Unity Software — Bull & Bear Memo
1. Business in one line
Consolidated revenue grew 2% in FY2025 [10-K], but the strategic business grew 38% in Q2'26 once ironSource and Supersonic were exited [Q2'26 call]. For a turnaround, the question is whether the new engine (Vector) keeps compounding or was a one-time catch-up. That is where the bear case sits.
2. Bull case — Peter Lynch pitch
The simple reason this stock could work is that Unity has turned its engine's reach into a performance-ad business that is growing 60%+ on 83% adjusted gross margins, and the old businesses that were hiding it are now gone.
A1Vector keeps winning share of mobile-game ad spend
- What must happen
- Vector needs to keep delivering better return on ad spend (ROAS) than rival networks, so advertisers move budget to it every quarter. Sequential growth needs to stay in the teens for another 4–6 quarters.
- Why Unity specifically
- Performance-ad spend follows measured returns. Unity ships model updates quickly (20+ Vector updates in Q2) and has added new products such as day-28 ROAS campaigns, which grew almost 3x from Q1 and are used by more than 25% of advertisers. Only $3M of Q2 Vector growth came from former ironSource customers.
- Evidence
- Vector grew 23% quarter over quarter in Q2'26 against a 12–13% plan, after four straight quarters of about 15%. Its annual run-rate is 'substantially over $1B'. Strategic Grow revenue was $329M, up 63% [Q2'26 call]. Q3 guide: Vector +19–21% quarter over quarter, Grow +68–70% year over year.
- What to monitor
- Vector sequential growth, and strategic Grow growth year over year.
A2Runtime data is a signal rivals can't buy
- What must happen
- Player-behaviour data from the Unity runtime needs to make Vector's predictions measurably better, and developers need to keep opting in.
- Why Unity specifically
- About 3 billion people a month play a Made-with-Unity game [Q2'26 call]. No other ad network sees in-game behaviour at the engine level across that many titles. This is what could turn a model race into a moat.
- Evidence
- Runtime signals went into production models only at the end of Q2'26, so this is still early. Opt-in to the developer data framework is above 90% [Q1'26 call].
- What to monitor
- Opt-in rate, any disclosure on runtime's contribution to returns, and Apple/Google policy changes.
A3The engine stays the default through the AI transition and feeds the ad flywheel
- What must happen
- Unity needs to hold its roughly 70% share of mobile game creation [Q1'26 call] as coding agents change how games get built. Pricing needs to move from seats to agent usage and consumption without revenue leaking out.
- Why Unity specifically
- Unity 7 (beta Q4'26, full release Q1'27) opens the engine to coding agents through a free MCP server and open APIs. It also switches on Vector, commerce and live-ops automatically, so new games start inside the Grow funnel. Unity AI pricing now counts agent connections as well as seats.
- Evidence
- Strategic Create revenue was $157M, up 14% excluding a one-time item. Growth came from price increases, minimum commitments and China, where Create grew almost 50% in FY2025 [Q4'25 call]. New Unity sign-ups rose 20% quarter over quarter in Q1'26.
- What to monitor
- Create growth excluding one-time items, Unity 7 adoption, and how much of Create growth comes from price versus volume.
A4Growth turns into GAAP profit and per-share value
- What must happen
- Margins need to keep expanding while stock comp and the share count fall, so free cash flow per share compounds.
- Why Unity specifically
- Incremental margins are high because the ad business sits on existing infrastructure. Exiting ironSource and Supersonic removes low-margin revenue. Management expects Supersonic's exit alone to add at least 200bp of operating margin [Q1'26 call].
- Evidence
- Adjusted EBITDA margin was 29% in Q2'26, up 800bp, with 33% guided for Q3. Q2 free cash flow was $202M, and H1 was $268M [10-Q]. Stock comp fell to 14% of revenue from 21% of FY2025 revenue [10-K]. Management expects its first GAAP profit in Q3'26. The $558M 2026 convertible will be repaid in cash in November.
- What to monitor
- GAAP net income, stock comp as a % of revenue, diluted share count, and recurring restructuring charges.
Vector keeps winning share of mobile-game ad spend
Vector needs to keep delivering better return on ad spend (ROAS) than rival networks, so advertisers move budget to it every quarter. Sequential growth needs to stay in the teens for another 4–6 quarters.
Evidence: Vector grew 23% quarter over quarter in Q2'26 against a 12–13% plan, after four straight quarters of about 15%. Its annual run-rate is 'substantially over $1B'. Strategic Grow revenue was $329M, up 63% [Q2'26 call]. Q3 guide: Vector +19–21% quarter over quarter, Grow +68–70% year over year.
Vector's lead is a model lead, and advertisers can leave the next day
Advertisers put budget wherever ROAS is best each day, with no contract and no integration lock-in. The 10-K names AppLovin and Moloco as competitors, and both run the same kind of self-learning bidding models. Some of Vector's 80% year-over-year growth (Q1'26) was recovery from a weak base, and that part won't repeat.
Confirms failure: Vector sequential growth below about 10% for two quarters, Q3/Q4 guidance cut, or management moving attention from Vector growth to 'platform' metrics.
Permanent if a rival's model lead widens, because ad-network scale reinforces itself in both directions.
Runtime data is a signal rivals can't buy
Player-behaviour data from the Unity runtime needs to make Vector's predictions measurably better, and developers need to keep opting in.
Evidence: Runtime signals went into production models only at the end of Q2'26, so this is still early. Opt-in to the developer data framework is above 90% [Q1'26 call].
The runtime 'moat' depends on permission Unity doesn't control
The data belongs to developers' players and reaches Vector only through developer opt-in, OS policy and privacy law. The 10-K flags Apple's tracking opt-out, planned Android controls, child-protection rules and a new DOJ restriction on sensitive-data flows. Big studios may also object to their players' behaviour training a network that sells installs to their competitors.
Confirms failure: Opt-in falling below 90%, an Apple/Google SDK or data-policy change, or regulatory action on in-game data used for advertising.
Permanent. Platform policy changes rarely reverse.
The engine stays the default through the AI transition and feeds the ad flywheel
Unity needs to hold its roughly 70% share of mobile game creation [Q1'26 call] as coding agents change how games get built. Pricing needs to move from seats to agent usage and consumption without revenue leaking out.
Evidence: Strategic Create revenue was $157M, up 14% excluding a one-time item. Growth came from price increases, minimum commitments and China, where Create grew almost 50% in FY2025 [Q4'25 call]. New Unity sign-ups rose 20% quarter over quarter in Q1'26.
AI weakens the seat-based engine business and shrinks the top of the funnel
If coding agents and world models do the work that seats used to do, fewer paid seats are needed for each game. Unity is opening its APIs and MCP server for free. Recent Create growth has come from price increases and minimum commitments, which have limits. Q3 Create is guided to +7–10%, down from 14%.
Confirms failure: Create growth excluding one-time items falling to low single digits, agent-based pricing failing to show up in revenue, or rival engines or AI-native tools taking mobile share.
Partly fixable if consumption pricing works. Permanent if the engine becomes a commodity layer.
Growth turns into GAAP profit and per-share value
Margins need to keep expanding while stock comp and the share count fall, so free cash flow per share compounds.
Evidence: Adjusted EBITDA margin was 29% in Q2'26, up 800bp, with 33% guided for Q3. Q2 free cash flow was $202M, and H1 was $268M [10-Q]. Stock comp fell to 14% of revenue from 21% of FY2025 revenue [10-K]. Management expects its first GAAP profit in Q3'26. The $558M 2026 convertible will be repaid in cash in November.
Headline profit flatters per-share economics, and the capital-allocation record is poor
Adjusted EBITDA excludes stock comp ($385M in FY2025) and restructuring charges that recur every year: $267M in 2024, $47M in 2025, $38M in H1'26. ironSource cost $2.9B in 2022 [FY22 10-K]. Its ad network has been shut down with a $279M impairment, and $3.16B of goodwill remains. The useful lives of the Wētā intangibles were cut to 1–3 years. In February management said ironSource would 'get smaller and smaller over time', then announced its shutdown six weeks later.
Confirms failure: Restructuring charges continuing in FY2027, stock comp moving back above 15% of revenue, or another impairment.
Fixable. Stock comp is already falling. But each new write-down weakens the case that management allocates capital well.
Why the market might be missing it: the consolidated numbers hid Vector until this year. FY2025 revenue grew 2% with a $401M net loss, and H1'26 GAAP results included a $279M impairment. Q3'26 is the first quarter with almost no non-strategic revenue, and the first where the reported numbers show what the business actually earns.
3. Bear case — Munger invert
The most likely way I lose money is that Vector turns out to be a catch-up spike in a model race with no switching costs. Its growth reverts toward market growth, and a stock priced for compounding is left with a mid-growth engine business and a history of write-downs.
Ordered by how permanent and value-destructive each risk is. Each card names the bull assumption it attacks.
B1Vector's lead is a model lead, and advertisers can leave the next dayattacks A1
- How it could fail
- Advertisers put budget wherever ROAS is best each day, with no contract and no integration lock-in. The 10-K names AppLovin and Moloco as competitors, and both run the same kind of self-learning bidding models. Some of Vector's 80% year-over-year growth (Q1'26) was recovery from a weak base, and that part won't repeat.
- What would confirm failure
- Vector sequential growth below about 10% for two quarters, Q3/Q4 guidance cut, or management moving attention from Vector growth to 'platform' metrics.
- Damage to economics / value
- Grow was about 71% of Q2 revenue and most of the incremental margin. If growth slows from 60% to around 15%, the operating leverage behind the 33% margin guide goes away while R&D and AI inference spend stay fixed.
- Permanent or fixable
- Permanent if a rival's model lead widens, because ad-network scale reinforces itself in both directions.
B2The runtime 'moat' depends on permission Unity doesn't controlattacks A2
- How it could fail
- The data belongs to developers' players and reaches Vector only through developer opt-in, OS policy and privacy law. The 10-K flags Apple's tracking opt-out, planned Android controls, child-protection rules and a new DOJ restriction on sensitive-data flows. Big studios may also object to their players' behaviour training a network that sells installs to their competitors.
- What would confirm failure
- Opt-in falling below 90%, an Apple/Google SDK or data-policy change, or regulatory action on in-game data used for advertising.
- Damage to economics / value
- Without runtime data, A2 disappears and A1 becomes a pure model race (B1). Unity would lose the one input it has that rivals can't copy.
- Permanent or fixable
- Permanent. Platform policy changes rarely reverse.
B3Headline profit flatters per-share economics, and the capital-allocation record is poorattacks A4
- How it could fail
- Adjusted EBITDA excludes stock comp ($385M in FY2025) and restructuring charges that recur every year: $267M in 2024, $47M in 2025, $38M in H1'26. ironSource cost $2.9B in 2022 [FY22 10-K]. Its ad network has been shut down with a $279M impairment, and $3.16B of goodwill remains. The useful lives of the Wētā intangibles were cut to 1–3 years. In February management said ironSource would 'get smaller and smaller over time', then announced its shutdown six weeks later.
- What would confirm failure
- Restructuring charges continuing in FY2027, stock comp moving back above 15% of revenue, or another impairment.
- Damage to economics / value
- Non-GAAP diluted shares are 489.5M against 437.9M GAAP, a 12% gap that includes 41M convertible shares. The $1.0B 2027 notes convert at $48.89. A new CEO award of 880K PSUs vests at $50/$60/$75. Say-on-pay got only about 72% support.
- Permanent or fixable
- Fixable. Stock comp is already falling. But each new write-down weakens the case that management allocates capital well.
B4AI weakens the seat-based engine business and shrinks the top of the funnelattacks A3
- How it could fail
- If coding agents and world models do the work that seats used to do, fewer paid seats are needed for each game. Unity is opening its APIs and MCP server for free. Recent Create growth has come from price increases and minimum commitments, which have limits. Q3 Create is guided to +7–10%, down from 14%.
- What would confirm failure
- Create growth excluding one-time items falling to low single digits, agent-based pricing failing to show up in revenue, or rival engines or AI-native tools taking mobile share.
- Damage to economics / value
- Create is the smaller and slower segment, so this hurts the flywheel more than the P&L. If fewer new games start in Unity, Vector gets fewer auto-configured titles and a thinner runtime data set.
- Permanent or fixable
- Partly fixable if consumption pricing works. Permanent if the engine becomes a commodity layer.
4. Signals to monitor
Each card is tagged with the assumption (A) or risk (B) it tracks.
Vector sequential growth
Strategic Grow revenue growth
Developer data-framework opt-in
Strategic Create growth (ex one-time)
Adj. EBITDA margin / GAAP net income
SBC % revenue; non-GAAP diluted shares
Restructuring & impairment charges
Greater China share of revenue
6. Bottom line
- Unity could work because Vector has turned the engine's reach into a 60%+ growth ad business on 83% adjusted gross margins, and exiting ironSource and Supersonic now lets that growth show up in reported profit.
- Vector needs to keep beating rival bidding models for several more quarters, runtime data needs to become a lasting edge that developers keep opting into, and Unity 7 needs to keep the engine as the default entry point as AI changes how games are built.
- The most important way the thesis breaks is that Vector's lead turns out to be a temporary model advantage in a market with no switching costs, which would bring growth down toward market rates while a record of write-downs and dilution remains.
- I would change my mind if Vector's sequential growth fell below about 10% for two straight quarters, or if developer opt-in or platform policy took away the runtime data advantage.
Section 5 (External challenge) not run. It is added only on request.
Sources
Annual & quarterly filings 3
- FY2025 10-K (filed 2026-02-11): segment revenue, FY2025 P&L, SBC, restructuring, competitors, risk factors
- Q2 FY2026 10-Q (filed 2026-08-06): Q2/H1 P&L, impairments, geography, convertible notes, AppsFlyer commitment, FCF
- FY2022 10-K: ironSource purchase consideration ($2,915.6M)
Current reports (8-K) and proxy votes 5
- 2026-02-10: board changes (Helgason, Bar-Zeev resign)
- 2026-03-26: planned exit of non-strategic businesses
- 2026-05-14: AGM results (say-on-pay ~72% for)
- 2026-07-16 / 07-27: director resignation; new principal accounting officer
- 2026-08-19: CEO special PPSU award (880K; $50/$60/$75 hurdles)
Earnings-call transcripts (third-party transcription) 4
- Q2 FY2026 (2026-08-06): Vector +23% QoQ, strategic Grow $329M +63%, Q3 guide, runtime data
- Q1 FY2026 (2026-05-07): Vector +80% YoY, >90% opt-in, ~70% mobile share, Supersonic margin uplift
- Q4 FY2025 (2026-02-11): ironSource 11% of Grow, China Create ~+50%
- Q3 FY2025 (2025-11-05): context only