H1 2026: A Record Half That Did Not Test the Thesis

Thesis status: Affirmed. The release calendar is confirmed, earnings quality has separated from earnings, and the incentive structure is now the clearest read available on what management itself expects from The Witcher 4.

Thesis Refresher

The central question is whether CD Projekt ($CDR.WA) can convert restored credibility into disciplined execution on The Witcher 4, and whether that execution translates long-cycle creative investment into durable value creation. The primer set the conditions under which that question gets answered. Memo No. 1 affirmed the thesis against FY25 and added two frame refinements, the GOG disposal and the Cyberpunk ecosystem strategy. Memo No. 2 recorded the Witcher bridge activating and the incentive programme tightening.

The horizon remains 36 to 48 months. This memo covers the H1 2026 reporting cycle and is the first struck off the completed reverse-DCF model.

H1 2026 in Brief

Every line in the table moved the right way, and the half deserves to be stated plainly before it is qualified.

The qualification is composition. Almost every headline above traces to a line that did not meaningfully exist twelve months ago. IP licensing rose to PLN 94.8m from over 100 partners, and its increase of PLN 85.4m is larger than the PLN 80.3m increase in total revenue. Own-project revenue fell 1.5% to PLN 340.3m. Cyberpunk own-project revenue declined as the Switch 2 launch quarter in the base lapped, and Witcher own-project revenue rose on the Game Pass entry Memo No. 2 identified as the opening move of the bridge.

Second quarter revenue of PLN 244.2m produced EBIT of PLN 148.3m, a 60.7% margin. Management characterised the licensing booked in it as spectacular.

The margin expansion is reported here rather than attributed. Splitting it between operating leverage and licensing mix requires a cost base for the licensing line, and the interim disclosure does not carry one. The range a reasonable attribution would span is too wide to be worth publishing.

Development spending set its own record. The PLN 355.6m of cash development expenditure in the half included PLN 205.8m in 2Q26, the largest quarter to date, taking the development-project balance to PLN 1,512.5m. Operating cash flow was flat while that outlay rose 47%, and the PLN 90.7m received on the GOG disposal covered part of the difference.

What the Price Requires

Struck at PLN 231.20, the 2 September close. The shares fell 4.8% from the 25 August mark of PLN 242.90 on a day the company reported a 23% revenue increase and a record margin.

The trailing twelve months carry a higher EBIT margin than any full year on record, and no margin the business can print funds this price, so the requirement has to be expressed in growth. At a held 50% EBIT margin, a level only FY2020 and FY2025 have cleared, the price requires a 26.1% revenue CAGR for nine consecutive years, which puts FY2035 revenue at PLN 7.66bn against FY2025's PLN 0.87bn.

The requirements fell against the August strike, and that is a caveat against our own number rather than good news: part is the lower price, most is the base. FY2026E revenue is now PLN 947m rather than PLN 800m, and the increase is largely one half of licensing — the requirement got easier because a possibly non-recurring item enlarged the base, which is why the held-margin line rather than the active-path figure is the one to carry.

This section characterises the financial conditions the current enterprise value implies. It is not a price target, a fair-value estimate, or a view on whether the shares are mispriced.

Watch Condition Assessment

C1: Development milestones and resource allocation

🟢 Affirmed → Headcount and composition affirm; the share metric reads down, and the share metric is what needs fixing.

The internal team reached 1,045 at 31 July 2026 from 975 at 30 April. The Witcher 4 team added six to 519, and its share of the internal workforce fell below half for the first time, to 49.7% — because the group grew around it, with Cyberpunk 2 up 21 to 184 and shared services up 23 to 196. A share test misreads a growing denominator, and a refinement to absolute headcount plus composition is proposed for the FY2026 restatement.

Songs of the Past runs to roughly 220 people outside these totals: 170 at Fool's Theory, 10 at CD Projekt, 40 outsourced. Two unannounced projects remain in capitalised development, down from four, consistent with Songs of the Past and the Switch 2 port having been announced.

The July components sum to 1,046 against a published total of 1,045.

C2: Capitalised development versus cash discipline

🟢 Affirmed → The record half was absorbed, and the capital retained to fund the rest of it.

The balance sheet carries no debt beyond leases. Cash, deposits and bonds stood at PLN 1,285.6m at 30 June 2026, down 3% from the FY25 close, net cash at PLN 1,249.8m, against a development-project balance of PLN 1,512.5m. Two new facts belong here. No dividend was declared from 2025 earnings — the full PLN 609.5m went to reserve capital, against PLN 99.9m paid in each of FY2024 and FY2025. A buy-back authorisation was granted on 23 June 2026.

The reading is capital retained to fund the ramp, not strain. Management declined to call a peak in development spending: they did not know where the end of the increases lay, and were not guiding to one.

C3: Management communication consistency

🟢 Affirmed → The strongest single communication data point since the primer, against two thinnings in disclosure.

The Stage 2 write-down is the behaviour the primer's standard asks for. Management judged the PLN 3bn target for 2024 to 2027 unlikely, reversed PLN 11.07m of previously recognised cost, and disclosed it unprompted in the same release as a record half and against their own compensation.

Two disclosures moved the other way. Customers passing 10% of revenue are no longer named: FY25 identified Valve at 44.2% and Sony ($6758.T) at 25.2%; H1 2026 gives client I at 37% and client II at 17%. And the licensing line is disclosed only in aggregate, with no partner detail, and management declined to guide H2 when asked directly. Neither is a breach. Concentration itself improved, and the hundred-plus licensing partners are the plausible reason — but the half's most consequential revenue line is now its least specified.

C4: Marketing posture

🟡 Developing → The condition's terms are re-anchored here: the confirmed 2028 window makes the original read points meaningless.

The original test was a gameplay campaign initiated at the right moment, with Summer Game Fest and Gamescom named as the read points. Both passed without a Witcher 4 showing. With a 2028 window company-confirmed, that moment is defined and two years out, so continuing to score those events would manufacture a signal out of a schedule. The condition re-anchors to the 2028 window, with the first meaningful read a campaign start during 2027 — a change to the registry's terms rather than a status movement.

The bridge produced a live demand read in the meantime: Songs of the Past drew 900,000 wishlist additions in the week after Gamescom and took the show's Best Trailer award. The 23m views the Management Board report cites cover both Witcher trailers together, not Songs of the Past alone.

C5: Cyberpunk ecosystem

🟢 Affirmed → The ecosystem stopped being adjacency and became revenue.

Memo No. 2 argued the Cyberpunk ecosystem was deliberate franchise management on the Capcom analogue. H1 settles that, and the re-frame cuts both ways: PLN 94.8m from more than 100 partners is 21.8% of the half, which validates the flywheel and simultaneously puts a material, lumpy and unguided line into a company whose thesis rests on game execution.

Whether the flywheel exists is now answered. Whether it recurs is not. A sub-monitor on licensing durability is proposed for the FY2026 restatement.

W1, W2 and W3 carry no separate prose this period. W1 is affirmed on the new short-term incentive plan rules of 19 August 2026 — PLN 50m threshold, FY2026 the first period — and no disclosed senior departure; W2 on own-project revenue holding to a 1.5% decline across a half with no new release; W3 on the absence of any engine-related qualifier in scope or timing language.

W4: Incentive programme

🟡 Developing → The clearest quantitative indication management has given of the scale it expects from The Witcher 4.

The basis is now unambiguous: cumulative net profit on continuing operations plus the phase's own valuation cost, rather than an adjusted figure. Stage 1, covering 2023 to 2026 against a PLN 2bn target, is 86% complete with roughly PLN 273m left to earn in H2, which management noted is more than the company earned across the first two quarters of the year. Stage 2 is written down.

Stage 2 was judged unlikely by the Management Board and PLN 11.07m of previously recognised cost reversed.

Management stated that they remain optimistic on the third and fourth targets, which is an affirmative statement rather than an inference from silence. Stage 3 runs to PLN 4bn cumulative across 2025 to 2028, PLN 0.8bn of that is earned, and PLN 3.2bn is left to earn across H2 2026, FY2027 and FY2028. Take an H2 2026 of roughly PLN 0.3bn, in line with the Stage 1 remainder management put at PLN 273m, and an FY2027 of roughly PLN 0.5bn, carried by a full expansion rather than a full-price release. That leaves roughly PLN 2.4bn falling in FY2028. FY2020, the Cyberpunk 2077 launch year, produced PLN 1.154bn.

The derivation is shown so a reader can reject it; the H2 and FY2027 assumptions are the places to push. Management is standing behind a target, not forecasting a result.

W5: Hadar ramp

🟡 Developing ↑ Direction improved, guidance softened.

Hadar reached 33 developers at 31 July from 24 at 30 April, reversing the two consecutive declines Memo No. 2 flagged, and remains in concept phase. Management restated the target on the call as up to 50 by the end of this year, softer than the 40 to 50 previously given.

Thesis Standing

The thesis is affirmed. The organisational and financial preconditions have now held for three consecutive reporting cycles, and nothing in this half disturbs them. Two developments change the frame.

The calendar stopped being an assumption. Three Witcher releases now sit in three consecutive years: the Witcher 3 remaster on 29 September 2026, Songs of the Past in 2027, The Witcher 4 in 2028. Every prior memo carried the launch year as inference; it is now stated, and the bridge strategy Memo No. 2 identified has a terminal date.

Earnings quality separated from earnings. The half's profit is real and its composition is new. A licensing line that flatters the interstitial years neither advances the thesis nor damages it — the thesis has always turned on a single flagship release event — but it does make the reported numbers a poorer proxy for progress against that event. The discipline for the next four quarters is to read own-project revenue separately, which is why it is broken out in the table above.

The burden of proof has moved accordingly. What is unresolved is execution, and the first real evidence of it arrives with a Witcher 4 campaign rather than with a results release.

Two corrections to prior memos. Memo No. 1 reported PLN 512m in cash generated from ongoing operations. FY25 operating cash flow was PLN 590.9m, and the PLN 512m figure was development spend, filed at PLN 513.2m. Memo No. 1's headcount table also mis-stated February 2025 Hadar as 14 against a correct 17, and October 2025 Witcher 4 as 477 against a correct 447. Both are corrected against slide 10 of the FY25 presentation and are reflected in the chart above.

What to Watch