NextFin News - Artificial intelligence is pushing up prices - memory chips, electricity, data-centre construction - but it is not what keeps Bank of England policymakers awake at night. With UK inflation at 3.1% in July, above the 2% target and rising for the first time in three months, the central bank's dominant worry is older and stickier: an energy shock from the Middle East feeding into wages and services prices. The AI bill is real, but it is the second-round spiral, not the chip shortage, that would force the Monetary Policy Committee's hand.
The distinction matters because it determines where the next rate move comes from. If AI-driven costs were the marginal driver, the appropriate response would be patience - supply constraints ease. If energy pass-through into wages is the driver, the response is tighter policy for longer. The data, and the committee's own language, point to the second channel.
The Numbers: Inflation Is Back Up, and It Is Not Chips Driving It
Britain's consumer price index including owner-occupiers' housing costs rose 3.1% in the 12 months to July, up from 2.8% in June, the Office for National Statistics said. The headline CPI measure rose 2.9%. Core CPIH - excluding energy, food, alcohol and tobacco - edged up to 2.9% from 2.8%.
The composition of the increase is the story. Goods inflation accelerated from 1.7% to 2.2%, while services inflation held at 3.6% on the CPIH measure and eased only marginally to 3.4% on the CPI measure. That is the signature of an imported cost shock - energy and intermediate inputs - rather than a broad-based domestic demand boom.
That services print is the number the committee watches. At 3.4%-3.6%, services inflation remains far above the 2% target and is the closest official proxy for domestically generated, wage-driven price pressure. Goods inflation can be imported and transitory; services inflation is made in Britain, and it is the channel through which an energy shock becomes entrenched.
The Bank of England has not moved rates since cutting Bank Rate to 3.75% in December 2025. At its meeting ending 29 July, the MPC voted 6-3 to stand pat, with three members pushing for a quarter-point rise to 4% - the largest hawkish minority in recent months. The split captures the dilemma: inflation has fallen sharply from its 2022 peak, yet it is heading back up, and the committee's own forecast says it will rise further through the rest of 2026 as higher global energy prices work through utility bills and supply chains.
The AI Inflation Channel Is Real, and It Is Quantifiable
Dismiss AI as an inflation driver at your peril. The mechanism is concrete: a surge of investment into data centres and AI hardware is running into scarce supply, pushing up prices for memory chips, processors and the electricity needed to run them. Global investment in data centres is approaching $1 trillion a year, nearly 1% of the world economy, according to the International Data Center Authority, and AI-related buildout spending is on track to top $700 billion in 2026.
The price evidence is stark. JPMorgan economists estimate that the cost of some computer memory chips will have risen as much as 400% between 2024 and the end of 2026. In the United States, electricity prices rose 5.9% in May from a year earlier, well above the 4.2% overall inflation rate, after having fallen back to around 2% in early 2025. Goldman Sachs economists forecast in February that power prices would rise 6% in 2026 and 2027, then a still-above-average 3% in 2028.
Britain is directly exposed. Data centres now consume about 5.9% of the UK's electricity, up from a government estimate of 2.5% in early 2025, and the government expects that share to quadruple by 2030. Ofgem says roughly 140 proposed data-centre projects - almost all driven by AI demand - would require more power than Great Britain's current peak electricity consumption. Oxford Economics forecasts UK data-centre power demand could grow fivefold over five years, with developers waiting as long as a decade for grid connections.
"We do know what effect AI is having on inflation now, and it is inflationary, not deflationary," wrote Dario Perkins, an economist at TSLombard.
Central bankers on both sides of the Atlantic have noticed. John Williams, president of the Federal Reserve Bank of New York and vice-chair of the Fed's rate-setting committee, said in July: "If this creates a sustained impulse to demand relative to supply in inflation, I do think that's the kind of situation where you don't look through this." The Bank of England has reportedly opened a probe into lending to data centres amid fears that an AI valuation collapse could transmit into the financial system.
Why the BOE Weighs It Differently: Scale, Persistence and the Second-Round Test
Here is why AI is not the BOE's biggest worry. A central bank does not react to a price level; it reacts to whether that price change will persist and propagate. The AI impulse, for all its visibility, fails two of the three tests that would make it a monetary-policy problem: it is small relative to the energy shock, and much of it is a one-off supply adjustment rather than a self-reinforcing loop.
Start with scale. Even if memory-chip costs quadruple, the weight of semiconductors and computing equipment in the UK consumer basket is tiny - a fraction of a percentage point. The pass-through to headline inflation is measurable but modest. By contrast, energy touches everything: the BOE's April Monetary Policy Report warned that CPI inflation had jumped to 3.3% in March on the back of the Middle East conflict, and that indirect effects would be largest for food prices as companies pass higher energy costs through supply chains.
Then there is the time profile. Chip prices are cyclical by nature - the semiconductor industry has run through boom-bust cycles for four decades, and today's shortage invites tomorrow's overcapacity. The electricity leg is more persistent, but it is still a cost-push shock whose inflationary effect fades once capacity is built. The BOE's own framework treats such first-round energy and supply effects as something to "look through" unless they seed something worse.
That "something worse" is the second-round effect - the wage-price spiral the committee has been hunting since 2022. This is the mechanism that turns a relative-price shock into persistent inflation: workers demand higher pay to keep up with the cost of living, firms raise prices to protect margins, and expectations ratchet upward. The BOE's July report put it plainly: "The risk of material second-round effects in price and wage-setting, against which policy needs to lean, is greater the longer higher energy prices persist." It added that there is "little evidence so far" of such effects, but the risk is tilted to the upside.
The wage data explain the caution. Regular pay growth ran at 3.4% in the three months to April, with public-sector settlements at 5.1% and the private sector at 2.9%. That is down from the 7%-plus peaks of 2023, but with services inflation at 3.4%-3.6%, real pay is barely advancing and workers still have reason to press for catch-up. The February Monetary Policy Report was explicit about what the committee needs to see: "Services price inflation and wage growth still need to fall further for the MPC to be confident that inflation will return to the target and stay there."
Expectations are the tripwire. The BOE's July report noted that "high inflation over the past five years may have made inflation expectations more sensitive to inflation than in the past," and that some measures of household inflation expectations have already risen sharply. If expectations de-anchor, the AI shock stops being a footnote and becomes kindling.
The Transmission Mechanism: Two Channels, One Decisive Difference
The cleanest way to see why the BOE ranks these risks as it does is to trace the two transmission channels side by side.
The AI channel runs through imported goods and capacity-constrained utilities. A memory-chip shortage raises the price of phones, laptops and servers. Data-centre power demand raises wholesale electricity prices. Both are real cost pressures, but both operate through relative prices in specific markets. Their inflationary effect is front-loaded: the price jumps when the shortage bites, then stabilises. That is precisely the pattern a central bank can look through, because it does not require a change in the policy stance to reverse - it reverses when supply catches up.
The energy-services channel runs through the wage-setting process. Higher energy and food prices are highly visible to households - the BOE singles them out for exactly that reason. Workers see their real pay erode and press for catch-up. Firms with compressed margins, which the BOE's agency intelligence has flagged, pass those labour costs into the prices of services that cannot be imported: haircuts, restaurant meals, repairs, hospitality. Because services make up roughly two-thirds of the UK consumption basket and are labour-intensive, a 1% acceleration in services inflation is far harder to reverse than a chip-price spike. It requires slack in the labour market, weaker demand, or a period of below-target inflation to break - all of which carry recession risk.
This is why the committee's language has narrowed so consistently to services inflation and wage growth. They are not the trigger of the current inflation uptick - energy is. They are the transmission mechanism, and transmission mechanisms are what monetary policy can actually affect.
The Cyclical Call: AI Is a Supply Shock; Services-Wage Persistence Is the Regime Question
This is the judgment that separates a headline from an analysis. The AI-driven cost impulse is, on current evidence, a cyclical supply shock: large in nominal investment terms, but mean-reverting as chip fabrication capacity catches up with demand and as data-centre buildout slows from its current fever pitch. The 400% chip-cost surge will not compound forever - it will unwind, as semiconductor cycles have done through every generation of the industry, from the DRAM shortages of the 1980s to the automotive-chip crunch of 2021.
The services-wage nexus is the structural question. Whether UK inflation settles sustainably at 2% depends less on the price of GPUs than on whether the labour market can absorb higher energy costs without a new round of pay settlements embedding 3%-plus services inflation. That is why the MPC's debate has narrowed to services inflation and wage growth: they are the transmission mechanism, not the trigger.
There is a second-order implication the market is underweighting. If the BOE concludes the AI impulse is transitory, it can afford to look through it - but only if the energy shock does not produce second-round effects. That makes every upcoming wage and services print a referendum on policy. A string of 0.3%-plus monthly services inflation readings would shift the hawks from a three-member minority toward the controlling bloc, because it would signal that the energy shock is propagating exactly as the committee fears. Conversely, if services inflation drifts back toward 3% while goods inflation moderates, the committee gains the room it wants to resume cutting rates - and the AI debate becomes academic.
The international context sharpens the point. The International Monetary Fund has warned that the UK is on course for the highest inflation rate among G7 nations in 2026, a reflection less of any single commodity price than of the persistence of domestic price-setting. That is the reputation cost the BOE is trying to avoid, and it is not a cost that chip prices drive.
The Counter-Thesis: What If AI Is the Structural Shock After All?
The strongest case against this reading is that AI is not a normal supply shock because demand for compute shows no sign of saturating. If data-centre power demand really quadruples by 2030 and grid constraints keep supply tight for a decade, the electricity-cost channel is not cyclical at all - it is a persistent, structural lift to the price level. In that scenario, the BOE would be making the classic error of "looking through" a shock that never reverses, the way many central banks looked through energy-price strength in 2021 and paid for it with a decade of above-target inflation.
That argument has force, and it is why policymakers are monitoring the data rather than declaring victory. But it still does not make AI the biggest worry today, for one reason: monetary policy cannot build power plants or fabs. Even if the BOE wanted to lean directly against AI-driven electricity costs, interest rates are a blunt instrument against a supply constraint. What rates can do is prevent second-round wage effects - which is precisely why services inflation and pay growth remain the operative variables. The counter-thesis only becomes the base case if wage settlements start consistently exceeding 4% while services inflation stalls above 3.5% - and even then, the policy response would be aimed at the wage spiral, not at AI.
What to Watch
The next decisive data points are the ONS services inflation and average-earnings releases, and the MPC's reaction function at its next meeting. Watch three signals:
- Services CPI: a sustained move above 3.8% would indicate the energy shock is embedding in domestic prices.
- Regular pay growth: a re-acceleration above 4% would revive the second-round risk the committee says it has not yet seen.
- The MPC vote split: a shift from a 6-3 hold to a narrower majority, or a rate rise to 4%, would mark the point at which persistence fears override transitory-supply comfort.
The falsifying signal for the view that AI is not the BOE's main worry is specific: if memory-chip and electricity costs keep rising while services inflation and wage growth continue to fall toward target, then AI is indeed the dominant marginal inflation driver and this analysis is wrong. The opposite pattern - sticky services and pay alongside moderating goods - confirms that the old wage-price question still owns the meeting room.
Bottom Line
Three scenarios frame the path from here, each tied to a trigger rather than a prediction.
The base case is that services inflation and wage growth grind slowly lower while goods inflation absorbs the energy and AI cost shocks. In that world, the BOE holds at 3.75% through the rest of 2026, looks through the AI impulse, and keeps its cutting option open for early 2027. The upside case for inflation - the one the hawks are pricing - is that monthly services readings stay at 0.3%-plus and regular pay growth re-accelerates toward 4%. That is the path that turns the 6-3 hold into a rate rise to 4% and keeps policy restrictive well into next year. The downside case is that the energy shock proves shorter-lived than feared, demand weakens faster than expected, and inflation falls back toward 2.5% before year-end - in which case the AI debate is overtaken by growth concerns.
AI is inflationary, and nobody at Threadneedle Street is ignoring it. But a central bank fights persistence, not price levels, and the persistence in the UK economy today comes from energy pass-through and the wage-setting behaviour it could trigger - not from the price of semiconductors. The chip bill will eventually be paid and the cycle will turn. What the BOE cannot look through is a labour market that keeps pricing 3%-plus services inflation into the cost of living. That is the inflation worry that actually moves rates.
Explore more exclusive insights at nextfin.ai.
