The short answer: Assassin's Creed Shadows runs on an upgraded Ubisoft Anvil, not Unreal Engine 5 or Snowdrop. Ubisoft lists Anvil as the engine for Shadows, and describes Anvil as the in-house tech crafted for the first Assassin's Creed and scaled for massive open worlds and even competitive multiplayer like Rainbow Six Siege on its technology page.
That choice looks conservative until you see what it had to carry. Shadows is the fourteenth main Assassin's Creed, developed by Ubisoft Quebec as the successor to Mirage, set in 16th-century Japan at the end of the Sengoku period with two very different leads — shinobi Naoe and samurai Yasuke. It launched March 20, 2025 for macOS, PS5, Windows and Xbox Series X/S, with a Switch 2 version December 2, 2025 and iPadOS to follow. It passed 1 million players on release day, 3 million in a week, and 5 million by July 2025, became the best-selling game of March 2025 in the U.S., posted the series' second-highest day-one revenue behind Valhalla, and outperformed Odyssey in consumer spending — all while costing over €100 million. Reviews were generally favorable, with 82% of critics recommending it on OpenCritic and 36/40 from Famitsu. The question is why Ubisoft trusted a 2007 house engine with its Japan rescue mission.
What Anvil actually is
Anvil started as Scimitar, written from the ground up in 2007 for the first Assassin's Creed by Ubisoft Montreal, using Autodesk HumanIK for hands and feet on climbs. It became AnvilNext for Assassin's Creed III in 2012 with global illumination, a new weather system, a rewritten renderer for up to 2,000 NPCs in real time, and no loading screens going from land to sea carried over from Far Cry 3 tech. AnvilNext 2.0 debuted with Unity in 2014, rebuilt to generate city blocks procedurally while still rendering Notre Dame near 1:1, then refined through Syndicate, Origins and Odyssey and reused for Rainbow Six Siege, For Honor, Ghost Recon Wildlands and Breakpoint. With Valhalla in 2020 it was rebranded Ubisoft Anvil, used for Immortals Fenyx Rising and Mirage, and then upgraded again for Shadows with enhanced lighting, breakable props and a seasonal system. The full lineage is documented on the Anvil page.
Ubisoft's own summary is blunter: Anvil is modular and scalable, focused on systemic gameplay and GPU-driven rendering, from massive living open worlds to Siege. That is not marketing poetry. It means one engine team feeds Quebec, Montreal, Bordeaux, Singapore and a dozen support studios without retraining everyone for Unreal.
What Shadows demanded from that old engine
Shadows is the first Assassin's Creed built only for ninth-generation consoles, dropping PS4 and Xbox One. Lead producer Karl Onnée said the newer hardware was necessary to support the new engine version that supplies light, shadows and dynamic weather. In practice that meant four linked upgrades, all described in hands-on reports:
1. Light as stealth. A new global illumination system creates dynamic light and shadows that affect enemy vision. You hide in shadows and create darkness by destroying lanterns or killing torch carriers, with a light meter showing concealment. Servants carry lanterns at night and can raise alarms. Naoe's Eagle Vision returns, plus a new Observe mode. IGN's 50 details breakdown lists the light meter, lantern destruction and servant system, and IGN's studio tour explains the Splinter Cell-like dynamic lighting intent.
2. Seasons as level design. The world progresses through spring, summer, autumn and winter dictated by campaign progress for historical authenticity. Spring and summer grow bushes and tall grass to hide in; autumn and winter kill them. Ponds you can hide in with a bamboo snorkel freeze over. Icicles form and can fall to reveal you. Summer samurai cut bushes to find you; winter guards hug fires, opening colder paths. Storms, fog, wind and snow mask sight and sound. The same IGN breakdown details frozen ponds, icicles and patrol behavior.
3. Everything breaks. Pretty much every prop can be dynamically damaged — katanas slice bushes Fruit Ninja-style with accurate marks, arrows leave puncture marks, blunt weapons smash pots. Enemy armor breaks in combat. Naoe's grappling hook is physics-simulated, attachable to every roof edge, usable to swing, to climb into ceilings, or to launch into assassinations. IGN documents the per-weapon damage replication and physics hook.
4. Less map vomit. The map is similar in size to Origins, covering Kansai — Kyoto, Kobe, Osaka, Iga — with castles rendered near 1:1 as centerpiece dungeons. Sync points no longer auto-plot icons; they are vantage points to survey highlights. Targets are non-linear, with hints and a spy network to narrow locations rather than exact markers. You can switch Naoe and Yasuke between objectives, with shared XP but separate skill trees and gear. Prone crawling, knockout takedowns, shoji-door assassinations and nightingale floors that creak if you rush complete the kit. Wikipedia summarizes the Origins-size world, scouts and vantage change.
What Anvil gained — and what it cost
What it gained is readability without scale bloat. Because Anvil already knew how to stream a 2,000-NPC Unity Paris and a Black Flag ocean without loads, Quebec could make one Origins-size Japan feel like four maps by toggling hiding spots, water access and patrol logic per season. Dynamic light did AI work: instead of scripting smarter guards, they made guards see worse in storms and better near lanterns. Breakable props did combat work: Yasuke's armor-break duels and Naoe's bamboo-and-shuriken stealth read instantly because every slash leaves a mark.
What it cost is familiar to anyone who played Unity at launch. AnvilNext 2.0 was rebuilt after Unity's technical issues and then fixed for Syndicate — that scar is why Shadows was delayed twice, from November 15, 2024 to February 14, 2025 after Star Wars Outlaws, then to March 20, 2025 while Ubisoft explored sale options, with day-one Steam launch and free expansion to compensate. Dropping last-gen cut the audience. The gore toggle and Japan-specific censorship to satisfy CERO, the Yasuke backlash and harassment of developers, and critiques of messy story and repetitive RPG structure show the engine could not fix content or culture. Steam peaked at 41,412 concurrent on launch day and 64,825 that weekend — the highest for Assassin's Creed on Steam — but half of early players by some estimates came via subscriptions, not sales. Anvil bought stability, not love.
My read: four contrarian takes you will not find in the reviews
Most coverage calls seasons eye-candy and Anvil outdated. I think that misses the point. These are interpretations, not verified facts, but they explain why the boring choice worked:
1. Seasons are anti-bloat, not graphics. Reviewers compare Shadows' seasons to Ghost of Yotei's dynamic weather around Mount Yotei — snow, rain, aurora — and declare Yotei prettier. Yotei sold 3.3 million by November 2025 after its October 2 PS5 launch with guiding wind, grappling from the start and Kurosawa Mode. Pretty is the wrong metric. Shadows' seasons quadruple infiltration routes without quadrupling art. One castle has a summer grass approach, an autumn exposed approach, and a winter frozen-moat approach. That is how a team that spent over $100 million still ships on time: reuse geography, change rules. Yotei uses weather as mood; Shadows uses seasons as level budget.
2. Darkness is currency. Everyone says Shadows finally brings back stealth. I say it brings back Splinter Cell's light meter because Anvil could not afford smarter AI. Making lantern oil destructible is cheaper than writing a new behavior tree for 15 studios to learn. Servants who cannot fight but can alarm, torch carriers you must snipe first, icicles that betray rooftops — that is an economy where light is ammo. It is inelegant and brilliant.
3. Yasuke is an engine excuse. The controversy over playing a real African samurai overshadowed a technical function: a foreigner needs scouts, hints and Observe mode. A native Iga shinobi would know every province. By pairing Naoe who knows with Yasuke who does not, Ubisoft justifies removing exact markers and the drone bird from Origins/Odyssey/Valhalla. The much-criticized scout system is not dumbing down; it is Anvil admitting it cannot hand-place 1,000 icons without choking, so it makes you earn the icon. Story covers tech debt.
4. Factory beats atelier for rescues. Sucker Punch's nameless in-house engine is a wind machine — wind as UI, grass navigation, minimal markers. Anvil is a factory — procedural blocks, HumanIK climbs, Siege netcode, Scalar cloud experiments. If you want one perfect duel on a mountain, pick the wind machine. If you need two protagonists, four seasons, breakable castles, spy networks and 15 support studios after a flop year, pick the factory. My hot take: Unreal with Nanite would have looked better in trailers but shipped later and broken co-op pipelines Ubisoft needs for Siege and future Hexe. Boring is a moat.
What indie devs can actually steal
You do not have Anvil or €100 million. Steal the principles:
- Tie art to mechanics on day one. If you add snow, remove a hiding spot. If you add pollen wind in spring, make it mask sound. Shadows' ecology — pollen gusts, falling leaves — is not decoration; it is stealth tuning.
- Delete the drone. Shadows removed the bird and forced ground reconnaissance with Observe and scouts. Your prototype does not need auto-markers; hints are cheaper and more fun.
- Make destruction legible, not realistic. You cannot simulate every pot, but you can do what Shadows does: katana leaves slice, arrow leaves hole, blunt smashes. Three rules, applied everywhere, feel systemic.
- Build one map, ship four. If you cannot afford Origins size, build one alley and change season, light and patrol warmth. Players remember routes, not square kilometers.
Feel the loop in five minutes in your browser
You will not learn Naoe's patience or Yasuke's timing from a trailer. If you have five minutes between downloads, try two free browser trainers that isolate the same verbs:
If you want the Brotherhood fantasy — hit-and-run against Templars, pirates and mythical bosses across Egypt, medieval cities, Valhalla hills and the feudal Orient — Assassin Hero distills it to roguelike positioning and knowing when to vanish. The page describes hit-and-run skills against hordes while chasing legendary assassin status. Play it as Naoe practice: do not trade blows, break contact, re-enter from darkness.
If you want Yasuke's duel fantasy — stance, block, parry, break armor — 3D Block Gladiator: Sword Draw forces the same discipline with a different input: you draw precise lines to attack, block and dodge instead of mashing. The page notes every slash counts and strategy beats button-mashing. Play it as armor-break practice: wait, draw the parry line late, punish.
Neither replaces Shadows' seasons or castles. Both teach the decision Shadows rewards: when to stay in shadow and when to draw.
Bottom line
Shadows bet that an 18-year-old house engine, upgraded for light, breakage and seasons, could carry Japan better than a licensed marvel. With 5 million players, second-best day-one revenue in series history and best-selling March in the U.S., the bet kept Ubisoft alive. It did not make Anvil modern — it made modern constraints (no last-gen, no markers, no drone) feel like design.
If you are choosing tech today, ask the Anvil question, not the Unreal question: what does your engine already know how to stream, break and light without retraining your team? Answer that honestly, then make the weather change the rules.


![AMR AI and Embodied Intelligence — Ch.18: GPU and Accelerator Design [KR]](https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fospxyzadkcmpo8xjqyok.png)