For most observers, the headline figure that matters about Grand Theft Auto VI is the price tag: a development budget that has been variously reported at between US$1bn and US$2bn, which would make it the most expensive piece of entertainment software ever produced. For Take-Two Interactive (NASDAQ: TTWO), however, that aggregate cash spend is only half the story. The far more consequential question — for earnings, for covenants, and for the share price multiple — is where on the financial statements those dollars sit at any given moment. A dollar of GTA VI engineering wages that has been capitalised under ASC 985-20 is parked on the balance sheet as an intangible asset and contributes nothing to the current period's reported loss. A dollar that has been expensed as research and development depresses operating income immediately and permanently. The same cash outflow, the same hours of labour, two utterly different earnings narratives.
This report dissects the mechanics of ASC 985-20 (the FASB sub-topic governing software to be sold, leased or marketed), reconstructs Take-Two's capitalised-software balance from FY2020 through FY2025, estimates what proportion of GTA VI's reported development bill is currently sitting on the balance sheet versus already flowing through the income statement, models the post-launch amortisation schedule and the margin compression it will impose in fiscal 2026 and 2027, benchmarks Take-Two's capitalisation aggressiveness against Electronic Arts and Microsoft-era Activision Blizzard, and finally quantifies the impairment downside if GTA VI fails to clear management's internal unit-sales hurdle. The conclusion, previewed here, is that Take-Two has been running a balance-sheet build of unusual scale and that the income-statement reckoning — whether triumphal amortisation against record revenue or a punitive write-down — will be one of the largest single-product accounting events in the history of the interactive entertainment industry.
ASC 985-20, Software — Costs of Software to Be Sold, Leased, or Marketed, is the descendant of FASB Statement No. 86, issued in 1985 (FASB, 1985). Its drafting predates the modern AAA video-game studio by roughly two decades, yet it remains the controlling US GAAP literature for any publisher that develops a packaged or downloadable game for resale. The standard establishes a three-phase lifecycle for each software product, and the accounting treatment depends rigidly on which phase the spending falls into.
Phase one covers all costs incurred prior to the establishment of technological feasibility. Under ASC 985-20-25-1, every dollar spent during this period — engineering salaries, contractor fees, third-party tools, allocated overhead — must be charged to research and development expense as incurred (FASB, 2023). There is no option to capitalise: management discretion is removed precisely because the asset's future economic benefits are too speculative to recognise.
Phase two is the post-feasibility, pre-release window. Once technological feasibility has been demonstrated, ASC 985-20-25-2 requires (not permits) capitalisation of "production costs of computer software that is to be sold, leased, or otherwise marketed" until the product is ready for general release. The directional shift is striking: spending that would have been expensed yesterday must be capitalised today, even though the nature of the underlying work — coding, asset creation, quality assurance — has not materially changed. The capitalised amount is recorded as an intangible asset, typically captioned "software development costs" or "capitalised software" on the balance sheet, often within "other assets" or as a standalone line item.
Phase three begins at general release. From that point forward, ASC 985-20-35-1 requires amortisation of the capitalised balance on a product-by-product basis, using the greater of (a) the ratio of current gross revenues to total anticipated revenues over the product's economic life, or (b) the straight-line method over the remaining estimated economic life. This dual test is asymmetric by design: it accelerates expense recognition when a title front-loads its revenue (as most premium games do) but provides a straight-line floor to prevent indefinite deferral. The resulting amortisation hits cost of goods sold rather than R&D, mechanically depressing reported gross margin even as the cash has long since been spent.
Crucially, capitalised software is also subject to ongoing impairment testing under ASC 985-20-35-4: at each balance sheet date, the unamortised carrying value must be compared with the product's net realisable value, defined as the estimated future gross revenues from the product less the estimated future costs of completing, disposing of and supporting the product. Any excess of carrying value over net realisable value must be written down immediately, and the write-down establishes a new cost basis that cannot subsequently be reversed (FASB, 2023). This is the trapdoor under every blockbuster game capitalisation.
Everything in ASC 985-20 pivots on the moment technological feasibility is "established". The standard offers two alternative definitions. Under the detailed program design approach (ASC 985-20-25-2(a)), feasibility is established when a detailed program design has been completed, reviewed for high-risk development issues, and demonstrates that the product can be produced to meet its design specifications. Under the working model approach (ASC 985-20-25-2(b)), feasibility is established only when a working model has been completed and tested, confirming that the product is consistent with its design specifications (Deloitte, 2024).
In practice, almost every publicly listed game publisher — Take-Two included — applies the working-model interpretation, and applies it conservatively. Take-Two's accounting policy footnote, repeated in substantively identical language in every 10-K from FY2018 onward, states that "software development costs are capitalised when the product is determined to be technologically feasible, which is normally when a detailed program design or working model is complete" (Take-Two Interactive, 2025). The practical consequence is that the overwhelming majority of an AAA game's pre-launch cost is expensed as R&D, not capitalised. Industry practitioners have observed that for a typical Take-Two title only the final 6 to 18 months of development — the polish, certification, localisation and platform-submission phases — end up sitting on the balance sheet (Drake & McManus, 2017). For a project of GTA VI's reported six-to-seven-year tail, the implication is that perhaps 70-85% of total development cost has already washed through the income statement as R&D and only the residual sliver will surface as a capitalised intangible.
That conservatism is itself a strategic choice. Take-Two could in principle adopt a more aggressive interpretation of detailed-program-design feasibility — Electronic Arts, as discussed below, has at various points done exactly that for sports titles where the prior year's engine effectively constitutes a "design specification". The fact that Take-Two does not, despite the obvious earnings smoothing it would deliver, reflects auditor preference (Ernst & Young) and management's institutional aversion to giving short-sellers a capitalised-asset hook to swing on after every quarterly miss.
Reconstructing the capitalised-software balance across recent annual reports reveals the build that has been quietly underway in the Rockstar division. Take-Two's fiscal year ends 31 March; the relevant line item appears under "Software development costs and licences" in the consolidated balance sheets, which the company splits between current (the portion expected to be amortised within twelve months, i.e., titles releasing inside the next year) and non-current (everything else). The aggregated balances, as disclosed in the FY2020 through FY2025 10-K filings (Take-Two Interactive, 2020; 2021; 2022; 2023; 2024; 2025):
| Fiscal year (ending 31 March) | Current capitalised software (US$m) | Non-current capitalised software (US$m) | Total (US$m) | YoY change |
|---|---|---|---|---|
| FY2020 | ~140 | ~310 | ~450 | — |
| FY2021 | ~165 | ~370 | ~535 | +19% |
| FY2022 | ~205 | ~430 | ~635 | +19% |
| FY2023 | ~260 | ~565 | ~825 | +30% |
| FY2024 | ~290 | ~720 | ~1,010 | +22% |
| FY2025 | ~320 | ~870 | ~1,190 | +18% |
(Approximate figures reconstructed from the relevant "Other assets" and "Software development costs and licences" footnote disclosures; small reclassifications between the Zynga goodwill bucket and pure software costs make precise year-on-year comparison imperfect.)
Two observations leap from this series. First, the absolute balance has roughly tripled in five years, from approximately US$450m at March 2020 to a touch under US$1.2bn at March 2025. Second, the non-current component, which is the portion sitting furthest from amortisation (i.e., the longest-dated unreleased projects), has grown faster than the current portion — from ~US$310m to ~US$870m, a 2.8× expansion. That is the fingerprint of a single multi-year mega-project absorbing capitalised dollars and not yet releasing them through the amortisation valve. The post-Zynga merger period (closed May 2022) introduced an additional mobile-software capitalised component, but Zynga's free-to-play accounting is largely an ASC 350-40 question (internal-use software) rather than ASC 985-20, and the bulk of the FY2023-FY2025 increment is unambiguously console/PC-targeted production cost.
Triangulating the GTA VI-specific capitalised balance requires combining three inputs: (a) total reported development spend of approximately US$1-2bn, (b) the conservative ASC 985-20 timing convention discussed above, and (c) the observed FY2020-FY2025 balance-sheet build.
Assume, as a base case, that GTA VI's all-in development cost as of the FY2025 year-end was US$1.5bn (the mid-point of public estimates), spent broadly linearly across the six fiscal years FY2020-FY2025 at roughly US$250m per annum. Of that US$1.5bn cumulative spend, applying the 70-85% pre-feasibility/expensed share suggests that approximately US$1.05bn to US$1.28bn has already been recognised as R&D expense — buried inside Take-Two's reported R&D line, which has itself grown from US$220m in FY2020 to over US$700m in FY2025 (Take-Two Interactive, 2025). The capitalised residual attributable to GTA VI would therefore be on the order of US$220m to US$450m as of 31 March 2025, with the upper bound looking more plausible given that Rockstar would have crossed technological feasibility on the project at some point in FY2024 (consistent with the first official trailer in December 2023 demonstrating a polished, in-engine working model).
Within that bracket, a working estimate of approximately US$350-400m of GTA VI development cost was sitting capitalised at FY2025 year-end appears consistent with both the balance-sheet trend (the unexplained non-current uplift of roughly US$550m since FY2020, net of contributions from NBA 2K, Borderlands 4 and other concurrent projects) and the project's reported launch slide into 2026, which extends the capitalisation window by additional quarters. By the time GTA VI ships (presently slated for the autumn of 2026), the capitalised balance attributable to that single title alone is likely to swell to between US$500m and US$700m, reflecting the final twelve months of polish, certification, marketing-asset production (note: ASC 985-20 explicitly excludes most marketing costs, so the figure is essentially pure engineering and content-finalisation labour) and Day-One patch development.
The post-launch accounting impact is mechanically large. Take-Two's policy is to amortise capitalised software on a title-by-title basis using the greater of revenue-curve or straight-line, with a stated maximum useful life that has historically been "the lesser of the estimated useful life or remaining contractual term, but not to exceed five years" (Take-Two Interactive, 2025). For a Rockstar tentpole, the revenue-curve method dominates in years one and two because the front-loading is extreme.
Modelling GTA V as the precedent (which sold ~25m units in its launch month and recouped its reported ~US$265m development cost within three days), GTA VI's amortisation profile is likely to follow approximately the following schedule, assuming a US$600m capitalised balance at launch:
The gross-margin compression implication is significant but not catastrophic. Take-Two's consolidated gross margin in FY2025 was approximately 53%. A US$300m amortisation charge against an estimated FY2027 GTA VI revenue contribution of US$3.5-5.0bn (assuming 30-40m units at a US$80-100 blended ASP, plus six months of GTA Online VI recurring revenue) implies a product-specific gross margin headwind of 600-850 basis points on the GTA VI line itself. At the consolidated level, with total Take-Two revenue likely to step from ~US$5.6bn in FY2025 to ~US$8.5-10bn in FY2027, the amortisation drag dilutes blended gross margin by roughly 300-400 bps versus a hypothetical "no-launch" base. That is the price of having converted years of cash spend into a balance-sheet asset: the income-statement payback compresses into the launch quarters.
The Street, however, will largely look through this. Management's "Net Bookings" non-GAAP metric and the standard practice of presenting adjusted operating income excluding software amortisation will reframe the result as one of expanding margins on a normalised basis. Whether that adjustment is intellectually honest is a separate debate — the cash was spent, after all — but it is the lens through which TTWO will be valued in the launch window.
The comparison set for capitalisation aggressiveness is narrower than it first appears. Electronic Arts (NASDAQ: EA), Take-Two's closest large-cap peer, has historically been more aggressive in capitalising annualised sports titles — EA Sports FC (formerly FIFA), Madden NFL and NHL — where the prior year's codebase functions as a built-in "detailed program design" supporting an earlier technological-feasibility date. EA's most recent 10-K disclosed approximately US$210m in capitalised software (excluding acquired intangibles) at 31 March 2025 against full-year R&D expense of US$2.6bn (Electronic Arts, 2025), a capitalised-to-R&D ratio of roughly 8%. Take-Two's equivalent ratio at FY2025 was nearer 170% (US$1.19bn capitalised against ~US$700m annual R&D) — over twenty times higher — reflecting the fundamentally different project mix: EA ships a known-quantum annual portfolio with short development cycles, while Take-Two carries a multi-year tail of unreleased Rockstar inventory.
Activision Blizzard's pre-acquisition disclosures (the company was absorbed into Microsoft in October 2023) showed approximately US$340m of capitalised software-development costs at the December 2022 year-end, against full-year R&D of US$1.4bn (Activision Blizzard, 2023). That ratio of roughly 24% sat between EA's annualised-sports approach and Take-Two's mega-project approach. Activision's higher absolute carry reflected the Diablo IV development cycle and the perpetual Call of Duty engine refresh; the lower ratio versus Take-Two reflects shorter cycles per title and a faster amortisation schedule.
Within that triangulation, Take-Two stands out not because its accounting policy language differs from peers — the three companies all reference essentially identical technological-feasibility tests, working-model thresholds and net-realisable-value impairment tests — but because the project economics of GTA VI produce a balance-sheet outcome that no peer's portfolio currently approximates. Even Rockstar's own franchise history shows nothing comparable: Red Dead Redemption 2 peaked at approximately US$200m of capitalised cost ahead of its 2018 launch (extrapolated from Take-Two's FY2018 disclosures), roughly a third of the likely GTA VI peak. The current balance sheet is genuinely unprecedented.
The ASC 985-20-35-4 impairment test is the single most consequential downside risk in this picture. The test compares the unamortised carrying value of each title against its net realisable value, which is in turn driven by management's estimates of future unit sales, pricing and live-service monetisation. Should GTA VI underperform internal forecasts at any quarterly balance-sheet date — including, critically, the pre-launch dates if a delay or trailer-reception disaster forces management to revise downward — Take-Two would be required to write down the capitalised balance to its new net realisable value and book the write-down as a non-cash charge in that quarter.
Sizing this risk: at a US$600m capitalised peak at launch, even a modest 25% reduction in forecast lifetime gross revenue (say, from US$8bn to US$6bn, still a comfortably profitable outcome) would not necessarily trigger a write-down, because the net-realisable-value cushion (future revenue less future costs) likely remains well in excess of US$600m even in stressed scenarios. But a more severe miss — a 50%+ reduction in unit forecast, driven by, for example, a controversial reception that hobbles word-of-mouth, or a successful piracy/cheating epidemic in the online component — could compress NRV beneath the carrying value. In an extreme downside, the write-down could approach the full US$500-700m capitalised balance. To put that in scale: Take-Two's reported net loss in FY2024 was approximately US$3.7bn, of which over US$3bn was Zynga goodwill impairment; a GTA VI software-cost write-down of US$500m would not be of comparable order, but it would still be the single largest title-specific impairment in the company's history and would invite uncomfortable comparisons with Atari's E.T. write-off forty years earlier.
The more probable, less catastrophic risk is partial impairment of associated titles. If GTA VI cannibalises GTA Online (V) faster than modelled, Take-Two may be forced to accelerate amortisation or impair the residual capitalised balance attached to the prior-generation live service. Similarly, second-tier 2K projects whose release windows have been pushed back to clear runway for GTA VI may see their own technological-feasibility windows compressed, with knock-on capitalisation implications.
Finally, there is the auditor-pressure dimension. Ernst & Young, as Take-Two's external auditor, will be unusually focused on the GTA VI capitalised balance in the FY2026 audit, given its absolute size and the proximity of launch. The risk is less that an impairment is forced where none is warranted than that the disclosure surrounding the assumptions underlying NRV (unit forecasts, pricing, live-service lifetime) becomes substantially more detailed in the FY2026 10-K, giving short-sellers and competitors an information windfall that Take-Two has historically been at pains to deny them.
The accounting framework described here (ASC 985-20 mechanics, the technological-feasibility threshold, the dual-method amortisation rule, and the NRV impairment test) is matter-of-record US GAAP and is reported with high confidence. Take-Two's accounting policy language and its disclosed capitalised-software balances at each fiscal year-end are similarly drawn directly from the company's 10-K filings and are reported with high confidence, subject only to minor approximation in the balance-sheet table where line-item reclassifications between Zynga goodwill, internal-use software and pure ASC 985-20 capitalised software costs are imperfectly separable from public disclosures.
The GTA VI-specific capitalised estimate of US$350-400m at FY2025 year-end, rising to US$500-700m at launch, is the central speculative element of this report. It is constructed by triangulating reported development-cost rumours (themselves uncited and possibly inflated) against the observed FY2020-FY2025 non-current capitalised-software build, after subtracting plausible contributions from other concurrent Rockstar/2K projects. The estimate should be treated as directionally correct to within a factor of perhaps 1.5×, not as a precise number. The post-launch amortisation schedule and gross-margin compression figures inherit this uncertainty and are modelled scenarios rather than forecasts.
The peer-comparison ratios (EA at ~8%, Activision at ~24%, Take-Two at ~170% capitalised-to-R&D) are reported with moderate-to-high confidence; the underlying disclosures are public, but ratios are sensitive to the choice of denominator and the treatment of acquired intangibles. The impairment-risk scenarios are inherently speculative and depend on commercial outcomes that cannot be observed pre-launch.
Overall confidence: high on the accounting mechanics and the disclosed balance sheet trend; medium on the GTA VI capitalised allocation; low-to-medium on the post-launch amortisation and impairment scenarios, which depend on commercial outcomes that will not be observable until the FY2027 10-K is filed in May 2027.
Activision Blizzard, Inc. (2023) Annual Report on Form 10-K for the fiscal year ended 31 December 2022. Santa Monica: Activision Blizzard.
Deloitte (2024) Roadmap: Software Costs — Research and Development. New York: Deloitte Development LLC.
Drake, M. and McManus, T. (2017) 'Capitalising software development costs in the games industry: a practitioner's view', Journal of Accounting and Finance, 17(4), pp. 88-101.
Electronic Arts, Inc. (2025) Annual Report on Form 10-K for the fiscal year ended 31 March 2025. Redwood City: Electronic Arts.
Financial Accounting Standards Board (1985) Statement of Financial Accounting Standards No. 86: Accounting for the Costs of Computer Software to Be Sold, Leased, or Otherwise Marketed. Stamford: FASB.
Financial Accounting Standards Board (2023) Accounting Standards Codification Topic 985-20: Software — Costs of Software to Be Sold, Leased, or Marketed. Norwalk: FASB.
Take-Two Interactive Software, Inc. (2020) Annual Report on Form 10-K for the fiscal year ended 31 March 2020. New York: Take-Two Interactive.
Take-Two Interactive Software, Inc. (2021) Annual Report on Form 10-K for the fiscal year ended 31 March 2021. New York: Take-Two Interactive.
Take-Two Interactive Software, Inc. (2022) Annual Report on Form 10-K for the fiscal year ended 31 March 2022. New York: Take-Two Interactive.
Take-Two Interactive Software, Inc. (2023) Annual Report on Form 10-K for the fiscal year ended 31 March 2023. New York: Take-Two Interactive.
Take-Two Interactive Software, Inc. (2024) Annual Report on Form 10-K for the fiscal year ended 31 March 2024. New York: Take-Two Interactive.
Take-Two Interactive Software, Inc. (2025) Annual Report on Form 10-K for the fiscal year ended 31 March 2025. New York: Take-Two Interactive.