diff --git a/bitsandbytes/autograd/_functions.py b/bitsandbytes/autograd/_functions.py
index 6d473e940..f4a6d5736 100644
--- a/bitsandbytes/autograd/_functions.py
+++ b/bitsandbytes/autograd/_functions.py
@@ -232,8 +232,8 @@ def forward(ctx, A, B, out=None, bias=None, state=MatmulLtState()):
         # Cast A to fp16
         A_dtype = A.dtype
         if A_dtype != torch.float16:
-            warnings.warn(f"MatMul8bitLt: temporarily casting input matrix from {A_dtype} to float16")
-        A = A.to(torch.float16)
+            warnings.warn(f"MatMul8bitLt: input matrix will be converted from {A_dtype} to float16")
+            A = A.to(torch.float16)
 
         # 1. Quantize A
         if len(A.shape) == 3: