diff --git a/.github/workflows/windows_cuda.yml b/.github/workflows/windows_cuda.yml
index c6346d9c4e932..701df5e5642e6 100644
--- a/.github/workflows/windows_cuda.yml
+++ b/.github/workflows/windows_cuda.yml
@@ -148,7 +148,6 @@ jobs:
       DocUpdateNeeded: false
       ONNXRUNTIME_TEST_GPU_DEVICE_ID: '0'
       AZCOPY_AUTO_LOGIN_TYPE: MSI
-      AZCOPY_MSI_CLIENT_ID: 63b63039-6328-442f-954b-5a64d124e5b4
 
   test:
     name: Windows GPU CUDA CI Pipeline Test Job
@@ -260,4 +259,3 @@ jobs:
       DocUpdateNeeded: false
       ONNXRUNTIME_TEST_GPU_DEVICE_ID: '0'
       AZCOPY_AUTO_LOGIN_TYPE: MSI
-      AZCOPY_MSI_CLIENT_ID: 63b63039-6328-442f-954b-5a64d124e5b4
diff --git a/.github/workflows/windows_cuda_plugin.yml b/.github/workflows/windows_cuda_plugin.yml
index 538c1d783cd68..ac54c79bb8b57 100644
--- a/.github/workflows/windows_cuda_plugin.yml
+++ b/.github/workflows/windows_cuda_plugin.yml
@@ -47,8 +47,26 @@ jobs:
       - name: Download CUDA SDK v12.8
         working-directory: ${{ runner.temp }}
         run: |
-          azcopy.exe cp --recursive "https://lotusscus.blob.core.windows.net/models/cuda_sdk/v12.8" .
-          dir
+          $cudaSdkDir = Join-Path $env:RUNNER_TEMP "v12.8"
+          $localCudaSdkDir = "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.8"
+
+          if (Test-Path (Join-Path $cudaSdkDir "bin\nvcc.exe")) {
+            Write-Host "CUDA SDK already present at $cudaSdkDir"
+          } else {
+            azcopy.exe cp --recursive "https://lotusscus.blob.core.windows.net/models/cuda_sdk/v12.8" .
+            $azcopyExitCode = $lastExitCode
+            if (($azcopyExitCode -ne 0) -and (Test-Path (Join-Path $localCudaSdkDir "bin\nvcc.exe"))) {
+              Write-Host "AzCopy download failed; falling back to local CUDA SDK at $localCudaSdkDir"
+              New-Item -ItemType Directory -Path $cudaSdkDir -Force | Out-Null
+              Copy-Item -Path (Join-Path $localCudaSdkDir "*") -Destination $cudaSdkDir -Recurse -Force -ErrorAction Stop
+              $azcopyExitCode = 0
+            }
+            if ($azcopyExitCode -ne 0) {
+              exit $azcopyExitCode
+            }
+          }
+
+          dir $cudaSdkDir
         shell: pwsh
 
       - name: Add CUDA to PATH
@@ -107,18 +125,35 @@ jobs:
           Remove-Item -Path $outputDir -Include "*.obj" -Recurse
         shell: pwsh
 
+      - name: Stage CUDA runtime test dependencies
+        shell: pwsh
+        run: |
+          $cudaStageDir = Join-Path $env:RUNNER_TEMP "cuda-test-deps\v12.8"
+          $cudaBinDir = Join-Path $cudaStageDir "bin"
+          $cuptiDir = Join-Path $cudaStageDir "extras\CUPTI\lib64"
+          New-Item -ItemType Directory -Path $cudaBinDir -Force | Out-Null
+          New-Item -ItemType Directory -Path $cuptiDir -Force | Out-Null
+          Copy-Item -Path "$env:RUNNER_TEMP\v12.8\bin\*" -Destination $cudaBinDir -Recurse -Force -ErrorAction Stop
+          Copy-Item -Path "$env:RUNNER_TEMP\v12.8\extras\CUPTI\lib64\*" -Destination $cuptiDir -Recurse -Force -ErrorAction Stop
+
       - name: Upload build artifacts
         uses: actions/upload-artifact@v6
         with:
           name: cuda-plugin-build-artifacts
           path: ${{ runner.temp }}\build
+
+      - name: Upload CUDA runtime test dependencies
+        uses: actions/upload-artifact@v6
+        with:
+          name: cuda-plugin-cuda-runtime
+          path: ${{ runner.temp }}\cuda-test-deps
+          if-no-files-found: error
     env:
       DOTNET_SKIP_FIRST_TIME_EXPERIENCE: true
       setVcvars: true
       ALLOW_RELEASED_ONNX_OPSET_ONLY: '0'
       ONNXRUNTIME_TEST_GPU_DEVICE_ID: '0'
       AZCOPY_AUTO_LOGIN_TYPE: MSI
-      AZCOPY_MSI_CLIENT_ID: 63b63039-6328-442f-954b-5a64d124e5b4
 
   test:
     name: Windows CUDA Plugin EP Test
@@ -141,6 +176,12 @@ jobs:
           name: cuda-plugin-build-artifacts
           path: ${{ runner.temp }}\build
 
+      - name: Download CUDA runtime test dependencies
+        uses: actions/download-artifact@v7
+        with:
+          name: cuda-plugin-cuda-runtime
+          path: ${{ runner.temp }}\cuda-test-deps
+
       - uses: actions/setup-python@v6
         with:
           python-version: '3.14'
@@ -161,20 +202,13 @@ jobs:
         working-directory: ${{ github.workspace }}
         shell: cmd
 
-      - name: Download CUDA SDK v12.8
-        working-directory: ${{ runner.temp }}
-        run: |
-          azcopy.exe cp --recursive "https://lotusscus.blob.core.windows.net/models/cuda_sdk/v12.8" .
-          dir
-        shell: pwsh
-
       - name: Add CUDA to PATH
         shell: powershell
         run: |
           Write-Host "Adding CUDA to PATH"
-          Write-Host "CUDA Path: $env:RUNNER_TEMP\v12.8\bin"
-          Add-Content -Path $env:GITHUB_PATH -Value "$env:RUNNER_TEMP\v12.8\bin"
-          Add-Content -Path $env:GITHUB_PATH -Value "$env:RUNNER_TEMP\v12.8\extras\CUPTI\lib64"
+          Write-Host "CUDA Path: $env:RUNNER_TEMP\cuda-test-deps\v12.8\bin"
+          Add-Content -Path $env:GITHUB_PATH -Value "$env:RUNNER_TEMP\cuda-test-deps\v12.8\bin"
+          Add-Content -Path $env:GITHUB_PATH -Value "$env:RUNNER_TEMP\cuda-test-deps\v12.8\extras\CUPTI\lib64"
 
       - name: Set OnnxRuntimeBuildDirectory
         shell: pwsh
@@ -213,5 +247,3 @@ jobs:
       setVcvars: true
       ALLOW_RELEASED_ONNX_OPSET_ONLY: '0'
       ONNXRUNTIME_TEST_GPU_DEVICE_ID: '0'
-      AZCOPY_AUTO_LOGIN_TYPE: MSI
-      AZCOPY_MSI_CLIENT_ID: 63b63039-6328-442f-954b-5a64d124e5b4
diff --git a/.github/workflows/windows_gpu_doc_gen.yml b/.github/workflows/windows_gpu_doc_gen.yml
index 5e50a970875fc..e75d3a61dae19 100644
--- a/.github/workflows/windows_gpu_doc_gen.yml
+++ b/.github/workflows/windows_gpu_doc_gen.yml
@@ -44,7 +44,6 @@ jobs:
       setVcvars: true
       ALLOW_RELEASED_ONNX_OPSET_ONLY: '0'
       AZCOPY_AUTO_LOGIN_TYPE: MSI
-      AZCOPY_MSI_CLIENT_ID: 63b63039-6328-442f-954b-5a64d124e5b4
     runs-on: [
         "self-hosted",
         "1ES.Pool=onnxruntime-github-Win2022-GPU-A10",
diff --git a/.github/workflows/windows_openvino.yml b/.github/workflows/windows_openvino.yml
index 52581c7d0a5f5..ce535e2061ba3 100644
--- a/.github/workflows/windows_openvino.yml
+++ b/.github/workflows/windows_openvino.yml
@@ -26,7 +26,6 @@ jobs:
     timeout-minutes: 240
     env:
       AZCOPY_AUTO_LOGIN_TYPE: MSI
-      AZCOPY_MSI_CLIENT_ID: 63b63039-6328-442f-954b-5a64d124e5b4
       OnnxRuntimeBuildDirectory: ${{ github.workspace }}
       DOTNET_SKIP_FIRST_TIME_EXPERIENCE: true
       ALLOW_RELEASED_ONNX_OPSET_ONLY: '1'
diff --git a/.github/workflows/windows_qnn_x64.yml b/.github/workflows/windows_qnn_x64.yml
index 35c620ca6f650..fc190e1b7d2be 100644
--- a/.github/workflows/windows_qnn_x64.yml
+++ b/.github/workflows/windows_qnn_x64.yml
@@ -29,7 +29,6 @@ jobs:
         QnnLibKind: [shared_lib, static_lib]
     env:
       AZCOPY_AUTO_LOGIN_TYPE: MSI
-      AZCOPY_MSI_CLIENT_ID: 63b63039-6328-442f-954b-5a64d124e5b4
       DOTNET_SKIP_FIRST_TIME_EXPERIENCE: true
       ALLOW_RELEASED_ONNX_OPSET_ONLY: '1'
 
diff --git a/.github/workflows/windows_tensorrt.yml b/.github/workflows/windows_tensorrt.yml
index d5710795942d1..1087254f0258f 100644
--- a/.github/workflows/windows_tensorrt.yml
+++ b/.github/workflows/windows_tensorrt.yml
@@ -154,7 +154,6 @@ jobs:
       DocUpdateNeeded: false
       ONNXRUNTIME_TEST_GPU_DEVICE_ID: '0'
       AZCOPY_AUTO_LOGIN_TYPE: MSI
-      AZCOPY_MSI_CLIENT_ID: 63b63039-6328-442f-954b-5a64d124e5b4
 
   test:
     name: Windows GPU TensorRT CI Pipeline Test Job
@@ -265,4 +264,3 @@ jobs:
       DocUpdateNeeded: false
       ONNXRUNTIME_TEST_GPU_DEVICE_ID: '0'
       AZCOPY_AUTO_LOGIN_TYPE: MSI
-      AZCOPY_MSI_CLIENT_ID: 63b63039-6328-442f-954b-5a64d124e5b4
diff --git a/cmake/external/onnxruntime_external_deps.cmake b/cmake/external/onnxruntime_external_deps.cmake
index ed3b0aa8192a7..3514f0555327d 100644
--- a/cmake/external/onnxruntime_external_deps.cmake
+++ b/cmake/external/onnxruntime_external_deps.cmake
@@ -822,6 +822,22 @@ if (onnxruntime_USE_WEBGPU)
     endif()
 
     onnxruntime_fetchcontent_makeavailable(dawn)
+
+    # Pre-create the emdawnwebgpu include directory at configure time to prevent a parallel
+    # build race condition. Dawn's src/emdawnwebgpu/CMakeLists.txt registers add_custom_command
+    # rules that copy webgpu_glfw.h and webgpu_enum_class_bitmasks.h into
+    # gen/src/emdawnwebgpu/include/webgpu/ using cmake -E copy. With -j32 parallel builds
+    # those copy commands can be triggered from multiple Makefile targets simultaneously
+    # before DawnJSONGenerator has created the destination directory, causing:
+    #   Error copying file "…/webgpu_enum_class_bitmasks.h" to
+    #   "…/gen/src/emdawnwebgpu/include/webgpu/webgpu_enum_class_bitmasks.h"
+    # Creating the directory here (at configure time) ensures it already exists when the
+    # build starts, regardless of whether the dawn_parallel_build_fix.patch was applied to
+    # the cached Dawn source on this runner.
+    FetchContent_GetProperties(dawn)
+    if(dawn_POPULATED)
+      file(MAKE_DIRECTORY "${dawn_BINARY_DIR}/gen/src/emdawnwebgpu/include/webgpu")
+    endif()
   endif()
 
   if (NOT CMAKE_SYSTEM_NAME STREQUAL "Emscripten")