Inflation turned up in July. The story underneath is cooling.

UK consumer price inflation rose to 2.9 per cent in the twelve months to July 2026, up from 2.6 per cent in June — the first increase in the annual rate since March. On the surface, that is an uncomfortable headline: inflation has been above the 2 per cent target continuously since 2024 — June marked the twenty-first consecutive month — and it is moving the wrong way again. But the July print is unusually easy to decompose, and the decomposition changes what the number means.

The rise was almost entirely mechanical and external. Ofgem's 13 per cent energy price cap increase took effect on 1 July, pushing the annual gas rate to 14.7 per cent — the largest since October 2022 — and electricity to 3.6 per cent. Those figures land against a July 2025 comparison month in which gas prices fell 7.2 per cent and electricity fell 3.8 per cent, so the arithmetic swing in the annual energy rate is close to 22 percentage points on gas alone.

Beneath that, the domestic picture continued to cool. Core CPI was unchanged at 2.6 per cent. Services inflation fell from 3.6 to 3.4 per cent. Food and non-alcoholic beverages slowed to 1.3 per cent, the smallest contribution to the headline rate since October 2021. That combination — an imported, administered price shock pushing the headline up while domestically generated inflation eases — is the whole story of this release, and it is what makes the policy question genuinely awkward.

The goods–services split is the analytically important line in this release. Goods inflation jumped half a percentage point, from 1.7 to 2.2 per cent, while services inflation fell 0.2 points, from 3.6 to 3.4. Goods are where imported energy and commodity costs land first; services are where domestic wage costs land. July shows the external channel opening and the domestic channel narrowing at the same time.

On the month, CPI rose 0.3 per cent against 0.1 per cent in July 2025. CPIH, which adds owner occupiers' housing costs, rose to 3.1 per cent from 2.8. RPI stood at 3.2 per cent.

Definitions that must not be mixed

Two housing figures circulated after the release, and they are both correct. The CPI housing and household services rate was 4.6 per cent; the CPIH equivalent was 4.1 per cent. CPIH includes owner occupiers' housing costs — a large and comparatively stable component that dilutes the energy spike. They are different indices answering different questions, and they are not interchangeable.

Components breakdown / July 2026

Across the twelve CPI divisions, education (5.1 per cent), communication (5.0) and housing and household services (4.6) sit at the top of the table; recreation and culture (1.4), food (1.3), furniture (1.0) and clothing and footwear (0.5) sit at the bottom. The spread between the top and bottom of that table is four and a half percentage points — this is a dispersed inflation, not a general one.

The July step-up is a base-effect and pass-through story, and it is worth being precise about the mechanism. Ofgem sets the default tariff cap quarterly. The increase that took effect on 1 July 2026 raised the typical dual-fuel direct debit bill to £1,862 a year — a rise of £221 — on the consumption basis previously used, with Ofgem explicitly citing higher wholesale gas prices caused by the ongoing conflict in the Middle East. Set against a comparison month in which energy prices fell, the annual gas rate swung from −7.2 to +14.7 per cent in a single print.

A definitional trap worth avoiding

Ofgem simultaneously quoted the new cap as £1,663 per year, because from 1 July it updated its typical consumption values to reflect households using around 7 per cent less electricity and 17 per cent less gas than at the last review. The two figures — £1,862 and £1,663 — describe the same price decision on different consumption assumptions. Only the £1,862 basis is comparable with earlier quarters, and CPI itself measures unit prices, not bills, so neither number enters the index directly. Anyone comparing £1,663 with last quarter's cap is comparing across bases.

Motor fuels tell the same story at a different amplitude: still up 15.5 per cent year-on-year in July, but down from 21.3 per cent in June. Brent stood at $84 per barrel and UK front-month gas at 136 pence per therm at the close on 28 July. The energy complex is elevated, not accelerating.

This is also not a repeat of 2022, and the difference matters for what happens next. In 2022 the energy shock hit an economy with an extremely tight labour market. In 2026 it is hitting one with vacancies at a twelve-year low outside the pandemic and payrolled employment falling.

For an energy shock to become persistent inflation, it has to pass through wages. The August labour market data describe an economy in which that transmission is weakening, not building. Regular pay growth was 3.5 per cent in April to June 2026 — roughly consistent with the 2 per cent target once trend productivity is allowed for — and the private-sector figure was just 2.8 per cent, arguably below it. Real regular pay grew 0.5 per cent deflated by CPIH: positive, but thin.

The quantity side is unambiguous. Unemployment stood at 4.9 per cent for April to June. Employment was down 0.2 points on the year. Vacancies fell to 707,000, the lowest outside the pandemic since late 2014. Payrolled employees were down 78,000 in the year to June and 94,000 on the July flash estimate, with ONS reporting employer feedback that some small firms are not recruiting because of higher labour costs. Falling employment alongside above-target inflation is the classic supply-shock signature.

The one genuinely uncomfortable wage statistic is the sector split: public-sector regular pay is running at 6.1 per cent against the private sector's 2.8 — a 3.3-point wedge. Public settlements do not clear in a market, and they are the most plausible route by which an energy shock acquires persistence. It is the single number in this release most worth watching over the autumn.

That vote is the policy context the July print lands in. The July decision was taken with June's 2.6 per cent in hand; July's data confirmed the energy pass-through the Bank expected without adding evidence of second-round effects — the specific risk the dissenters cited. The Committee's own framing is that economic weakness is likely to help contain the strength of second-round effects by limiting pricing and bargaining power. Huw Pill dissented on precisely that point, citing more insidious second-round effects driven by catch-up dynamics in wage and price setting.

The Bank's central projection has CPI peaking at around 3.2 per cent in the fourth quarter of 2026 before falling to 1.7 per cent by the first quarter of 2028. The HM Treasury compilation of independent forecasts, published alongside the CPI release, puts Q4 2026 at 3.4 per cent, easing to 2.2 per cent by Q4 2027. Both are conditioned paths, not promises — and the Bank published multiple scenarios in July rather than one, which is itself informative about the uncertainty.

Our base case — and it is a house judgement built on those anchors, not a point forecast — is that headline CPI peaks in the 3.2 to 3.5 per cent range in Q4 2026 and then falls back through 2027 as the energy base effects reverse, provided services inflation continues to decelerate and private-sector pay stays near 3 per cent. On that path, Bank Rate most likely stays at 3.75 per cent into 2027. The single largest near-term input — Ofgem's cap level for 1 October to 31 December — was unpublished at our cut-off, which is precisely why the range is a range.

What the market is pricing

The 10-year gilt yield stood near 5.01 per cent on 25 August, with markets pricing one Bank Rate increase by year-end and a further quarter-point by early 2027; the Bank's July Report was conditioned on a market path implying a high chance of two hikes by Q3 2027. The market, in other words, is discounting an energy path, not a wage path — rate pricing did not improve in July even as the domestic cost data did. Energy itself may resolve the argument: the EIA's August outlook has Brent averaging about $85 in Q3 2026 and falling to a $69 average in 2027, assuming roughly 0.6 million barrels per day of disruption persists to end-2027. If that path holds, energy turns from a headline tailwind into a headwind during 2027.

One further date belongs in the forecast calendar: the Budget and the OBR's updated Economic and fiscal outlook land on 28 October 2026. Indirect taxes, duties and administered prices feed CPI directly and within months, and the fiscal arithmetic is tighter with the 10-year near 5 per cent. The OBR's March forecast — CPI averaging 2.3 per cent in 2026 — predates the energy shock and stands as the benchmark for how much the world has moved.

The strongest challenge to our base case runs like this: this is the second energy shock in four years; the public sector is settling at 6.1 per cent; inflation has been above target for nearly two years; and expectations may not survive another year above 3 per cent. Three MPC members already voted to raise, and markets agree with them — pricing a high chance of two hikes by Q3 2027. If that reading is right, the “look through the energy shock” case is a repeat of the 2021 mistake.

The counter is the data underneath: core flat at 2.6, services falling, food at a five-year-low contribution, private pay at 2.8, employment shrinking. Second-round effects require bargaining power that the quantity data say is absent. Both readings are coherent; they disagree about whether the labour market or the expectations channel dominates, and the next two CPI prints will arbitrate.

The cyclical–structural split is the cleanest way to hold the two horizons. The energy component is cyclical and reverses mechanically as base effects drop out during 2027. The structural risks are the public–private pay wedge and nearly two years of above-target inflation working on expectations — they are the ones that would still be there when the energy arithmetic has washed through. And the downside deserves equal weight: the EIA's 2027 path, continued payroll contraction and a flat October cap could bring CPI back to target faster than either the Bank or the consensus expects — at which point, with Bank Rate at 3.75, real rates would be tightening passively into a weakening labour market.

Two evidence caveats, stated plainly rather than buried. The Bank of England's own website refused automated access during this research, so the MPC vote and projection figures rest on a verbatim republication of the Bank's summary and on Reuters reporting — consistent across independent outlets, but second-hand. And the gilt yield is sourced from a market-data aggregator rather than an exchange feed; it is positioning colour, and no material figure in this piece depends on it alone.

First, Ofgem's cap for 1 October to 31 December, unpublished at our cut-off. It is the decisive near-term input: a rise above roughly 5 per cent pushes the Q4 peak above the consensus 3.4 and breaks the assumption that Q4 marks the top; a flat or falling cap compresses the peak sharply.

Second, the 16 September CPI release (August data). Services inflation is the line to read: continued deceleration supports the base case; two consecutive readings above 3.6 per cent would signal domestic persistence and invalidate it. Third, the MPC decision on 17 September — specifically whether the 6–3 split narrows or widens. A 5–4 vote would make a Q4 hike a live prospect.

Fourth, private-sector regular pay in the September labour market release: near 2.8 supports the base case; back above roughly 3.5 per cent invalidates the judgement that second-round effects are absent. Fifth, payrolled employees: continued contraction from the 30.3 million July flash estimate weakens the case for tightening. Sixth, Brent against the EIA's path from $85 toward a $69 average in 2027 — that is the variable that determines whether energy turns disinflationary next year. Seventh, the 28 October Budget and whatever indirect-tax and administered-price measures it carries. And eighth, the 10-year gilt: a sustained fall below 5 per cent would signal markets abandoning their hike pricing.

The July print, in short, is a worse headline than it is a release. The number went up; nearly everything that determines where inflation settles — services, core, pay, employment — went down or held. Whether that distinction survives the autumn is what the next two prints will tell us.