← All posts / Industry

0.4% Against a Zero Forecast: The ONS Just Put AI in Britain's GDP Numbers

Britain's economy grew 0.4% in July against expectations of stagnation, and for the first time the Office for National Statistics explicitly named AI and cloud-computing firms as a measurable driver of the print.

0.4% Against a Zero Forecast: The ONS Just Put AI in Britain's GDP Numbers

On 11 September 2026, the United Kingdom’s Office for National Statistics published its monthly GDP estimate for July — and delivered the kind of surprise that statisticians rarely get to write. The economy grew by 0.4% in a month when analysts, polled in advance, had predicted no growth at all. May had been flat; June had delivered 0.3%. July’s print was the strongest monthly performance since early 2025, and it lifted annual GDP growth to 1.6% in the twelve months to July — the fastest pace in eighteen months.

But the headline number is not the real story. The real story is a single paragraph buried in the services section of the bulletin, in which the UK’s official statistics agency — an institution famous for its caution — explicitly attributed part of the growth to artificial intelligence.

The number that beat the forecast

Monthly real GDP is estimated to have grown by 0.4% in July 2026, after 0.3% in June and stagnation in May. All three main sectors contributed positively: services rose 0.4%, production 0.2%, and construction 0.1% — a rare month in which no sector dragged. On a three-month rolling basis, GDP grew 0.4% in the three months to July compared with the previous three months, the eighth consecutive three-month-on-three-month expansion.

The largest contributions came from where you would least expect them in a British economy that has spent two years worrying about stagnation. Administrative and support service activities rose 3.7% on the month, driven by rental and leasing (up 7.9%) and services to buildings (up 4.0%). But the second-largest contributor was the one that caught everyone’s attention: information and communication grew 2.4% in July, and within it, computer programming, consultancy and related activities expanded 3.5% in a single month, contributing 0.14 percentage points to services output and 0.12 percentage points to headline GDP.

To put that in perspective: one narrow subsector — essentially software houses and IT consultancies — delivered nearly a third of July’s entire GDP growth on its own.

The ONS breaks its silence on AI

Here is the sentence that turned a routine statistical release into a global news story. The ONS wrote:

“There is evidence that across computer programming, consultancy and related activities, and information services activities (up 1.1% in July 2026), many of the businesses reporting the largest turnover in July 2026 are involved in activities related to artificial intelligence and cloud computing. However, because of the nature of our data collection, it is difficult for us to quantify the exact impact of these types of activities on turnover.”

Read it carefully, because every clause matters. The agency is not saying “AI added X% to GDP” — it explicitly says it cannot quantify the exact impact. What it is saying is something arguably more significant: when ONS analysts looked at which firms inside the booming IT subsector were reporting the biggest turnover, they found AI and cloud-computing businesses overrepresented at the top of the distribution. The pattern, per ONS director of economic statistics Liz McKeown, was visible not just in July but in May and June as well.

This is a quiet methodological milestone. For years, economists have argued about whether AI’s productivity effects would ever show up in the national accounts — a modern variant of Solow’s famous observation that you can see the computer age everywhere except in the productivity statistics. The ONS has now gone on record, in an official statistical bulletin, linking a visible growth beat to firms whose business is AI. The effect is being detected through the survey data itself: the biggest turnover reporters in software services just happen to be AI companies.

Three months of compounding software strength

The July snapshot is not an isolated spike. In the three months to July, computer programming, consultancy and related activities grew 4.4%, pulling the whole information and communication subsector up 2.5%. Professional, scientific and technical activities — another AI-adjacent category — rose 2.1%, driven by scientific research and development, which surged 7.0% over the same period.

The shape of the growth tells a consistent story: the UK’s expansion is increasingly being carried by knowledge-intensive services, while physical sectors struggle. Production fell 0.5% and construction fell 0.5% over the three-month window. Education declined, with the ONS noting school closures during June’s heatwave. Wholesale trade fell 3.0% over three months and was the single largest negative contributor to July’s GDP.

In other words, the British economy in mid-2026 looks like a barbell: a heavyweight technology-services engine pulling forward, while traditional industry and consumer-facing sectors sag under the weight of elevated energy prices and borrowing costs.

Resilience with an asterisk

Chancellor John Healey called the figures evidence of “a welcome resilience, despite serious global uncertainty,” noting that growth — “although still fragile” — was the fastest in the G7 in the first half of the year. He presents his first Budget in October against borrowing costs sitting near three-decade highs, and the opposition was quick to point out that construction and production are shrinking and unemployment has risen.

Economists, meanwhile, were nearly unanimous in treating July as good news with a short shelf life. Paul Dales of Capital Economics said the data showed first-half resilience continuing, but warned that higher energy and borrowing costs “would soon start to hit growth.” KPMG’s chief economist Yael Selfin noted that the headline figure “masks a weaker picture for households”: consumer-facing services contracted in July, with retail and hospitality falling back. Richard Carter of Quilter Cheviot suggested activity “is likely to stall ahead of the Budget.” The Bank of England’s rate decision looms the following week, with some analysts pencilling in a hike before year-end as Middle East conflict pushes oil, energy, and fuel prices upward.

Why this matters beyond Britain

The UK print arrives amid an accumulating body of evidence that AI investment has started showing up in macroeconomic data. Singapore raised its 2026 growth forecast to 5.5% citing an AI investment boom. Taiwan posted its fastest growth in nearly four decades on AI hardware demand. Australia’s capex surge prompted warnings of capacity constraints. The UK case is different in one important respect: it is not a data-center construction story or a chip-export story. It is a services-utilization story — firms buying, building, and deploying AI, and generating measurable turnover in the process.

That distinction matters for every advanced economy wondering where its AI dividend will come from. Britain does not manufacture the GPUs; it cannot rival the hyperscalers’ capex. What it apparently can do is monetize the deployment layer — consultancies, software houses, and AI-native service firms billing for implementation at a pace strong enough to register in monthly GDP.

There is also a sobering counterpoint running through the BBC’s reporting. Rob Arnold, co-founder of the nine-person AI firm Ascendea, claims his company develops apps “100 times quicker at a 50th of the cost” thanks to AI — but says the UK government needs to invest more in the sector, warning that he knows several small British AI firms that have moved to the US or are considering it because better opportunities lie across the Atlantic. His other warning deserves repeating: AI “can be dangerous if not understood properly. It’s like playing with a weapon.”

The measurement problem comes into focus

The ONS’s honesty about its own limits — “it is difficult for us to quantify the exact impact” — highlights the measurement challenge now facing every statistical agency. AI’s economic footprint is diffuse: it shows up as faster software development, cheaper consultancy delivery, and new product categories that sit awkwardly inside industrial classifications designed decades ago. When the biggest turnover reporters in “computer programming” are AI firms, the old categories still work. When AI becomes invisible infrastructure inside law firms, hospitals, and logistics companies, they will not.

Britain’s next GDP releases — including Blue Book 2026 revisions due in October — will be watched closely for exactly this reason. If the AI-linked strength in software services persists through August and September, the UK will have assembled the first sustained, officially recognized evidence trail of AI-driven growth inside a major G7 economy’s national accounts. If it fades, July 2026 will be remembered as the month the AI trade met the business cycle — and the business cycle won.

Either way, the era of arguing that AI exists only in the stock market and not in the statistics is quietly drawing to a close. The ONS just wrote its obituary in the most bureaucratic language possible — and that is precisely what makes it credible.