diff --git a/src/App.zig b/src/App.zig
index 8f749890f..15e06a03a 100644
--- a/src/App.zig
+++ b/src/App.zig
@@ -106,6 +106,18 @@ fn threadedWarmup() void {
             log.warn("font warmup thread spawn failed err={}", .{err});
         }
     }
+
+    // Same for the renderer's graphics API (e.g. Metal), which pays
+    // one-time framework initialization costs on first use.
+    if (comptime @hasDecl(renderer.Renderer.API, "warmup")) {
+        if (std.Thread.spawn(
+            .{},
+            renderer.Renderer.API.warmup,
+            .{},
+        )) |thr| thr.detach() else |err| {
+            log.warn("renderer warmup thread spawn failed err={}", .{err});
+        }
+    }
 }
 
 /// Initialize the main app instance. This creates the main window, sets
diff --git a/src/renderer/Metal.zig b/src/renderer/Metal.zig
index 6c7432d21..1ad6b812b 100644
--- a/src/renderer/Metal.zig
+++ b/src/renderer/Metal.zig
@@ -407,6 +407,23 @@ pub inline fn beginFrame(
     return try Frame.begin(.{ .queue = self.queue }, renderer, target);
 }
 
+/// Warm up the Metal device machinery. The first Metal device query in
+/// a process takes multiple milliseconds; once warm, subsequent queries
+/// are effectively free. Calling this early (e.g. on a background
+/// thread at app startup; Metal device queries are thread-safe) moves
+/// that one-time cost off the critical path of the first surface's
+/// renderer initialization.
+///
+/// We deliberately do NOT cache the chosen device: the device set can
+/// change at runtime (e.g. an eGPU being plugged in or removed) and
+/// chooseDevice prefers removable GPUs, so every renderer init must
+/// re-choose. Only the underlying framework initialization is a
+/// one-time cost.
+pub fn warmup() void {
+    const device = chooseDevice() catch return;
+    device.release();
+}
+
 fn chooseDevice() error{NoMetalDevice}!objc.Object {
     var chosen_device: ?objc.Object = null;