diff --git a/tests/kernels/attention/test_cpu_attn.py b/tests/kernels/attention/test_cpu_attn.py
index 6af1bfe1e7ac..c4bd10215486 100644
--- a/tests/kernels/attention/test_cpu_attn.py
+++ b/tests/kernels/attention/test_cpu_attn.py
@@ -9,7 +9,7 @@
 
 from vllm.platforms import CpuArchEnum, current_platform
 from vllm.utils.torch_utils import set_random_seed
-from vllm.v1.attention.backends.cpu_attn import _get_attn_isa
+from vllm.v1.attention.backends.cpu_attn import CPUAttentionBackend, _get_attn_isa
 
 if not current_platform.is_cpu():
     pytest.skip("skipping CPU-only tests", allow_module_level=True)
@@ -59,6 +59,10 @@ def get_attn_isa(
     )
 
 
+def test_cpu_backend_requires_hnd_kv_cache_layout():
+    assert CPUAttentionBackend.get_required_kv_cache_layout() == "HND"
+
+
 # rand number generation takes too much time, cache rand tensors
 @functools.lru_cache(maxsize=128, typed=False)
 def tensor_cache(
diff --git a/vllm/v1/attention/backends/cpu_attn.py b/vllm/v1/attention/backends/cpu_attn.py
index e28e20045148..801a16d319e1 100644
--- a/vllm/v1/attention/backends/cpu_attn.py
+++ b/vllm/v1/attention/backends/cpu_attn.py
@@ -25,6 +25,7 @@
     MultipleOf,
 )
 from vllm.v1.attention.backends.utils import (
+    KVCacheLayoutType,
     split_decodes_and_prefills,
 )
 from vllm.v1.kv_cache_interface import AttentionSpec, CrossAttentionSpec
@@ -94,6 +95,10 @@ def get_kv_cache_shape(
     ) -> tuple[int, ...]:
         return 2, num_blocks, num_kv_heads, block_size, head_size
 
+    @classmethod
+    def get_required_kv_cache_layout(cls) -> "KVCacheLayoutType | None":
+        return "HND"
+
     @staticmethod
     def use_cascade_attention(*args, **kwargs) -> bool:
         return False