Amazon's AI Game Mandate Backfired: Project Trident Team Laid Off Anyway

In October 2025, developers at Amazon Game Studios received the same notification that has become all too familiar across the industry: their jobs were gone. What makes this particular case stand out is not the layoff itself, but the circumstances leading up to it. A team working on an internal project called Project Trident was pressured to incorporate generative AI into their game, scrambled to make it work, and then got laid off anyway.

According to multiple sources who spoke to Eurogamer, the project went through at least three distinct pivots before its cancellation. The original concept was a Shadow of the Colossus-style action game where players would topple Jotuns using flying mounts and grappling hooks. Sources described it as something "everybody was excited by" internally.

Then mid-2024, an "AI mandate" was introduced. The team was told to either integrate generative AI into their game or face shutdown. This forced a complete redesign. The project shifted to a Helldivers-esque multiplayer experience with roguelite mechanics, where players would drop in, drop out, and use AI for mission generation and story elements. The timeline was compressed to 18 months for a ship-ready product (basically nothing in game development time).

Halfway through that compressed schedule, the deadline was lifted, and the team shuffled to a third concept: a single-player third-person action game set in a comedy-focused Nordic world. The protagonist works for a fictional parody company called Valhalla Ventures. The key differentiator was the use of large language models to enable dynamic communication between players and NPCs.

Gameplay footage reviewed by Eurogamer showed players commanding an NPC named Thor through verbal or text commands. The LLM would recognize the command and trigger special attacks. Environmental puzzles required players to direct Thor to take specific actions, with the AI-powered NPC reacting accordingly. In another sequence, players could attempt to convince captured enemies to join Valhalla Ventures through custom dialogue, with the enemy's personality powered by an LLM determining whether they agreed or rejected the offer.

The art, music, story, and core gameplay remained hand-crafted by developers. Generative AI only powered player-to-NPC interaction and improved animation quality through dynamically generated lip-synching. The team was close to having a demonstrable demo ready for the first half of 2026 when the layoffs hit.

This wasn't an isolated incident. Amazon laid off approximately 14,000 employees in late 2025, with significant reductions across its gaming division. New World, one of Amazon Games' few success stories, was dealt a fatal blow with its team's deconstruction coinciding with an end-of-life announcement following its tenth season. The Lord of the Rings MMO project also appears to have been cancelled, though Amazon claims it is still exploring something with the IP.

In an internal memo at the time, Steven Boom, head of Amazon Games, Twitch, and audio, wrote that the company would halt "a significant amount of our first-party AAA game development work — specifically around MMOs — within Amazon Game Studios." This strategic pivot away from big-budget internal development affected teams across the company, from New World to Lost Ark to Throne and Liberty.

When contacted for comment, Jeff Gattis, head of Amazon Games, stated: "AI was not the reason behind role reductions in Games. Those changes were the result of a strategic shift in our business and a refocus on the areas where Amazon can deliver the most value to players. Great games are made by talented people and we think AI should expand what's possible."

The disconnect between corporate messaging and developer experience is stark. One source told Eurogamer: "I think we did discover the best ways and the worst ways that [generative AI implementation] can happen," and yet, "[Amazon] laid off everyone that was an expert in the best and worst ways to implement AI in regards to game development." (a problem that has plagued users for years, frankly).

Independent reporting from PC Gamer corroborates the timeline and scope of the changes, noting the team was given two years to make the AI-integrated version work before the October 2025 layoffs.

The physical reality of this development cycle is worth noting. Developers spent months iterating on voice command systems, testing how players would interact with NPCs through microphones or text boxes. They built systems where Thor's personality would respond differently based on player dialogue choices. They created puzzle mechanics that required the LLM to parse player intent and trigger appropriate in-game responses. All of this work, all of this experimentation, was discarded when the strategic pivot came down.

What does this tell us about the current state of AI in game development? The technology itself wasn't the problem. Project Trident showed that generative AI could create genuinely novel gameplay mechanics when integrated thoughtfully. The issue was corporate strategy that demanded AI integration without providing adequate time, resources, or long-term commitment to see projects through.

Other studios are watching closely. The pattern here is clear: mandate AI adoption, compress timelines, then cut the team when the project doesn't meet unrealistic expectations. It's like trying to parallel park a freight train while someone keeps changing the parking spot dimensions.

Whether the industry learns anything from this remains to be seen. Amazon's statement about using AI "thoughtfully and responsibly" rings hollow when the team that actually figured out how to do that was let go. The real question isn't whether AI belongs in games. It's whether companies will give developers the time and stability needed to make it work properly.

For now, the Project Trident team is scattered across the industry, carrying knowledge about what works and what doesn't with generative AI in games. Whether that expertise finds a home elsewhere, or whether the next corporate mandate will be just as arbitrary, is something only time will tell. Whether users actually pay for AI-integrated games remains the real question.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn