diff --git a/onnxruntime/test/providers/qnn/matmul_test.cpp b/onnxruntime/test/providers/qnn/matmul_test.cpp
index e721ccbcb45a9..3073dde9d8e4c 100644
--- a/onnxruntime/test/providers/qnn/matmul_test.cpp
+++ b/onnxruntime/test/providers/qnn/matmul_test.cpp
@@ -112,12 +112,13 @@ TEST_F(QnnCPUBackendTests, MatMulOp) {
 }
 
 // Test MatMul broadcasting
-// Note slight inaccuracy in CPU backend:
+// Failed randomly on Linux
+// Value of: expected_tensor.DataAsSpan<float>()
 // Expected: contains 896 values, where each value and its corresponding value in 16-byte object
-// <80-03 00-00 00-00 00-00 40-00 34-DD F7-01 00-00> are an almost-equal pair
-// Actual: 16-byte object <80-03 00-00 00-00 00-00 40-00 23-DD F7-01 00-00>,
-// where the value pair (73.68116, 73.680809) at index #80 don't match, which is -0.000350952 from 73.6812
-TEST_F(QnnCPUBackendTests, MatMulOp_Broadcast) {
+// <80-03 00-00 00-00 00-00 40-B8 53-08 CC-7F 00-00> are an almost-equal pair
+// Actual: 16-byte object <80-03 00-00 00-00 00-00 C0-B7 43-08 CC-7F 00-00>, where the value pair
+// (-5.19657087, 0) at index #29 don't match, which is 5.19657 from -5.19657
+TEST_F(QnnCPUBackendTests, DISABLED_MatMulOp_Broadcast) {
   // Create two matrices with element values in the range [-10.0, 10.0].
   std::vector<float> input_a = GetFloatDataInRange(-10.0f, 10.0f, 28 * 64);
   std::vector<float> input_b = GetFloatDataInRange(-10.0f, 10.0f, 64 * 32);