Computing, Communications and Cognition: Why Moore's Law Hasn't Delivered Productivity Growth (1976–2025)

The electrification analogy is instructive. Paul David's famous 1990 paper demonstrated that electricity was invented in the 1880s, yet factories didn't achieve full productivity gains until the 1920s--a lag of 30-40 years. The reason: factories had to be completely redesigned. Early electric factories simply replaced steam engines with electric motors, keeping the same layout. Only when engineers realised they could distribute small motors throughout a factory--eliminating the central power shaft, redesigning the floor plan, introducing assembly lines--did productivity surge. We may be at a similar inflection point with AI today.

The Productivity Paradox

Despite this extraordinary technological progress, the productivity numbers tell a sobering story.

TFP Growth by Sub-Period

The pattern is striking. The 1995-2005 period--when businesses finally integrated internet technologies into their operations--produced a genuine productivity acceleration. But it was brief. After 2005, despite the smartphone revolution, cloud computing, and social media, TFP growth collapsed to just 0.5% per year. Robert Gordon, the Northwestern economist, called this "the great stagnation" and argued that the internet's productivity impact was fundamentally smaller than that of electricity, the internal combustion engine, or indoor plumbing.

Three explanations have been advanced for the paradox.

The measurement problem. GDP statistics were designed for an economy that produces physical goods. How do you measure the productivity gain from free services like Google Search, Wikipedia, or GPS navigation? How do you account for the quality improvement when a smartphone replaces a camera, a map, a calculator, a music player, and a telephone simultaneously? Erik Brynjolfsson and Andrew McAfee at MIT have argued that conventional GDP statistics miss perhaps 40% of the value created by digital technologies because that value accrues to consumers as free services rather than as market transactions.

The J-curve effect. New General Purpose Technologies initially reduce measured productivity. This seems counterintuitive, but the logic is sound. When a transformative technology arrives, firms invest heavily in it (reducing current profits), reorganise their operations (disrupting existing processes), and retrain workers (temporarily lowering output per hour). Only after this painful reorganisation period--which can last a decade or more--do the productivity gains materialise. The J-curve hypothesis, developed by Brynjolfsson, Rock, and Syverson, suggests that the 2005-2019 productivity slowdown was partly the downward stroke of a J-curve, as firms invested in AI, cloud infrastructure, and data capabilities without yet reaping the full returns.

The distribution problem. A significant share of IT investment has gone into activities that redistribute value rather than create it. High-frequency trading, targeted advertising, social media engagement optimisation, and algorithmic pricing are highly sophisticated technologies. They generate enormous returns for the firms that deploy them. But they don't necessarily increase total economic output. They shift profits from one firm to another, or from consumers to producers, without creating new goods or services.

The Balance Sheet Inversion

Perhaps the most consequential transformation of the C3 era is invisible to casual observers. It is the inversion of the corporate balance sheet.

In 1975, according to research by Ocean Tomo, 83% of the market value of S&P 500 companies was attributable to tangible assets: factories, equipment, inventory, real estate, and cash. Intangible assets--patents, brands, software, customer relationships, organisational capability--accounted for just 17%.

By 2020, those proportions had reversed. Intangible assets accounted for 90% of S&P 500 market value. Tangible assets: 10%.

This is not a subtle shift. It is a wholesale inversion of what constitutes corporate value. A company like Apple is worth over $3 trillion. Its factories (largely outsourced to Foxconn) represent a fraction of that value. Its design capability, its brand, its ecosystem of apps and services, its customer loyalty, its supply chain orchestration--these intangible assets are where the value lives.

Two landmark accounting standards attempted to close this gap, but succeeded only partially.

SFAS 86 (1985). The Financial Accounting Standards Board issued Statement 86, which for the first time required companies to capitalise the costs of developing computer software intended for sale. Before SFAS 86, all software development costs were expensed immediately--treated as current-period costs rather than long-term assets. SFAS 86 allowed companies to put software on the balance sheet once "technological feasibility" was established. This was a first tentative step toward recognising intangible assets, but it applied only to software for external sale, not internal-use software.

IAS 38 (1998). The International Accounting Standards Committee issued IAS 38, the first comprehensive international standard for intangible assets. IAS 38 defined an intangible asset as "an identifiable non-monetary asset without physical substance" and established criteria for recognition: the asset must be identifiable, the entity must control it, and future economic benefits must flow from it. IAS 38 was a conceptual breakthrough. But it contained a fatal limitation: internally generated goodwill, brands, mastheads, customer lists, and similar items could not be recognised as assets. Only acquired intangibles--those purchased through a business combination--could appear on the balance sheet.

Key Figures: The Architects

Gordon Moore (1929-2023)

Moore co-founded Intel in 1968 and in 1965 published the observation that would become "Moore's Law": the number of transistors on a chip doubles approximately every two years. This wasn't a physical law but an empirical prediction about the pace of semiconductor innovation. It held for over fifty years, driving the exponential growth in computing power that enabled every technology in this lesson. Moore's insight was that progress in computing was not linear but exponential--a pattern that humans find deeply unintuitive.

Tim Berners-Lee (b. 1955)

A British computer scientist working at CERN, Berners-Lee invented the World Wide Web in 1989. His proposal--a system of interlinked hypertext documents accessible via the internet--was initially dismissed as "vague but exciting" by his supervisor. Crucially, Berners-Lee made the web's protocols (HTTP, HTML, URLs) open and royalty-free. Had he patented the web, the digital revolution would have unfolded very differently. The web is perhaps the most valuable intangible asset ever created, and it appears on nobody's balance sheet.

Steve Jobs (1955-2011)

Jobs didn't invent the smartphone, the personal computer, or the digital music player. He integrated existing technologies into products that redefined consumer expectations. The iPhone (2007) combined a phone, an iPod, and an internet communicator. But Jobs's real innovation was the App Store (2008), which created a platform for third-party developers and generated an entirely new software economy. Jobs understood something that technology companies routinely forget: the value isn't in the hardware. It's in the ecosystem.

The Workshop Floor: Hidden Stories of Productivity

Solow's Quip and the Statistician's Nightmare

Robert Solow won the Nobel Prize in Economics in 1987--the same year he made his famous remark about computers and productivity statistics. The irony was exquisite. Solow had won the Nobel partly for developing the "Solow residual," the growth accounting method that attributes productivity growth to technological progress after accounting for capital and labour inputs.

The Solow residual was designed for an industrial economy. You measure inputs (hours of labour, tonnes of steel, kilowatt-hours of electricity) and outputs (cars produced, tons of wheat harvested). The residual--the gap between input growth and output growth--is productivity. But what happens when the output is "better search results" or "more convenient communication" or "entertainment consumed during a commute"? These outputs are real. People value them enormously. But they don't appear in GDP statistics because nobody pays for Google Search and nobody accounts for the time saved by GPS navigation.

Solow's quip was prescient. The computers were generating productivity. The statistics just couldn't see it.

The Software That Ate the Balance Sheet

In 1980, the entire US software industry had revenues of roughly $2 billion. By 2020, global enterprise software revenues exceeded $500 billion. Yet for decades, most of this value was invisible to accountants.

Before SFAS 86 in 1985, every dollar spent developing software was expensed immediately. A company could spend $100 million building a software platform that would generate revenue for twenty years, and the entire cost would appear as a current-year expense. The balance sheet showed nothing. The income statement was distorted. Investors had no way to distinguish between a company burning cash on worthless projects and one building an enormously valuable asset.

Even after SFAS 86, the rules applied only to software developed for external sale. Internal-use software--the custom systems that run supply chains, manage customer relationships, and coordinate operations--remained invisible. A company might spend $50 million building a proprietary logistics platform that gave it a decisive competitive advantage. On the balance sheet: nothing.

Today, software is the most valuable asset class in the global economy. It is also the most poorly measured.

Go Deeper: Recommended Reading

Connection to Today

The C3 era leaves us with a paradox and a measurement crisis.

The paradox: the most powerful technologies ever invented have delivered less measured productivity growth than railways, electricity, or the assembly line. The measurement crisis: the assets that drive modern productivity--software, data, algorithms, brands, organisational capability--are largely invisible to accounting standards designed for an industrial age.

These two problems are deeply connected. If your measurement system cannot see the most valuable inputs and outputs in the economy, it will systematically understate productivity growth. The IT productivity paradox may not be a paradox at all. It may simply be a measurement failure.

Consider the electrification analogy once more. Electricity was invented in the 1880s. It took 30-40 years for factories to reorganise around it. During that lag period, productivity growth was disappointing. Observers questioned whether electricity mattered. Then, in the 1920s, productivity surged as factories finally adopted the new organisational model that electricity made possible.

We may be living through a similar lag with artificial intelligence. LLMs arrived in 2022. Most organisations are still in the early adoption phase--bolting AI onto existing processes rather than redesigning work around AI capabilities. If the electrification analogy holds, the real productivity gains may still be a decade away.

But this time, there is an additional challenge. When electrification finally delivered its productivity surge, the gains showed up clearly in GDP statistics because the outputs were physical: more cars, more appliances, more houses. When AI delivers its productivity surge, the gains may be invisible--better decisions, faster diagnoses, more creative solutions, reduced friction--because they accrue as quality improvements and time savings rather than as additional physical output.

This is why measurement matters. This is why the gap between market value and book value--the intangible premium--is not merely an accounting curiosity. It is the central challenge of modern economic policy. If we cannot measure what drives growth, we cannot manage it, incentivise it, or invest in it rationally.

Opagio's Valuator exists to close this gap: to make the invisible assets visible, to quantify what accounting standards cannot see, and to give businesses and investors a true picture of where value lives. The Pioneers of 1776 created the first productivity revolution and the first framework for intellectual property. Two hundred and fifty years later, we are still trying to build a measurement system worthy of the economy those Pioneers set in motion.

This is Lesson 5 of the Productivity 250 series. Previous: The Golden Age of Productivity (1926-1975) | Next: Five Big Challenges for the Next 50 Years of Productivity Growth