Witchfire CEO says Divinity devs are "definitely not evil" for exploring AI, but the backlash stems from people wanting "to have contact with other humans through art"

Adrian Chmielarz, CEO of The Astronauts, defended Larian Studios' exploration of generative AI during early ideation on Divinity, saying the studio is "definitely not evil" and is conducting internal deliberations, while confirming his own studio Witchfire only uses AI for minor placeholder text cleanup and not for game assets. The comments follow consumer backlash focused on ethical and environmental concerns about AI-generated art, highlighting reputational and PR risk for developers but presenting limited direct financial or market impact.

Analysis

Market structure: Generative-AI tooling favors infrastructure and creative-software incumbents (NVIDIA, Adobe, Unity, AWS/MSFT GCP) by lowering marginal cost of asset creation by an estimated 30–70% over 12–36 months, while increasing price competition among mid/small studios. Demand will bifurcate — a mass-market segment chasing cheaper, faster releases and a premium “human-made” cohort that can command 5–30% higher willingness-to-pay for branded/artisanal titles. Cross-asset impact: stronger GPU/semiconductor pricing (supporting NVDA) and small widening of credit spreads for leveraged indie publishers; FX/commodities immaterial outside energy impact from extra datacenter compute demand.

Risk assessment: Tail risks include copyright class actions or restrictive regulation (e.g., EU AI Act language) leading to fines or injunctive relief that could cost a mid-sized studio >$50–100M; probability non-negligible within 12–24 months. Short-term (days/weeks) risks are PR-driven revenue hits (pre-order cancellations 5–20%); medium-term (6–18 months) operational shifts as studios retool pipelines; hidden dependency: platform gatekeepers (Steam/Epic) and community sentiment can instantly reprice title profitability by 10–40%.

Trade implications: Tactical long positions in AI infra/creative SaaS (NVDA, ADBE, U) and short selective small-cap gaming exposure (short ESPO or puts) capture the infrastructure upside and reputational downside respectively; use 3–12 month horizons. Options: buy 3–6 month NVDA call spreads (debit) to capture sustained GPU demand; pair trade = long NVDA (2–3% portfolio) / short ESPO (1%) sized to net delta ~0 and profit from structural divergence. Entry window: act on pullbacks within 2–6 weeks; hold 6–12 months.

Contrarian angles: Markets underprice the “human-made” premium — if 3+ major studios pledge ‘no-AI’ within 90 days, branded IP owners (TTWO, MSFT-owned studios) can re-capture pricing power and outperform smaller studios by 10–30% over 6–18 months. Conversely, if broad adoption accelerates, regulatory backlash could temporarily lift infra vendors then compress margins; historical parallel = photography/CGI: democratization expanded supply but created a durable high-end artisan segment. Watch lawsuits and platform policy shifts as early signals.

Key Decisions for Investors

• Establish a 2–3% portfolio long in NVIDIA (NVDA) within 2 weeks, buy on any ≤5% intraday pullback, target +25% in 6–12 months, stop-loss at -12%; rationale: sustained GPU demand from generative-AI art pipelines.
• Add a 1–2% long position in Adobe (ADBE) over the next 4–8 weeks on pullbacks ≤3%, target +15–20% in 6–12 months as creative-suite AI tools monetize, stop-loss -10%.
• Initiate a 1% short or buy 3-month put spread on ESPO (VanEck Video Gaming & eSports ETF) within 2–6 weeks to capture reputational/earnings downside among small/mid-cap studios; size risk to lose no more than 1% of portfolio.
• Reduce exposure to small-cap/public game developers (examples: Embracer OTC: EMBRF and similar) by 50% over the next 4 weeks and reallocate proceeds into NVDA/ADBE; rationale: higher execution, PR and legal risk for smaller studios.
• Monitor regulatory/industry triggers: if EU AI Act or 3+ AAA studios announce strict 'no-AI-generated content' policies within 90 days, allocate 1% into Take-Two Interactive (TTWO) within 30 days (expect 6–18 month outperformance of branded human-made IP); if the opposite (widespread AI adoption by 3+ majors), add another 1% to NVDA.