diff --git a/vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe.py b/vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe.py
index c2b884c058d3a..5470deb768450 100644
--- a/vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe.py
+++ b/vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe.py
@@ -398,7 +398,8 @@ class CompressedTensorsW4A4MoeMethod(CompressedTensorsMoEMethod):
                 quant_type_id=scalar_types.float4_e2m1f.id,
                 apply_router_weight_on_input=apply_router_weight_on_input,
                 global_num_experts=global_num_experts,
-                expert_map=expert_map)
+                expert_map=expert_map,
+                workspace=layer.workspace)
 
         # FlashInfer fused experts path
         if self.fused_experts is not None:
@@ -940,7 +941,8 @@ class CompressedTensorsW8A8Fp8MoEMethod(CompressedTensorsMoEMethod):
                 quant_type_id=scalar_types.float8_e4m3fn.id,
                 apply_router_weight_on_input=apply_router_weight_on_input,
                 global_num_experts=global_num_experts,
-                expert_map=expert_map)
+                expert_map=expert_map,
+                workspace=layer.workspace)
 
         assert self.fused_experts_func is not None
 
diff --git a/vllm/model_executor/layers/quantization/fp8.py b/vllm/model_executor/layers/quantization/fp8.py
index 49ff87df93c31..254cc2be05ee6 100644
--- a/vllm/model_executor/layers/quantization/fp8.py
+++ b/vllm/model_executor/layers/quantization/fp8.py
@@ -1103,7 +1103,8 @@ class Fp8MoEMethod(FusedMoEMethodBase):
                 quant_type_id=scalar_types.float8_e4m3fn.id,
                 apply_router_weight_on_input=apply_router_weight_on_input,
                 global_num_experts=global_num_experts,
-                expert_map=expert_map)
+                expert_map=expert_map,
+                workspace=layer.workspace)
         elif self.flashinfer_moe_backend == FlashinferMoeBackend.CUTLASS:
             assert self.block_quant is None
             assert (not renormalize and custom_routing_function is not None)
diff --git a/vllm/model_executor/layers/quantization/modelopt.py b/vllm/model_executor/layers/quantization/modelopt.py
index 9b99931e7b43f..60a79e53e8141 100644
--- a/vllm/model_executor/layers/quantization/modelopt.py
+++ b/vllm/model_executor/layers/quantization/modelopt.py
@@ -1474,7 +1474,8 @@ class ModelOptNvFp4FusedMoE(FusedMoEMethodBase):
                 quant_type_id=scalar_types.float4_e2m1f.id,
                 apply_router_weight_on_input=apply_router_weight_on_input,
                 global_num_experts=global_num_experts,
-                expert_map=expert_map)
+                expert_map=expert_map,
+                workspace=layer.workspace)
 
         if self.fused_experts is not None:
             assert self.allow_flashinfer and \