GSK’s Core Problem 

In pharmaceutical company valuation, nobody guides on EBITDA. They guide on Core — and Core removes precisely the impairments, amortisation and contingent payments that show whether a decade of dealmaking created value. In the second quarter of 2026 GSK’s Core operating profit rose 6% to £2.8bn. Its Total operating profit fell 76% to £481m. 

The pharma & life sciences paper in the Bloor Finance series: a CFO-perspective benchmark of GSK plc (LSE: GSK) against AstraZeneca and Roche, on the 2025 full-year record and the half-year results to 30 June 2026, published between 23 and 28 July 2026. 

1. Where the opening paper left off

The opening paper in this series made one claim and asked twelve questions. The claim: wherever an organisation embeds AI, cost leaves the line EBITDA (earnings before interest, tax, depreciation and amortisation) sees and reappears on lines it does not. The technology paper showed it at Oracle as compute debt. The banking paper showed it at Lloyds as labour debt. This is the pharma paper the opening paper promised, and its sector line was the bluntest of the twelve. 

Pharma’s distortion is self-inflicted and respectable. ‘Core’ or ‘adjusted’ earnings strip out precisely the M&A amortisation and impairment that reveal whether serial dealmaking created or destroyed value. 

The opening paper cited a £471m write-off on one terminated GSK programme in 2025. The programme was belrestotug, an anti-TIGIT antibody; GSK’s full-year release records ‘an impairment charge of £471 million related to the termination of the belrestotug development programme’, within £880m of intangible impairments for the year. A week before the opening paper was published, GSK reported a larger one: £1,334m on camlipixant. 

GSK is the anchor because it carries every one of the sector’s lies in one set of accounts, in sterling, under UK tax rules. AstraZeneca and Roche, both named in the opening paper, are the peers. 

2 · Why does EBITDA mislead in pharmaceutical company valuation?

Because the costs that decide what a pharmaceutical company is worth failed research, bought pipelines, payments to partners when drugs succeed are the costs EBITDA and its pharma cousin, Core, are built to remove. The majors do not guide on EBITDA. They guide on Core, which is the same logic applied with more precision. 

They say so plainly. ‘GSK provides earnings guidance to the investor community on the basis of Core results’, its half-year release states, and it ‘is not able to give guidance for Total results as it cannot reliably forecast certain material elements’. AstraZeneca ‘is unable to provide guidance on a Reported basis’ for the same reason, naming intangible impairments among the items it cannot forecast. Roche guides on core earnings per share. 

Here is what that means in one quarter. GSK’s second-quarter Core operating profit was £2,800m, up 6%. Total operating profit the IFRS number was £481m, down 76%. Of every pound of Core profit, about 17 pence survived to the statutory line. 

The half-year shows the shape. Core operating profit was £5,450m; Total was £2,774m. The £2,676m between them 49% of Core was £2,067m of intangible impairment, £385m of intangible amortisation, £770m of transaction-related charges, £47m of restructuring, less a £593m credit from divestments and other items. A year earlier the gap was £925m, or 18%. 

Core is not a lie. It answers a narrower question than the one a board should be asking. 

None of this is hidden. GSK publishes the bridge line by line. The problem is which number the market prices, which number management is paid on, and which number a CFO forecasts. All three are Core. 

3· R&D capitalisation: why is bought research an asset and home-grown research an expense?

Under IAS 38, research a company does itself is expensed as incurred, and development is capitalised only when strict criteria are met which in pharma usually means very late. Research a company buys a licence, a milestone, a whole biotech is capitalised on day one as an intangible asset, and when it fails, the write-off is excluded from Core. 

That asymmetry is the first lie. A failed in-house programme costs Core every year it runs. A failed acquired programme costs Core nothing, ever. 

GSK’s first half makes it concrete. Core R&D was £3,214m, 20.0% of turnover. Total R&D was £5,158m, up 48%. The £1,944m difference was almost entirely impairment: £1,877m of intangible impairment charged to the R&D line. More than a third of what GSK reported as research and development in the half was not research at all. It was the write-down of research it had bought. 

The largest item was camlipixant. GSK bought it with Bellus Health in April 2023 for about $2bn, roughly double Bellus’s previous closing share price. In 2026, after the CALM-1 and CALM-2 pivotal trials, GSK decided not to progress it in refractory chronic cough, and impaired it by £1,334m. Its carrying value at 30 June 2026 is £104m, for a different indication. In the same quarter GSK fully impaired £371m following the termination of assets under its Alector collaboration. 

The pipeline that remains sits on the balance sheet as £16,802m of other intangible assets and £7,381m of goodwill at 30 June 2026. How much of it is capitalised development, software or acquired in-process R&D is not split out in the half-year release, and is shown here as n/d. 

Then there is AI. GSK’s chair says the company continues ‘to invest significantly in the transformational capability afforded by AI/ML’, and the savings in its new Accelerate Growth programme are ‘expected to be enabled by technology and AI’. The opening paper asked the question this raises, and it still stands: is that spend research expense, capitalised asset, or an off-balance-sheet commitment to a discovery platform and over what life is it carried? The documents cited here do not say. The amount and the line are n/d. 

THE QUESTION CFOs AREN’T ASKING

Over a full cycle, does the amortisation and impairment you exclude from Core roughly equal the acquisition premiums you paid that is, has your serial dealmaking created value or merely recycled it? And what is AI drug-discovery actually costing you, on which line, over which useful life? 

4 · Milestone payments: why does a successful drug produce a charge outside Core?

When a pharma company buys an asset with payments that depend on future milestones or sales, it books a contingent consideration liability and remeasures it every quarter. Good news for the drug is a charge against Total profit and Core removes the charge. 

GSK carries the sector’s clearest example. At 30 June 2026 its contingent consideration liabilities were £6,781m £1,376m current and £5,405m non-current mostly owed to Shionogi under the former ViiV Healthcare joint venture arrangements. In the first half the charge on that liability was £680m: £487m of remeasurement and £193m of discount unwind. The second-quarter charge of £392m arose, in GSK’s words, from higher sales forecasts. 

The cash is real. GSK paid £757m of contingent consideration in the half, £710m of it to Shionogi. None of the £680m charge reached Core. The full-year pattern is the same: other operating expense of £488m in 2025, after £1,839m in 2024, ‘principally arising from the remeasurement of CCLs’ and the Pfizer put option. 

In pharma, success has a price. Core does not show it. 

Milestones flow the other way too. AstraZeneca reports a separate Collaboration Revenue line, which in the first half of 2026 was $77m, all sales milestones $44m on Farxiga and $32m on Crestor. It is small and clearly labelled. It is also revenue that depends on a partner’s sales, not on AstraZeneca’s own, and it belongs in a different valuation multiple. 

5 · Royalty income: recurring earnings or a one-off?

Royalty income sits inside Core operating profit, where a valuation multiple treats it as recurring. Some of it is. Some of it is a legal settlement. 

GSK’s royalty income was £879m in 2025, up from £639m, and the release explains why: ‘The full year included historic royalties recognised in association with the settlement of an IP dispute.’ The first half of 2026 shows the reversal £399m against £426m, and £204m in the second quarter against £246m, the comparative having carried those historic royalties. GSK guides to £850–900m for 2026. 

At 9.0% of 2025 Core operating profit and 7.3% of the first half’s, royalty income is not a rounding item. Capitalised at a pharma multiple, a one-off settlement inside Core becomes a permanent piece of enterprise value. 

Roche shows the other side of the ledger. Its pharmaceuticals division booked CHF 489m of royalty income in the first half of 2026, and CHF 853m of royalty expense. A royalty is a share of someone else’s product. The honest gauge separates the contractual run-rate from catch-ups and settlements, and nets what is paid against what is received. 

6· R&D tax credits (RDEC): where does the UK credit land?

The UK’s R&D expenditure credit is a taxable credit that large companies account for above the tax line, so it lifts operating profit rather than reducing the tax charge. Since 1 April 2024 the merged scheme gives relief at 20% of qualifying expenditure. 

HMRC’s own manual describes the credit as one ‘to be brought into account as a receipt’ in calculating profits. For a research-intensive group that is a subsidy presented as performance. It sits in operating profit, and nothing in GSK’s definition of Core adjusting items removes it, so it sits in Core as well. 

How much it is, and on which line, is not disclosed in any document cited here not by GSK, and not by AstraZeneca, the other UK-headquartered major. It is shown as n/d. This paper does not estimate it. 

The point is not the size. It is portability. A UK credit is a function of where the research is done and of a tax policy that has changed several times since 2013. A buyer who would move the research or a government that would change the rate would strip it out. A board should know how much of its Core margin is policy. 

7 · The patent cliff: what happens to EBITDA when exclusivity ends?

A patent cliff does not appear in EBITDA until it happens, and then it appears all at once. The honest gauge is revenue at risk by year of loss of exclusivity, set against the cost of replacing it. 

GSK’s cliff is dolutegravir, the backbone of its HIV franchise. Dolutegravir products sold £5,648m in 2025 17.3% of Group turnover of £32,667m and lose exclusivity in 2028–2030. GSK expects ‘a stable to improving operating margin’ through that period and a 2031 sales outlook of more than £40bn. 

The plan to get there includes Accelerate Growth: £1.9bn of annual savings, fully realised by 2029, for expected total costs of £2.4bn, of which £2.1bn is cash, most of it in 2026 and 2027. GSK treats the programme as a Major restructuring programme, so its costs sit in Adjusting items. The savings will not. 

The cost of crossing the cliff is adjusted out. The benefit is not. 

The peers show what the cliff looks like when it arrives. AstraZeneca’s Farxiga revenue fell 6% to $4,042m in the first half of 2026, with US sales down 17% as generics launched; Brilinta fell 64% after generic entry. Roche says the first US biosimilars of Xolair could enter in the second half of 2026, and does not expect US biosimilars of Perjeta before 2028. 

8 · How do GSK, AstraZeneca and Roche compare on Core versus reported profit?

All three guide on Core; all three report a statutory profit materially below it; none discloses its R&D tax credit or what AI-enabled discovery costs and on which line. GSK’s gap is the widest this half because its largest impairment landed in it not because its accounting is different. 

The three report in three currencies, so the table compares ratios only. Each company’s absolute figures are in its own reporting currency, under IFRS. 

Honest gauges GSK against AstraZeneca and Roche (half-year to 30 June 2026, results published 23–28 July 2026) 

Company (currency) Core vs reported operating profit, H1 2026 Gap as % of Core (H1 2025) Amortisation & impairment excluded, H1 2026 Core R&D % of revenue Guidance basis Patent cliff in view RDEC; AI cost line 
GSK (£m) Core 5,450; Total 2,774 49% (18%) 385 amortisation + 2,067 impairment 20.0% Core; ‘cannot reliably forecast’ Total Dolutegravir, 17.3% of 2025 sales; LoE 2028–30 n/d; n/d 
AstraZeneca ($m) Core 10,510; Reported 7,410 29% (n/d) 2,536 amortisation & impairment 23.2% Core; ‘unable to provide guidance on a Reported basis’ Farxiga US generics Q2 2026; Brilinta −64% n/d; n/d 
Roche (CHF m) Core 11,856; IFRS 9,671 18% (14%) c.400 amortisation + c.1,000 impairment 19.0% Core EPS Xolair US biosimilars possible H2 2026; Perjeta US not before 2028 n/d; n/d 

GSK, AstraZeneca and Roche figures are from each company’s half-year results (28, 27 and 23 July 2026). Gap = Core less reported operating profit, divided by Core; Core R&D % = Core R&D ÷ turnover/total revenue/sales Bloor arithmetic. AstraZeneca’s H1 2025 Core operating profit was not retrieved, so its comparative gap is n/d. Roche amortisation and impairment are the approximate pre-tax figures Roche gives in its half-year report. Roche’s guidance basis is as reported by Teleborsa (23 July 2026). n/d not disclosed in the cited document; never treated as zero. Currencies differ; only ratios are compared. No company is ranked; the comparison is indicative. 

Two contrasts deserve a line each. AstraZeneca excludes the most amortisation and impairment in absolute terms, reflecting a decade of acquisitions; its gap is steadier because less of it is impairment. Roche’s impairments rose from about CHF 0.2bn to about CHF 1.0bn in a year, across three named programmes the same pattern as GSK’s, at a smaller scale relative to Core. 

The empty column is the finding. Three of the world’s largest research organisations, all investing in AI-enabled discovery, and none of them tells a reader what it costs, where it sits or how long it is carried. 

9 · What will a buyer do to your EBITDA?

A buyer of a pharmaceutical business does not pay for EBITDA. It pays for a risk-adjusted pipeline and it structures the price so that the seller carries the risk EBITDA hides. 

The working method is risk-adjusted net present value (rNPV): each asset’s expected cash flows, weighted by its probability of reaching the market at its current phase, discounted, and summed with the marketed portfolio. It is the honest alternative to an EBITDA multiple because it prices the failures in advance rather than adjusting them out afterwards. 

GSK’s own deals show the mechanics. Camlipixant was bought at roughly double the target’s share price; three years later the asset carries £104m. Nuvalent, completed on 15 July 2026, was bought at $124 a share about 40% above the prior close for an equity value of about $10.6bn, or $9.4bn net of cash, with low single-digit percentage dilution to Core EPS expected in 2026–2028. Its premium becomes intangible assets whose amortisation, and any impairment, Core will exclude. 

A buyer looking at a pharma target normalises in four places. It treats contingent consideration as debt for GSK, £6,781m beside £15,132m of net debt. It strips settlement royalties out of the run-rate. It treats restructuring as recurring when it recurs. And it removes credits, like RDEC, that will not survive a change of owner or jurisdiction. Then it values the pipeline asset by asset, and pays part of the price later, in milestones, so that if the asset fails the seller shares the write-off. 

10 · What should pharma CFOs and boards track instead of EBITDA?

In pharma the remedy is not a new metric but the existing ones read in full. The pharma edition of the depreciation-aware dashboard, applied to GSK’s first half, reads as follows. 

— The Total-to-Core gap, over a cycle, not a quarter. GSK: £2,676m or 49% of Core in H1 2026; £925m or 18% in H1 2025; £1,851m or 19% in FY2025. A gap that never closes is not an adjustment. It is a cost. 

— Impairment against purchase price, programme by programme. GSK: belrestotug £471m (2025); camlipixant £1,334m against a c.$2bn acquisition; Alector £371m. Every acquired asset should be tracked to its outcome. 

— Core and Total R&D intensity side by side. GSK: 20.0% Core, 32.2% Total. The difference is research bought and lost. 

— Contingent consideration as debt. GSK: £6,781m of contingent consideration beside £15,132m of net debt £21,913m together. 

— Royalty quality: contractual run-rate separated from settlements. GSK: £879m in 2025 including historic royalties; £850–900m guided for 2026. 

— R&D tax credits on a disclosed line. GSK: n/d. A credit that sits in operating profit should be visible there. 

— Revenue at risk by year of loss of exclusivity, against the cost of replacing it. GSK: dolutegravir 17.3% of 2025 turnover, 2028–2030; Accelerate Growth costs £2.4bn, excluded from Core. 

— Cash against Core. GSK: free cash flow £2,809m in H1 2026, 52% of Core operating profit; £4.0bn in 2025, 41%. 

THE QUESTION CFOs AREN’T ASKING 

If your restructuring is excluded from Core and its savings are not, how many years of ‘one-off’ restructuring has your Core margin quietly absorbed and what would that margin be if the cost and the benefit sat on the same line? And when your next acquired programme fails, will your board learn its price from the impairment note, or from a pipeline valuation it saw a year earlier?

— AI-enabled discovery: amount, line and useful life. GSK: n/d.

11 · The Bloor lens

For a pharma CFO the work is specific. Publish the Total-to-Core bridge with the same prominence as Core, and over five years, not one. Track every acquired programme from purchase price to outcome. Put contingent consideration in the net debt that management discusses. Split royalty income into run-rate and one-off. Disclose the R&D tax credit and its line. And say what AI-enabled discovery costs, where it sits and over what life it is carried. 

GSK is not the weak case. Core operating profit rose 6% in the half, free cash flow rose 54%, guidance moved to the upper half of the range, and the bridge from Core to Total is published in full. That is why it is the right case. If the best-disclosed UK pharma major can lose half its first-half Core profit on the way to the statutory line, the sector’s Core numbers should be read with the bridge beside them. 

Pharmaceutical company valuation done honestly starts where Core stops: with the failures, the deferred payments and the cliff, priced in advance. 

THE BLOOR TEST 

Add nothing back. Over five years, take Total operating profit, subtract the cash paid for acquisitions and contingent consideration, and set it beside Core. If the two lines converge, Core was a timing convenience. If they do not, Core was the valuation and the difference is what dealmaking cost. 

Reform adjusts the past. Reset designs the future.

Watch for the remaining When EBITDA Lies industry papers as they publish. For boards that want an early diagnostic read of their sector’s honest gauges before their industry paper lands, the Bloor Finance Desk is available for a confidential conversation. 

Analysis only; not investment advice. 

Frequently asked questions

1. How do you value a pharmaceutical company? 

Usually by risk-adjusted net present value: each pipeline asset’s expected cash flows weighted by its probability of success and discounted, added to the value of marketed products. EBITDA or Core multiples are a cross-check at best. GSK and AstraZeneca both guide on Core and both state that they cannot forecast their statutory results, mainly because they cannot forecast impairments and contingent consideration. 

2. What is R&D capitalisation under IAS 38?

IAS 38 requires research to be expensed and allows development to be capitalised only when specific criteria are met, including technical feasibility and the intention and ability to complete and use or sell the asset. Acquired research licences, in-process R&D bought in a deal is capitalised as an intangible asset. When it fails, it is impaired: GSK’s first-half 2026 R&D line included £1,877m of such impairment.

3. What is the pharma patent cliff?

The fall in revenue when a drug loses exclusivity and generics or biosimilars enter. GSK’s dolutegravir products sold £5,648m in 2025, 17.3% of turnover, and lose exclusivity in 2028–2030; AstraZeneca’s US Farxiga sales fell 17% in the first half of 2026 as generics launched. 

4. How are milestone payments accounted for? 

When a buyer agrees to pay future milestones or sales-linked amounts, it records a contingent consideration liability and remeasures it each period through the income statement. GSK held £6,781m of such liabilities at 30 June 2026 and charged £680m against the main one in the half, all excluded from Core. A company receiving milestones may report them as revenue, as AstraZeneca does in its Collaboration Revenue line. 

5. What is the R&D expenditure credit (RDEC)? 

A UK tax relief for R&D, paid as a taxable credit that large companies account for above the tax line, so it increases operating profit. Under the merged scheme in force since 1 April 2024 the rate is 20% of qualifying expenditure. Neither GSK nor AstraZeneca discloses the amount in the results documents cited here. 

9 · Sources & basis

Every figure in this paper is drawn from a named, dated document. The principal sources are listed below with links, and a full Source Note listing the origin, pinpoint reference and status of every figure accompanies this paper.

Gaps, ratios and intensities are Bloor arithmetic on reported figures. R&D tax credits, the cost and accounting line of AI-enabled discovery, and the split of intangible assets are not disclosed in the cited documents and are shown as n/d, never as zero.

No company is ranked. This is analysis for a CFO audience and is not investment advice.

Print Article for this blog Download here