are stable infrastructure. Even new payment methods (contactless, digital wallets) plug into existing architecture.
4. Insurance
Semantic: Policy types are standardized. “Term life” means term life. Actuarial categories are consistent across carriers. The basic concepts—premium, indemnity, risk pool—have been stable since Venetian merchants formalized them in the 1300s.
Economic: Underwriting profit plus investment income. Combined ratios, loss ratios, and expense ratios are comparable metrics.
Technical: Risk modeling is mature. Even catastrophe modeling follows established methodologies.
5. Generic Pharmaceuticals
Semantic: A generic is bioequivalent to its reference product. FDA standards define what “generic” means.
Economic: Price competition drives margins. Manufacturing scale determines profitability. ANDA filing costs and timelines are predictable.
Technical: API synthesis and formulation science are mature. Regulatory pathways are defined.
The Age Puzzle
Here’s what I find interesting: the correlation between age and entropy isn’t what you’d expect.
Low Entropy Markets are old, but not uniformly ancient:
- Insurance: ~670 years (14th century Venice and Genoa)
- CPG: ~190 years (P&G founded 1837)
- Utilities: ~144 years
- Payment Networks: ~68 years
- Generic Pharmaceuticals: ~42 years
High Entropy Markets are older than they appear:
- Quantum Computing: ~45 years (Feynman’s 1981 MIT talk)
- AI Agents: ~40 years (distributed AI research) or ~3 years (GenAI agents)
- Metaverse: ~34 years (Snow Crash) or ~5 years (current hype)
- Synthetic Biology: ~25 years
- Web3: ~17 years (Bitcoin whitepaper)
Payment networks and generic pharma achieved low entropy in decades. Quantum computing remains high entropy after a similar timeframe.
The difference isn’t age. It’s whether definitions stabilized.
Insurance crystallized early because the basic problem, merchants needed to pool risk on cargo ships crossing the Mediterranean, was concrete and urgent. Payment networks crystallized around clear standards (the four-party model, interchange categories) that everyone adopted. Generic pharma crystallized around regulatory clarity (the Hatch-Waxman Act, bioequivalence standards). In each case, semantic clarity came first. The technology and economics followed.
Quantum computing and blockchain lack this semantic foundation. When the basic terms remain contested, economic and technical uncertainty persist regardless of how much time passes.
The Velocity Question
This suggests something about the current moment. We might be seeing markets reach low entropy faster than before—or at least, we have examples of markets that achieved stability relatively quickly.
The question for high entropy markets isn’t “how old are they?” but “what would it take for definitions to stabilize?”
For AI agents, that might mean industry consensus on what the term actually means—which company or consortium will play the role the card networks played in payments? For quantum computing, it might mean one technical paradigm winning decisively. For Web3, it might mean the centralization-versus-decentralization question getting resolved by market forces rather than philosophy.
Until then, operating in these markets requires a different playbook than standard strategic planning. The frameworks I’ve used for exploring unknown markets and decomposing unfamiliar problems apply here. Recognize the entropy, account for it, and watch for the moments when it starts to decrease.
Those moments, when semantic uncertainty begins to resolve, are where strategic advantage gets created.