When ONS Redefines Digital Infrastructure, the Capital Structure Was Already There

On 24 August 2026 the Office for National Statistics published its updated estimate of UK digital-infrastructure investment using an expanded definition. The headline finding: £11.2 billion in market-sector digital-infrastructure investment for 2025. The old headline measure? Just £3.6 billion. The gap between them — a factor of 3.1 — is not new investment that appeared overnight. It is investment that was already happening, already priced by entrepreneurs and engineers, already generating returns in the real economy. The ONS simply had not been counting it.

This is not a story about government stimulus or industrial policy. Government capital grants funded just 5.7% — or £0.7 billion — of all market-sector digital-infrastructure investment in 2024, up from £0.2 billion in 2023. The remaining 94.3% was deployed by private actors responding to profit signals, consumer demand, and their own estimates of future value. The capital structure had already formed. The ONS was late to the reckoning.

Eugène von Böhm-Bawerk identified this dynamic in 1884. Producers, he explained, must anticipate consumer demand across time. They borrow from present savers to finance production processes that will not yield consumption goods until some future date. The rate of this borrowing — the interest rate, or more broadly the time preference of the community — determines how much capital is accumulated, how long production chains can be, and therefore how productive the economy becomes. The crucial word is anticipate. The capital structure is built before the statistician counts it. By the time an official framework is updated to include software licences, hyperscale data campuses, and cryptographic infrastructure, those assets have already been deployed, depreciated, upgraded, and in some cases superseded. Measurement lags the market because measurement is retrospective by design, while entrepreneurial activity is prospective by necessity.

The ONS article itself confirms this. It notes that adopting an expanded methodology "would continue to bring our estimates of UK digital infrastructure investment closer to external estimates than our current headline measure." The gap between the ONS's £3.6 billion figure and its £11.2 billion estimate represents investment that market participants had already priced into software licences, database infrastructure, data-centre construction, and fibre-optic rollout. The administrative measure lagged the market because administrative measures always lag markets. They are snapshots taken from the outside of processes that run continuously from the inside.

Between 2020 and 2025, 77.6% of total market-sector digital-infrastructure investment concentrated in just two asset categories: software and databases, and other buildings and structures and transfer costs. This is the anatomical picture of a modern capital structure — not bridges and motorways but intellectual-property products and climate-controlled server halls. The ONS methodology was developed for traditional infrastructure such as water and electricity. It was not designed to capture a sector where the dominant assets are intangible, rapidly depreciating, and cross-classified across multiple existing data categories.

Data centres illustrate the measurement lag with particular clarity. The Department for Energy, Security and Net Zero estimates that operational data centres in Great Britain increased from 221 to 239 between 2020 and 2024. Oxford Economics' historical series stretches further back: from 52 in 2000 to 266 in 2024. These facilities — now formally designated as critical national infrastructure (CNI) by the UK Government — were being planned, financed, and built throughout this entire period. The national accounts simply lacked the statistical framework to include them.

The European System of Accounts (ESA 2010) — the EU's accounting framework implemented in 2014 and aligned with the global System of National Accounts 2008 — does not explicitly define infrastructure. The ONS notes that infrastructure is therefore not covered in SNA 2025 or ESA 2010; the functional definition of economic infrastructure used here is the ONS's own published-infrastructure approach, developed mainly for traditional infrastructure such as water and electricity, not an ESA classification of digital assets. The ONS is clear that its proposed changes "will not lead to the revision of the main national accounts statistics" in the current cycle. This distinction matters: the existing GDP, gross fixed capital formation, and national balance-sheet estimates will not be revised. What changes is the supplementary analytical framework — the lens through which policymakers and researchers attempt to understand the economy that already exists.

This is the same epistemic problem that Böhm-Bawerk diagnosed at the level of the individual firm, scaled to the level of statistical administration. The capital goods are there whether or not the accounts capture them. The production processes are running. The future-oriented activity has already occurred. What the ONS has done — usefully, but retrospectively — is close the gap between what was being priced and what was being counted.

There is a measurement system that runs continuously rather than on a publication calendar. On Ethereum mainnet, every transfer, every collateralised position, every oracle attestation settles against a single reference unit in hours, not years. The price of the future that participants are willing to pay for in the present is recorded at the moment of exchange, by code, in public state. No statistical reclassification is required; no supplementary framework, no methodology committee, no seven-year lag between the build and the count.

This is not an investment recommendation. It is an observation about the form of economic measurement. The gap between an on-chain settlement record and a retrospective statistical article is, at its core, a gap between a market that continuously discovers prices and an administrative system that retrospectively classifies them. The ONS does not lose the argument — it never had the speed. Its methodology is the visible price a centrally administered statistical system pays for the authority to publish a single number.

The ONS deserves credit for attempting to close its own gap. An expanded £11.2 billion estimate is more analytically useful than a narrower £3.6 billion figure. But the episode reveals something deeper: the capital structure, in the Austrian sense, was never absent. It was simply uncounted. Entrepreneurs, engineers, and investors had already made their time-preference calculations. They had already built the data centres, licensed the software, and deployed the databases. What the official statistics provided — belatedly, and with caveats about their limitations — was confirmation of what the market had already decided.

Böhm-Bawerk would not have been surprised. The capital structure is not a statistical artefact. It is a real phenomenon that exists independently of how we choose to measure it. The 3.1 factor between the ONS's old headline and its new estimate is not evidence of a policy success. It is evidence of how far behind the statistics always lag the economy that is actually being built.

Sources: Office for National Statistics, Redefining investment in digital infrastructure in the UK: 2026 (24 August 2026), ONS article; ONS, Data centres and the UK National Accounts (methodology); ONS release summary (24 August 2026). ESA 2010 framework: Eurostat ESA 2010 overview.