diff --git a/be/src/gutil/cpu.cc b/be/src/gutil/cpu.cc
index 91cf92ef63..a7d9bc6540 100644
--- a/be/src/gutil/cpu.cc
+++ b/be/src/gutil/cpu.cc
@@ -70,6 +70,7 @@ CPU::CPU()
           has_popcnt_(false),
           has_avx_(false),
           has_avx2_(false),
+          has_avx512_(false),
           has_aesni_(false),
           has_non_stop_time_stamp_counter_(false),
           is_running_in_vm_(false),
@@ -201,6 +202,8 @@ void CPU::Initialize() {
                    (xgetbv(0) & 6) == 6 /* XSAVE enabled by kernel */;
         has_aesni_ = (cpu_info[2] & 0x02000000) != 0;
         has_avx2_ = has_avx_ && (cpu_info7[1] & 0x00000020) != 0;
+        has_avx512_ = has_avx2_ && (cpu_info7[1] & 0x00010000) != 0 &&
+                   (cpu_info7[1] & 0x40000000) != 0 && (cpu_info7[1] & 0x80000000) != 0;
     }
     // Get the brand string of the cpu.
     __cpuid(cpu_info, 0x80000000);
@@ -253,6 +256,7 @@ void CPU::Initialize() {
 #endif
 }
 CPU::IntelMicroArchitecture CPU::GetIntelMicroArchitecture() const {
+    if (has_avx512()) return AVX512;
     if (has_avx2()) return AVX2;
     if (has_avx()) return AVX;
     if (has_sse42()) return SSE42;
diff --git a/be/src/gutil/cpu.h b/be/src/gutil/cpu.h
index f7a12bbe43..82b87a1fb3 100644
--- a/be/src/gutil/cpu.h
+++ b/be/src/gutil/cpu.h
@@ -60,6 +60,7 @@ public:
         SSE42,
         AVX,
         AVX2,
+        AVX512,
         MAX_INTEL_MICRO_ARCHITECTURE
     };
     // Accessors for CPU information.
@@ -81,6 +82,7 @@ public:
     bool has_popcnt() const { return has_popcnt_; }
     bool has_avx() const { return has_avx_; }
     bool has_avx2() const { return has_avx2_; }
+    bool has_avx512() const { return has_avx512_; }
     bool has_aesni() const { return has_aesni_; }
     bool has_non_stop_time_stamp_counter() const { return has_non_stop_time_stamp_counter_; }
     bool is_running_in_vm() const { return is_running_in_vm_; }
@@ -107,6 +109,7 @@ private:
     bool has_popcnt_;
     bool has_avx_;
     bool has_avx2_;
+    bool has_avx512_;
     bool has_aesni_;
     bool has_non_stop_time_stamp_counter_;
     bool is_running_in_vm_;
diff --git a/be/src/olap/rowset/segment_v2/bitshuffle_wrapper.cpp b/be/src/olap/rowset/segment_v2/bitshuffle_wrapper.cpp
index 7ad20f210c..7e569f92b2 100644
--- a/be/src/olap/rowset/segment_v2/bitshuffle_wrapper.cpp
+++ b/be/src/olap/rowset/segment_v2/bitshuffle_wrapper.cpp
@@ -34,6 +34,17 @@
 #undef bshuf_compress_lz4
 #undef bshuf_decompress_lz4
 
+// Include the bitshuffle header again, but this time importing the
+// AVX512-compiled symbols by defining some macros.
+#undef BITSHUFFLE_H
+#define bshuf_compress_lz4_bound bshuf_compress_lz4_bound_avx512
+#define bshuf_compress_lz4 bshuf_compress_lz4_avx512
+#define bshuf_decompress_lz4 bshuf_decompress_lz4_avx512
+#include <bitshuffle/bitshuffle.h> // NOLINT(*)
+#undef bshuf_compress_lz4_bound
+#undef bshuf_compress_lz4
+#undef bshuf_decompress_lz4
+
 using base::CPU;
 
 namespace doris {
@@ -54,7 +65,11 @@ decltype(&bshuf_decompress_lz4) g_bshuf_decompress_lz4;
 // the cost of a 'std::once' call.
 __attribute__((constructor)) void SelectBitshuffleFunctions() {
 #if (defined(__i386) || defined(__x86_64__))
-    if (CPU().has_avx2()) {
+    if (CPU().has_avx512()) {
+        g_bshuf_compress_lz4_bound = bshuf_compress_lz4_bound_avx512;
+        g_bshuf_compress_lz4 = bshuf_compress_lz4_avx512;
+        g_bshuf_decompress_lz4 = bshuf_decompress_lz4_avx512;
+    } else if (CPU().has_avx2()) {
         g_bshuf_compress_lz4_bound = bshuf_compress_lz4_bound_avx2;
         g_bshuf_compress_lz4 = bshuf_compress_lz4_avx2;
         g_bshuf_decompress_lz4 = bshuf_decompress_lz4_avx2;