RELX's 19x multiple needs an AI revenue crack—not another scare

Two lawyers from Ledlin Lawyers review work on a tablet outside an Australian law courts building. The real workplace photograph fits a valuation debate about professional workflow and the information layer behind it.[6]

Priced: investors have already cut the old information-services premium because a general-purpose AI agent might sit between professionals and RELX's databases. New: the first half of 2026 produced no revenue crack. Legal underlying revenue grew 10%, group margin widened, and the shares still traded at about 19 times trailing adjusted earnings.[1][3][4]

That is not proof that RELX has beaten AI disruption. It is evidence that the market's February scare arrived before the operating damage. At an intraday 2,534.21p on July 23, the stock now asks a more exact question: will better models erode RELX's paid workflows before its own AI products deepen them?

The scare arrived before the revenue evidence

On February 3, Anthropic's release of legal plugins for Claude helped send RELX down as much as 15% in a day. The sell-off expressed a coherent fear: if a capable agent can review contracts, research law, and orchestrate tasks from a user's files, perhaps the expensive incumbent interface becomes a replaceable layer.[3]

The first-half result pushes against the timing of that thesis. RELX reported 7% underlying group revenue growth, 9% underlying adjusted operating-profit growth, and a 35.5% adjusted operating margin, up from 34.8%. Legal—the segment closest to the February argument—grew underlying revenue 10% and underlying adjusted operating profit 13%. Scientific, Technical & Medical grew underlying revenue 6%, while Risk grew 8%.[1]

RELX attributes double-digit growth in Law Firms & Corporate Legal to adoption of Lexis+ with Protégé, its AI-enabled research platform and integrated legal assistant, and says renewals and new sales were strong across Legal's key segments. That is useful management evidence, but it is not a disclosed product-level retention cohort.[1]

Those figures do not reveal whether every new sale was AI-related, nor do they show retention by product. They do establish a boundary: through June, the feared substitution had not become a visible slowdown in the segment most directly challenged.

The mechanism is data → workflow → renewal → margin

RELX is often described as a publisher, but the multiple rests on a more useful mechanism. Risk combines public, licensed, contributed, and proprietary data with decision tools. Elsevier places scientific and medical content inside research and clinical workflows. LexisNexis sells legal information, analytics, and drafting tools to customers who care about provenance, currency, and whether an answer can survive professional scrutiny.[1][2]

That last mile matters. A model can make the interface easier to reproduce; it does not automatically recreate permissioned content, editorial maintenance, citations, customer-specific integrations, or accountability. RELX can therefore use AI to make its own corpus more valuable. It can also lose if an outside agent becomes the preferred interface and reduces the incumbent to a lower-priced data supplier.

The renewal clock makes both outcomes slow. Legal generated 85% of its 2025 revenue from subscriptions, often under multi-year contracts. Across RELX, subscriptions supplied £2.633 billion of the £4.871 billion reported in the first half of 2026.[1][2] Embedded revenue softens an immediate shock, but it can also delay the evidence. A customer may experiment with a new tool today and change seats, modules, or price negotiations only when a contract comes up for renewal.

The causal chain to watch is therefore:

AI usefulness → customer workflow share → renewal and pricing → underlying revenue → margin.

The first-half print clears the final two links. It does not yet settle the first three.

Freeze the multiple before telling the moat story

RELX reported 128.5p of adjusted earnings per share for 2025, including 63.5p in the first half. Replace that old half with the 68.6p reported for the first half of 2026 and trailing adjusted EPS becomes 133.6p. Dividing the 2,534.21p intraday share price by that observed denominator produces 19.0 times earnings, or an earnings yield of roughly 5.3%.[1][2][4]

That calculation is cleaner than doubling a half-year result or mixing the current price with reported EPS. A simple annualisation of the new half would produce 137.2p and 18.5 times, but it would assume away seasonality. RELX's adjusted measure excludes acquisition-related intangible amortisation and other specified items; readers who prefer statutory earnings should demand a higher apparent multiple. The trailing approach also avoids treating constant-currency EPS growth as guaranteed full-year sterling growth.

What does 19 times imply? It does not require RELX to return to a glamour multiple. It does require high-single-digit operating growth to remain durable enough that earnings compound while the multiple holds. If Legal and STM slow because customers shift workflow share to general-purpose agents, the denominator and the multiple can fall together. If RELX keeps the interface, the trusted data, and the renewal, today's valuation leaves room for the earnings base—not exuberant rerating—to do most of the work.

The cheapest-looking number is partly bought

The strongest counterweight to the “quality at a discount” reading sits on the balance sheet. RELX plans a £2.25 billion buyback in 2026, roughly 5% of the approximately £43 billion market capitalisation cited in the interim release. It had completed £1.75 billion in the first half and another £100 million after June.[1]

That retirement supports per-share growth, but it is not free operating progress. Net debt rose to £8.733 billion at June 30, and net debt to EBITDA reached 2.3 times, versus 2.0 times at the end of 2025. Cash conversion remained strong at 98%, yet the higher leverage means part of the EPS acceleration comes alongside a more indebted equity claim.[1][2]

The distinction matters for the multiple. A buyback below intrinsic value can create value; a buyback financed while the competitive duration is under debate also spends balance-sheet capacity before the debate is resolved. Investors should separate three contributors:

underlying revenue and margin growth;
fewer shares outstanding;
any future rerating once AI fears recede.

Only the first is proof that customers still assign more value to the products.

A three-multiple sensitivity

Using the 133.6p trailing adjusted denominator makes the burden visible. These are arithmetic cases, not target prices or consensus forecasts:

15x / 2,004p: Legal growth fades, renewal pressure appears, and the market values RELX as a mature information vendor rather than a durable workflow compounder. That is about 21% below the July 23 intraday price.
19x / 2,538p: the current case. High-single-digit underlying growth broadly persists, but investors keep a discount for unresolved AI substitution and elevated leverage.
21x / 2,806p: Legal and STM show that AI is raising product value, margin progress continues, and the balance sheet begins to normalise. That is about 11% above the freeze price.

The asymmetry is instructive. Upside does not need a return to the former narrative; it needs operating durability plus modest multiple repair. Downside does not need Claude to replace LexisNexis outright; slower renewals and ordinary compression are enough.

Strongest counterweight: six clean months do not settle a platform shift

The bullish reading of the first half is powerful because the most exposed segment accelerated. It is also incomplete. Legal represents about one-fifth of group revenue, so RELX can post healthy consolidated growth while a product-level contest is still forming. Multi-year contracts make the income statement resilient, but they also turn disruption into a cohort problem: the damage may first appear in new-logo wins, module expansion, seat counts, or concessions before it reaches reported revenue.[1][2]

Management says AI lets RELX launch higher-value products faster and keep cost growth below revenue growth. The margin result is consistent with that claim, but the release does not disclose enough customer-level evidence to prove it. The missing data are as important as the reported numbers: renewal rates for AI-enabled legal products, pricing versus non-AI modules, usage, and how much new business is expansion rather than migration from older RELX products.

That leaves the valuation posture at wait for proof, not “AI risk disproved.” The stock looks inexpensive only if the current earnings stream remains a durable base.

Falsifier

The cautious view is falsified if RELX carries Legal's roughly 10% underlying growth through the full year, expands group margin again, and brings net debt to EBITDA back toward its year-end 2025 level after completing the buyback. That combination would connect AI-era demand, operating leverage, and balance-sheet repair. A strong share-price reaction by itself would not.

The bearish AI thesis gets its own clean test: a material Legal deceleration accompanied by weaker group margin, especially if management points to renewal, pricing, or competitive pressure rather than currency or event timing. That would be the revenue crack the February sell-off anticipated.

Watchlist

October 22, 2026 — nine-month trading update: require Legal to remain a growth leader and look for any change in language around new sales, AI-enabled products, customer demand, or pricing. The update may be qualitative, so wording changes matter.[5]
By December 31, 2026 — buyback completion: RELX says the remaining £400 million will be deployed before year-end. The relevant question is not whether the programme finishes, but whether strong cash generation prevents another step up in leverage.[1]
Full-year 2026 results — renewal and balance-sheet proof: compare Legal's full-year underlying growth, group margin, cash conversion, and net debt to EBITDA with the first-half anchors. The exact release date was not yet listed on RELX's calendar at the July 23 freeze point.[1][5]

RELX has not proved that proprietary information businesses are immune to general-purpose agents. It has shown that the feared disruption is not yet in revenue, even where the challenge is most direct. At 19 times trailing adjusted earnings, the market is no longer paying for invulnerability. It is waiting to learn whether the moat is the screen, the data behind it, or the professional habit that binds both together.

cronfeed.work