diff --git a/vllm/model_executor/layers/fused_moe/unquantized_fused_moe_method.py b/vllm/model_executor/layers/fused_moe/unquantized_fused_moe_method.py
index 91de16f79c..d9cfce2c21 100644
--- a/vllm/model_executor/layers/fused_moe/unquantized_fused_moe_method.py
+++ b/vllm/model_executor/layers/fused_moe/unquantized_fused_moe_method.py
@@ -171,6 +171,11 @@ class UnquantizedFusedMoEMethod(FusedMoEMethodBase, CustomOp):
         replace_parameter(layer, "w13_weight", w13_new, prefer_copy=is_weight_update)
         replace_parameter(layer, "w2_weight", w2_new, prefer_copy=is_weight_update)
 
+        # AITER backend requires weights to be marked as shuffled.
+        if self.unquantized_backend == UnquantizedMoeBackend.AITER:
+            layer.w13_weight.is_shuffled = True
+            layer.w2_weight.is_shuffled = True
+
         if not is_weight_update:
             # Setup moe kernel only on the first call. For the unquantized
             # method, moe_quant_config is either the constant
diff --git a/vllm/model_executor/layers/quantization/fp8.py b/vllm/model_executor/layers/quantization/fp8.py
index 1c9237d3f6..30e9a6344b 100644
--- a/vllm/model_executor/layers/quantization/fp8.py
+++ b/vllm/model_executor/layers/quantization/fp8.py
@@ -28,6 +28,7 @@ from vllm.model_executor.layers.fused_moe.config import (
 )
 from vllm.model_executor.layers.fused_moe.layer import UnquantizedFusedMoEMethod
 from vllm.model_executor.layers.fused_moe.oracle.fp8 import (
+    Fp8MoeBackend,
     convert_to_fp8_moe_kernel_format,
     make_fp8_moe_kernel,
     make_fp8_moe_quant_config,
@@ -766,6 +767,11 @@ class Fp8MoEMethod(FusedMoEMethodBase):
         replace_parameter(layer, f"w13_{self.weight_scale_name}", w13_scale)
         replace_parameter(layer, f"w2_{self.weight_scale_name}", w2_scale)
 
+        # AITER backend requires weights to be marked as shuffled.
+        if self.fp8_backend == Fp8MoeBackend.AITER:
+            layer.w13_weight.is_shuffled = True
+            layer.w2_weight.is_shuffled = True
+
         self.moe_quant_config = self.get_fused_moe_quant_config(layer)
         if self.moe_quant_config:
             assert self.experts_cls is not None
diff --git a/vllm/model_executor/layers/quantization/mxfp4.py b/vllm/model_executor/layers/quantization/mxfp4.py
index d6fef0b3d3..d3f3033be7 100644
--- a/vllm/model_executor/layers/quantization/mxfp4.py
+++ b/vllm/model_executor/layers/quantization/mxfp4.py
@@ -350,6 +350,11 @@ class GptOssMxfp4MoEMethod(FusedMoEMethodBase):
             self.w13_precision_config = w13_scale
             self.w2_precision_config = w2_scale
 
+        # AITER backend requires weights to be marked as shuffled.
+        if self.mxfp4_backend == Mxfp4MoeBackend.AITER_MXFP4_BF16:
+            layer.w13_weight.is_shuffled = True
+            layer.w2_weight.is_shuffled = True
+
         if w13_bias is not None and w2_bias is not None:
             replace_parameter(layer, "w13_bias", w13_bias)
             replace_parameter(layer, "w2_bias", w2_bias)
@@ -678,6 +683,11 @@ class Mxfp4MoEMethod(FusedMoEMethodBase):
             self.w13_precision_config = w13_scale
             self.w2_precision_config = w2_scale
 
+        # AITER backend requires weights to be marked as shuffled.
+        if self.mxfp4_backend == Mxfp4MoeBackend.AITER_MXFP4_BF16:
+            layer.w13_weight.is_shuffled = True
+            layer.w2_weight.is_shuffled = True
+
         if w13_bias is not None and w2_bias is not None:
             replace_parameter(layer, "w13_bias", w13_bias)
             replace_parameter(layer, "w2_bias", w2_bias)