diff --git a/packages/cli/src/commands/render.ts b/packages/cli/src/commands/render.ts
index c70899cb4..3fe44e20d 100644
--- a/packages/cli/src/commands/render.ts
+++ b/packages/cli/src/commands/render.ts
@@ -1102,6 +1102,7 @@ let deParallelRouterTrialFiredThisProcess = false;
  * resetting outside a test process where many independent test cases share
  * one imported module instance.
  */
+// fallow-ignore-next-line unused-export
 export function __resetDeParallelRouterTrialStateForTests(): void {
   deParallelRouterTrialManagedByUs = false;
   deParallelRouterTrialFiredThisProcess = false;
diff --git a/packages/cli/src/telemetry/events.ts b/packages/cli/src/telemetry/events.ts
index 05a0bb50f..d7593d84f 100644
--- a/packages/cli/src/telemetry/events.ts
+++ b/packages/cli/src/telemetry/events.ts
@@ -51,6 +51,7 @@ export interface RenderObservabilityTelemetryPayload {
   captureDeWorkerInversion?: string;
   captureDePreInversionWorkers?: number;
   captureDeParallelRouter?: string;
+  captureDeGpuRenderer?: string;
   captureDePreRouterWorkers?: number;
   captureDeSelfVerifyFallback?: boolean;
   captureDeFallbackReason?: string;
@@ -108,6 +109,7 @@ function renderObservabilityEventProperties(props: RenderObservabilityTelemetryP
     de_worker_inversion: props.captureDeWorkerInversion,
     de_pre_inversion_workers: props.captureDePreInversionWorkers,
     de_parallel_router: props.captureDeParallelRouter,
+    gpu_renderer: props.captureDeGpuRenderer,
     de_pre_router_workers: props.captureDePreRouterWorkers,
     de_self_verify_fallback: props.captureDeSelfVerifyFallback,
     de_fallback_reason: props.captureDeFallbackReason,
@@ -183,7 +185,7 @@ export function trackRenderComplete(
     deParallelRouter?: string;
     dePreRouterWorkers?: number;
     deGateReason?: string;
-    /** Raw WebGL renderer string from DE session init (ANGLE backend + GPU vendor). */
+    /** Low-cardinality GPU bucket from DE session init (`<backend>/<vendor>`, e.g. `d3d11/nvidia`). */
     gpuRenderer?: string;
     deWorkerEncode?: boolean;
     deVerifyArmed?: number;
@@ -377,6 +379,10 @@ export function trackRenderError(
       elapsed_ms: props.elapsedMs,
       peak_memory_mb: props.peakMemoryMb,
       memory_free_mb: props.memoryFreeMb,
+      // gpu_renderer arrives via renderObservabilityEventProperties below:
+      // on the failure path perfSummary is never built, so live capture
+      // observability is the only source. Backend attribution matters MOST
+      // here — a win32 D3D11 crash is what the rollout is watching for.
       ...renderObservabilityEventProperties(props),
     },
     props.distinctId,
diff --git a/packages/cli/src/telemetry/renderObservability.ts b/packages/cli/src/telemetry/renderObservability.ts
index 3df009f4d..87554d693 100644
--- a/packages/cli/src/telemetry/renderObservability.ts
+++ b/packages/cli/src/telemetry/renderObservability.ts
@@ -43,6 +43,7 @@ export function renderObservabilityTelemetryPayload(
     captureDeWorkerInversion: capture.deWorkerInversion,
     captureDePreInversionWorkers: capture.dePreInversionWorkers,
     captureDeParallelRouter: capture.deParallelRouter,
+    captureDeGpuRenderer: capture.deGpuRenderer,
     captureDePreRouterWorkers: capture.dePreRouterWorkers,
     captureDeSelfVerifyFallback: capture.deSelfVerifyFallback,
     captureDeFallbackReason: capture.deFallbackReason,
diff --git a/packages/engine/src/services/drawElementService.test.ts b/packages/engine/src/services/drawElementService.test.ts
index a21f9602e..364542507 100644
--- a/packages/engine/src/services/drawElementService.test.ts
+++ b/packages/engine/src/services/drawElementService.test.ts
@@ -1,6 +1,7 @@
 import { describe, expect, it, vi } from "vitest";
 import type { Page } from "puppeteer-core";
 import {
+  classifyGpuRenderer,
   detectGpuBackend,
   detectSwiftShader,
   resolveDrawElementCaptureMode,
@@ -48,6 +49,42 @@ describe("detectGpuBackend", () => {
   });
 });
 
+// ── classifyGpuRenderer ────────────────────────────────────────────────────────
+
+describe("classifyGpuRenderer", () => {
+  it("buckets real ANGLE renderer strings to <backend>/<vendor>", () => {
+    expect(
+      classifyGpuRenderer("ANGLE (Apple, ANGLE Metal Renderer: Apple M4 Pro, Unspecified Version)"),
+    ).toBe("metal/apple");
+    expect(
+      classifyGpuRenderer(
+        "ANGLE (NVIDIA, NVIDIA GeForce RTX 3080 Direct3D11 vs_5_0 ps_5_0, D3D11)",
+      ),
+    ).toBe("d3d11/nvidia");
+    expect(
+      classifyGpuRenderer("ANGLE (Intel, Intel(R) UHD Graphics 630 Direct3D11 vs_5_0 ps_5_0)"),
+    ).toBe("d3d11/intel");
+    expect(classifyGpuRenderer("ANGLE (AMD, AMD Radeon RX 6800 Direct3D11 vs_5_0 ps_5_0)")).toBe(
+      "d3d11/amd",
+    );
+    expect(classifyGpuRenderer("Google SwiftShader")).toBe("swiftshader/other");
+  });
+
+  it("drops the GPU model — the bucket must stay low cardinality", () => {
+    // Two different NVIDIA cards must collapse to ONE bucket, otherwise the
+    // property is unbounded and useless for aggregation.
+    expect(classifyGpuRenderer("ANGLE (NVIDIA, NVIDIA GeForce RTX 4090 Direct3D11)")).toBe(
+      classifyGpuRenderer("ANGLE (NVIDIA, NVIDIA GeForce GTX 1060 Direct3D11)"),
+    );
+  });
+
+  it("returns undefined for missing input rather than a bogus bucket", () => {
+    expect(classifyGpuRenderer(null)).toBeUndefined();
+    expect(classifyGpuRenderer(undefined)).toBeUndefined();
+    expect(classifyGpuRenderer("")).toBeUndefined();
+  });
+});
+
 // ── resolveDrawElementCaptureMode ──────────────────────────────────────────────
 
 describe("resolveDrawElementCaptureMode", () => {
diff --git a/packages/engine/src/services/drawElementService.ts b/packages/engine/src/services/drawElementService.ts
index a5d08eb08..05f8ab820 100644
--- a/packages/engine/src/services/drawElementService.ts
+++ b/packages/engine/src/services/drawElementService.ts
@@ -110,15 +110,58 @@ export interface GpuBackendInfo {
   isSwiftShader: boolean;
   /**
    * Raw UNMASKED_RENDERER_WEBGL string (e.g. "ANGLE (Apple, ANGLE Metal
-   * Renderer: Apple M4 Pro, ...)", "ANGLE (NVIDIA, D3D11 ...)"), or null when
-   * WebGL / the debug extension is unavailable. Carried to render telemetry:
-   * drawElement failure modes proved compositor-backend-specific during the
-   * macOS rollout, so the win32/D3D11 cohort needs damage clusters
-   * attributable to a GPU vendor + ANGLE backend, not just `os`.
+   * Renderer: Apple M4 Pro, ...)", "ANGLE (NVIDIA, GeForce RTX 3080 Direct3D11
+   * vs_5_0 ps_5_0, D3D11)"), or null when WebGL / the debug extension is
+   * unavailable. LOCAL USE ONLY — this is unbounded driver-supplied text and
+   * must not be shipped to telemetry verbatim; send
+   * {@link classifyGpuRenderer}'s bucket instead.
    */
   renderer: string | null;
 }
 
+/**
+ * Low-cardinality bucket for a raw WebGL renderer string: `<backend>/<vendor>`
+ * (e.g. `metal/apple`, `d3d11/nvidia`, `swiftshader/other`).
+ *
+ * drawElement failure modes proved compositor-backend-specific during the
+ * macOS rollout, so the win32/D3D11 cohort needs damage attributable to an
+ * ANGLE backend + GPU vendor. The raw string can't do that job in telemetry:
+ * it is unbounded, driver-authored, carries specific GPU model names, and is
+ * joined across parallel sessions — high cardinality by construction. The
+ * bucket keeps the analytic signal (which backend, which vendor) and drops
+ * everything else, matching how `deGateReason` is a sanitized bucket rather
+ * than the full fallback trigger. Pure; exported for tests.
+ */
+export function classifyGpuRenderer(renderer: string | null | undefined): string | undefined {
+  if (!renderer) return undefined;
+  const r = renderer.toLowerCase();
+  const backend = r.includes("swiftshader")
+    ? "swiftshader"
+    : r.includes("metal")
+      ? "metal"
+      : r.includes("direct3d11") || r.includes("d3d11")
+        ? "d3d11"
+        : r.includes("direct3d9") || r.includes("d3d9")
+          ? "d3d9"
+          : r.includes("vulkan")
+            ? "vulkan"
+            : r.includes("opengl") || r.includes("angle")
+              ? "opengl"
+              : "other";
+  const vendor = r.includes("apple")
+    ? "apple"
+    : r.includes("nvidia")
+      ? "nvidia"
+      : r.includes("amd") || r.includes("radeon")
+        ? "amd"
+        : r.includes("intel")
+          ? "intel"
+          : r.includes("microsoft")
+            ? "microsoft"
+            : "other";
+  return `${backend}/${vendor}`;
+}
+
 /**
  * Detect the page's WebGL backend: SwiftShader vs a real GPU, plus the raw
  * renderer string for telemetry.
diff --git a/packages/engine/src/services/frameCapture.ts b/packages/engine/src/services/frameCapture.ts
index 88e044cb4..4751f7a05 100644
--- a/packages/engine/src/services/frameCapture.ts
+++ b/packages/engine/src/services/frameCapture.ts
@@ -34,6 +34,7 @@ import {
   shouldDefaultCaptureBeyondViewport,
 } from "./screenshotService.js";
 import {
+  classifyGpuRenderer,
   detectGpuBackend,
   injectDrawElementCanvas,
   captureDrawElementFrame,
@@ -147,10 +148,12 @@ export interface CaptureSession {
   /** True if running on SwiftShader (detected at init). Undefined before init. */
   isSwiftShader?: boolean;
   /**
-   * Raw WebGL UNMASKED_RENDERER_WEBGL string, captured alongside the
-   * SwiftShader probe at DE session init (e.g. "ANGLE (NVIDIA, D3D11 ...)").
-   * Surfaces in CapturePerfSummary → render telemetry so backend-specific
-   * drawElement damage (Metal vs D3D11 vs GL) clusters attributably.
+   * Low-cardinality GPU bucket (`<backend>/<vendor>`, e.g. `d3d11/nvidia`)
+   * derived from the WebGL renderer at DE session init. Surfaces in
+   * CapturePerfSummary → render telemetry so backend-specific drawElement
+   * damage (Metal vs D3D11 vs GL) clusters attributably. The raw
+   * driver-supplied string is deliberately NOT retained — see
+   * classifyGpuRenderer.
    */
   gpuRenderer?: string;
   /** drawElementImage canvas was injected and is ready for capture. */
@@ -713,7 +716,7 @@ async function initDrawElementOrTransparentBackground(
   if (useDrawElement) {
     const gpuBackend = await detectGpuBackend(page);
     session.isSwiftShader = gpuBackend.isSwiftShader;
-    session.gpuRenderer = gpuBackend.renderer ?? undefined;
+    session.gpuRenderer = classifyGpuRenderer(gpuBackend.renderer);
     const transparent = session.options.format === "png";
     async function routeToFallback(): Promise<void> {
       session.captureMode = session.launchCaptureMode;
diff --git a/packages/engine/src/types.ts b/packages/engine/src/types.ts
index 9253e00df..4f30c1efb 100644
--- a/packages/engine/src/types.ts
+++ b/packages/engine/src/types.ts
@@ -277,11 +277,11 @@ export interface CapturePerfSummary {
   /** Final capture mode this session used: "drawelement" | "screenshot" | "beginframe". */
   captureMode: string;
   /**
-   * Raw WebGL UNMASKED_RENDERER_WEBGL string from DE session init (ANGLE
-   * backend + GPU vendor, e.g. "ANGLE (Apple, ANGLE Metal Renderer: ...)").
-   * Undefined when drawElement was never attempted. Lets telemetry cluster
-   * backend-specific damage now that DE engages on both Metal (darwin) and
-   * D3D11 (win32).
+   * Low-cardinality GPU bucket from DE session init: `<backend>/<vendor>`
+   * (e.g. `metal/apple`, `d3d11/nvidia`). Undefined when drawElement was
+   * never attempted. Lets telemetry cluster backend-specific damage now that
+   * DE engages on both Metal (darwin) and D3D11 (win32). Bucketed, not raw —
+   * see `classifyGpuRenderer`.
    */
   gpuRenderer?: string;
   /**
diff --git a/packages/producer/src/services/render/observability.ts b/packages/producer/src/services/render/observability.ts
index 23bc93814..39111960c 100644
--- a/packages/producer/src/services/render/observability.ts
+++ b/packages/producer/src/services/render/observability.ts
@@ -72,6 +72,13 @@ export interface RenderCaptureObservability {
   dePreInversionWorkers?: number;
   /** DE parallel-router outcome: "routed" (fired, held) | "reverted" (fired, self-verify retry rolled back). */
   deParallelRouter?: "routed" | "reverted";
+  /**
+   * Low-cardinality GPU bucket (`<backend>/<vendor>`) from the DE probe
+   * session. Lives on capture observability (not just perfSummary) so a hard
+   * failure — crash / OOM / timeout — still reports which GPU backend it hit:
+   * that is precisely the cohort the win32 D3D11 rollout must attribute.
+   */
+  deGpuRenderer?: string;
   /** Worker count the resolver would have used absent the router; undefined if it never fired. */
   dePreRouterWorkers?: number;
   /**
diff --git a/packages/producer/src/services/renderOrchestrator.ts b/packages/producer/src/services/renderOrchestrator.ts
index 5417ef3fd..b6bb80af2 100644
--- a/packages/producer/src/services/renderOrchestrator.ts
+++ b/packages/producer/src/services/renderOrchestrator.ts
@@ -498,7 +498,7 @@ export interface RenderPerfSummary {
     preRouterWorkers?: number;
     /** Engine init-time gate: swiftshader | css_effect:* | at_risk_timeline | 3d_init_failed | supersampling | render_mode_hint. */
     gateReason?: string;
-    /** Raw WebGL renderer string from DE session init (ANGLE backend + GPU vendor); |-joined across parallel sessions. */
+    /** Low-cardinality GPU bucket from DE session init (`<backend>/<vendor>`, e.g. `d3d11/nvidia`); |-joined across parallel sessions (bounded: one bucket per distinct backend on the host). */
     gpuRenderer?: string;
     /** Worker-encode drain (the verified path) was active. */
     workerEncode: boolean;
@@ -2644,6 +2644,11 @@ async function executeRenderPipeline(input: {
       // any resource-pressure failure unique to this cohort.
       dePreInversionWorkers: deWorkerInversion ? preRoutingWorkerCount : undefined,
       dePreRouterWorkers: deParallelRouter ? preRoutingWorkerCount : undefined,
+      // Same rationale as the counters above: carried on live capture
+      // observability, not only the success-path perfSummary, so a crash /
+      // OOM / timeout still reports which GPU backend it happened on. That
+      // is the cohort the win32 D3D11 rollout most needs to attribute.
+      deGpuRenderer: probeSession?.gpuRenderer,
     });
     observability.checkpoint("worker_resolution", "resolved", {
       workerCount,