Moore's Law vs Metcalfe's Law

Supply vs. Demand: The Two Engines of Tech Economics

Moore's Law and Metcalfe's Law operate on different axes of the same economic system. Moore's Law is a supply-side phenomenon: it makes computation cheaper over time, creating the conditions for new categories of products and services. Metcalfe's Law is a demand-side phenomenon: it explains why, once those products connect people, the resulting networks become exponentially more valuable and harder to displace. The PC revolution was a Moore's Law story — processing power became cheap enough for individuals. The social media revolution was a Metcalfe's Law story — Facebook's value wasn't its code but its two-billion-node social graph. The current AI infrastructure buildout combines both: Moore's Law (and its successor, Huang's Law) drives down the cost of inference, while Metcalfe dynamics determine which AI platforms achieve dominant network positions.

The Mathematics: Exponential Time vs. Quadratic Scale

Moore's Law describes exponential improvement over time — transistor counts double on a roughly fixed schedule. From 2,300 transistors on the Intel 4004 in 1971 to 336 billion on NVIDIA's Rubin in 2026, that's a 146-million-fold increase over 55 years. Metcalfe's Law describes quadratic growth over scale — value increases with the square of users at any given moment. A network of 1,000 users supports ~500,000 unique connections; at 1 million users, that number reaches ~500 billion. The practical difference matters: Moore's Law rewards patience (wait and hardware improves), while Metcalfe's Law rewards speed (capture users before competitors do). This is why hardware companies plan in multi-year roadmaps while social platforms pursue blitzscaling strategies. For a deeper look at how these exponential dynamics interact, see Red Queen Effect vs. Exponentials.

Where Each Law Breaks Down

Moore's Law hit its first major inflection point when Dennard Scaling collapsed around 2006 — transistors kept shrinking, but power density stopped falling proportionally, ending the era of free clock-speed increases. By 2026, leading-edge fabs operate at 3nm and below, where quantum tunneling effects make further miniaturization extraordinarily expensive. TSMC's Arizona fab costs over $40 billion. The industry has responded with architectural innovation: chiplets, 3D stacking, gate-all-around (GAA) nanosheet transistors, and domain-specific accelerators. Classical Moore's Law is slowing; the spirit of Moore's Law — that compute cost-performance improves relentlessly — continues through different means. See Moore's Law vs. Wright's Law for how manufacturing learning curves complement transistor scaling.

Metcalfe's Law overstates value in practice because it assumes all connections are equally valuable. Researchers Odlyzko and Tilly demonstrated that realistic network value scales closer to n·log(n) — still superlinear, but far less explosive than n². This correction explains why not every network becomes a monopoly and why niche platforms can coexist with giants. It also explains network fragmentation: your ten closest contacts matter far more than a million strangers, which is why messaging apps like Discord can thrive alongside Meta's platforms. For a comparison of network value models, see Metcalfe's Law vs. Reed's Law.

The Convergence in AI and the Agentic Economy

The most important current application of both laws is in AI infrastructure. Moore's Law (via Huang's Law) has delivered roughly 1,000× GPU performance improvement over a decade — NVIDIA's Rubin delivers 50 petaflops of AI inference per chip, up from Blackwell's ~10 petaflops just one generation prior. This supply-side deflation is crashing the cost of AI inference: per-million-token pricing has fallen from $30 in early 2023 to as low as $0.10 in 2026. Meanwhile, Metcalfe dynamics are emerging in AI ecosystems: as more agents, APIs, and tools interconnect, the value of each node in the agentic AI ecosystem grows quadratically. The platform that achieves the densest web of AI agent interoperability — the largest effective n — will benefit from Metcalfe-style lock-in, just as Facebook did with social connections.

Platform Strategy: When to Ride Which Law

For technology strategists, the distinction between these laws maps directly onto build-vs.-network decisions. If your competitive advantage comes from raw computational capability — training the largest AI model, running the fastest simulation, processing the most data — you are riding Moore's Law (and its successors). Your moat is technical and temporary; competitors can catch up as hardware improves for everyone. If your competitive advantage comes from user connections and ecosystem density — the most creators, the deepest social graph, the richest plugin marketplace — you are riding Metcalfe's Law. Your moat is structural and compounding. Companies like Roblox demonstrate both: Moore's Law enables increasingly rich 3D experiences on cheap devices, while Metcalfe's Law makes the platform stickier as millions of creators and players form an interconnected social layer. The flywheel economics of such platforms combine both laws into a single reinforcing loop.

The Deflationary Intersection

Both laws contribute to the broader pattern of deflationary technology. Moore's Law drives cost deflation directly: the same computation costs half as much every ~two years. Metcalfe's Law drives value inflation: each user added to a network makes the network disproportionately more valuable. When combined, the result is that technology platforms deliver exponentially more value at exponentially lower cost — the fundamental dynamic that has driven three decades of internet economics. However, this dynamic also creates winner-take-all outcomes and raises questions about platform economics and market concentration. As AI infrastructure costs continue to fall while AI network effects intensify, these twin dynamics will shape which companies dominate the next era of computing.