diff --git a/pcsx2/GS/Renderers/Metal/GSTextureMTL.h b/pcsx2/GS/Renderers/Metal/GSTextureMTL.h
index b79826fded..9574e938bb 100644
--- a/pcsx2/GS/Renderers/Metal/GSTextureMTL.h
+++ b/pcsx2/GS/Renderers/Metal/GSTextureMTL.h
@@ -93,9 +93,6 @@ public:
 	void Flush() override;
 
 private:
-	// TODO: Is there an optimal transfer pitch alignment for Metal?
-	static constexpr u32 PITCH_ALIGNMENT = 32;
-
 	GSDownloadTextureMTL(GSDeviceMTL* dev, MRCOwned<id<MTLBuffer>> buffer, u32 width, u32 height, GSTexture::Format format);
 
 	GSDeviceMTL* m_dev;
diff --git a/pcsx2/GS/Renderers/Metal/GSTextureMTL.mm b/pcsx2/GS/Renderers/Metal/GSTextureMTL.mm
index 7c84dc0943..66882ceb09 100644
--- a/pcsx2/GS/Renderers/Metal/GSTextureMTL.mm
+++ b/pcsx2/GS/Renderers/Metal/GSTextureMTL.mm
@@ -17,10 +17,14 @@
 #include "GS/Renderers/Metal/GSTextureMTL.h"
 #include "GS/Renderers/Metal/GSDeviceMTL.h"
 #include "GS/GSPerfMon.h"
+#include "common/Align.h"
 #include "common/Console.h"
 
 #ifdef __APPLE__
 
+// Uploads/downloads need 32-byte alignment for AVX2.
+static constexpr u32 PITCH_ALIGNMENT = 32;
+
 GSTextureMTL::GSTextureMTL(GSDeviceMTL* dev, MRCOwned<id<MTLTexture>> texture, Type type, Format format)
 	: m_dev(dev)
 	, m_texture(std::move(texture))
@@ -119,7 +123,7 @@ bool GSTextureMTL::Map(GSMap& m, const GSVector4i* _r, int layer)
 	GSVector4i r = _r ? *_r : GSVector4i(0, 0, m_size.x, m_size.y);
 	u32 block_size = GetCompressedBlockSize();
 	u32 blocks_wide = (r.width() + block_size - 1) / block_size;
-	m.pitch = blocks_wide * GetCompressedBytesPerBlock();
+	m.pitch = Common::AlignUpPow2(blocks_wide * GetCompressedBytesPerBlock(), PITCH_ALIGNMENT);
 	if (void* buffer = MapWithPitch(r, m.pitch, layer))
 	{
 		m.bits = static_cast<u8*>(buffer);