diff --git a/vllm/model_executor/layers/activation.py b/vllm/model_executor/layers/activation.py
index 43056786..ab4be153 100644
--- a/vllm/model_executor/layers/activation.py
+++ b/vllm/model_executor/layers/activation.py
@@ -33,7 +33,31 @@ class SiluAndMul(CustomOp):
 
         d = x.shape[-1] // 2
         output_shape = (x.shape[:-1] + (d, ))
-        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        # ═══════════════════════════════════════════════════════════════
+        # CCCL dispatch_transform.cuh CacheAsyncConfiguration pattern:
+        #   "This computation MUST NOT depend on any runtime state of the
+        #    current API invocation (like num_items), since the result
+        #    will be cached."
+        #
+        # For element-wise transforms, the output tensor shape is
+        # deterministic from the input shape. During decode, input shape
+        # is stable (num_seqs × hidden_dim doesn't change between steps).
+        # Cache the output tensor to avoid cudaMalloc on every step.
+        #
+        # CCCL also uses spread_out_items_per_thread to dynamically
+        # adjust tile size for small problems — analogously, we only
+        # cache when shapes are stable (decode), not during prefill
+        # where shapes vary per request.
+        # ═══════════════════════════════════════════════════════════════
+        _cache_key = (output_shape, x.dtype, x.device)
+        _cached = getattr(self, '_out_cache', {}).get(_cache_key)
+        if _cached is not None and _cached.shape == output_shape:
+            out = _cached
+        else:
+            out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+            if not hasattr(self, '_out_cache'):
+                self._out_cache = {}
+            self._out_cache[_cache_key] = out
         ops.silu_and_mul(out, x)
         return out