Halmas 21% EPS Record|-14% Crash on AI Customer Concentration

Chapter 1: Record Profits, Wrong Stock

Halma PLC shares fell 14% on 11 June 2026, the same day the company reported its best full-year results in its history. That is the paradox the market left unresolved. Adjusted earnings per share rose 21% to 114.05p. Revenue climbed 15% to £2,582.3m. Adjusted EBIT margin expanded 140 basis points to 23.0%. Every headline number was a record. The stock lost roughly a seventh of its value in a single session.

The bottleneck is not in the earnings — it is in the disclosure. Buried inside the results was a detail that invalidated the valuation frame Halma has traded on for two decades: one unnamed customer now accounts for roughly 20% of group revenue and generated approximately half of all organic growth in the year just reported.

That customer sits inside Halma's photonics division, Avo Photonics, which grew revenue 52% in the year to 31 March 2026 driven entirely by demand from that single relationship — a data centre operator whose capital expenditure programme Avo supplies with fibre-optical interconnects. When the year ahead guidance came alongside the record results, photonics growth was guided at 30%, not 52%. The deceleration is not a reversal. But the single-name exposure behind it turned a compounder into something closer to a contract manufacturer with one anchor client.

The residue of chapter 1: the selloff answered one question — what investors thought of the disclosure — but opened a harder one. Was the market right to reprice, or did it overcorrect?

Chapter 2: How a $9 Million Acquisition Became a Group-Defining Relationship

The concentration did not appear overnight. Its origin is a 2011 acquisition that cost Halma an initial $9.0m and contingent payments of up to $11.0m. At the time, Avo Photonics had revenues of $5.7m and profit before tax of $1.0m. Halma bought a small specialist fibre-optic component maker with no indication it would become the engine of the group's most explosive growth period.

By FY2026, Avo had grown to the point where its photonics division was generating approximately £516m of revenue — implying roughly a fifth of Halma's £2,582.3m total. A business acquired for low double-digit millions fifteen years ago now sits at the structural heart of the group's growth profile.

This is where the buried paradox emerges. Halma's equity story has rested on diversification: 40-plus autonomous subsidiaries across safety, environment, and health sectors, with no single division ever dominating. The FY2026 disclosure punctures that reading. Not because management obscured it — the results were transparent — but because investors who owned Halma for its compounder model owned something different by the time the results landed.

Diploma, the FTSE 100 distributor, also reported data centre demand in its recent results and was well received. Diploma's announcement carried no equivalent single-customer concentration. The contrast is instructive: similar demand exposure, different ownership structure, opposite market reaction. The market is not repricing AI data centre demand generally — it is repricing single-name dependency specifically.

Halma invested £123m in R&D and £56m in capex in FY2026, and completed two further acquisitions since year-end for approximately £75m combined. That spending signals confidence. But spending from a position where one relationship accounts for half of organic growth is different from spending as a genuinely diversified compounder. The tension between the group's historical model and its current revenue structure is the question the AGM on 23 July cannot avoid.

Chapter 3: The Analyst Conflict the Market Left Open

Within hours of Halma's 14% — at one point 17% — intraday fall, Panmure Liberum analyst Alex O'Hanlon raised his earnings per share forecast and lifted his target price from 4,270p to 4,620p. His argument was direct: Halma's equity story is "about much more than just Photonics", the company is "investing from a place of strength", and the selloff "creates a buying opportunity". Projected net debt to EBITDA of 0.6x by 2027 leaves what he described as "significant firepower for further acquisitions".

Citigroup took the same directional view, upgrading Halma to 'buy' and setting a 4,600p target. Exane BNP raised its recommendation to 'outperform' with a 4,550p target. On the other side, Goldman Sachs — which had held a 'buy' rating — cut its target from 5,060p to 5,010p, a reduction modest in magnitude but directionally opposite to the upgraders.

These are not vague market reactions. They are named analysts drawing opposing conclusions from the same filing on the same day. The upgraders argue that the selloff mispriced a temporary concentration within a structurally sound compounder. Goldman's cut implies the previously assumed target was based on a revenue mix that turns out to be different from what Halma actually is.

The conflict between those two readings — reset or rerating — is the decision the holder faces today. A reset means the shares are cheaper than before the results on unchanged fundamentals. A rerating means the model itself has changed, and the previous multiple was never the right anchor.

What neither camp addresses directly is the customer's own spending trajectory. Avo Photonics supplies a single data centre client. If that client's AI infrastructure programme holds or accelerates, 30% photonics growth next year maintains a meaningful contribution. If the client slows, defers, or diversifies suppliers, there is no secondary revenue stream to offset the gap. The upgraders' case rests on a data centre spending assumption that is not in the results filing.

Chapter 4: The July AGM Test and the Single Variable That Resolves It

The verification anchor for Halma's contested read is the AGM on 23 July 2026. The meeting carries two functions. The first is formal: shareholder approval of the 7% increased final dividend at 15.11p per share. That approval is rarely denied and would confirm uninterrupted cash return to shareholders. The second function is the more important one: management will face questions on photonics guidance and customer structure that the full-year results filing did not fully resolve.

The specific variable to watch is not the dividend — it is any updated commentary on the unnamed data centre customer's contract status or spending outlook for the year ahead. The 30% photonics growth guidance in the results was stated without customer-level detail. If the AGM produces even qualified confirmation that the relationship is contracted or on long-term framework terms, the reset reading is supported. If management continues to carry the guidance without that confirmation, the rerating reading retains its force.

The counter-evidence against the buying case is not trivial. OpenAI paused its Stargate UK data centre project in April 2026, citing unfavourable regulatory conditions and high energy costs. That pause does not name Halma's customer, but it signals that large-scale data centre commitments in the UK are not unconditional. A holder of Halma is implicitly long a single unnamed AI infrastructure programme in a market where the largest AI lab publicly deferred its own UK build-out.

Management guided that adjusted EBIT margin for the coming year would be in line with FY2026, excluding a one-off from the Nuvonic transaction. That guidance holds even with a 30%-not-52% photonics growth rate, implying the broader portfolio absorbs the deceleration without a margin crack. That is the part of the compounder model that remains intact. But intact margins at 30% photonics growth do not answer whether the stock was de-rated correctly or excessively.

A holder should watch for AGM commentary on contract length and customer concentration before deciding whether the reset has fully occurred. A watch-list investor should not enter ahead of that disclosure — the 14% drop means the shares reflect partial resolution, not full resolution. The invalidation condition: evidence that photonics growth falls materially below 30%, or that the unnamed customer has reduced its commitment, would confirm a rerating rather than a reset and would remove the basis for the current analyst upgrade consensus.