diff --git a/doc/source/aws.rst b/doc/source/aws.rst
index d428ccd24..1dae2fd1d 100644
--- a/doc/source/aws.rst
+++ b/doc/source/aws.rst
@@ -141,7 +141,7 @@ Selecting the ``aws`` driver adds the following options to the
       .. code-block:: yaml
 
          max-resources:
-           'L-43DA4232': 224
+           'L-43DA4232': 448
 
       See `instance quotas`_ for more information.
 
@@ -498,7 +498,7 @@ Selecting the ``aws`` driver adds the following options to the
          .. code-block:: yaml
 
             max-resources:
-              'L-43DA4232': 224
+              'L-43DA4232': 448
 
          See `instance quotas`_ for more information.
 
diff --git a/doc/source/configuration.rst b/doc/source/configuration.rst
index 25c57b78a..f0dec9ab4 100644
--- a/doc/source/configuration.rst
+++ b/doc/source/configuration.rst
@@ -557,7 +557,7 @@ Options
           max-servers: 10
           max-cores: 200
           max-ram: 16565
-          'L-43DA4232': 224
+          'L-43DA4232': 448
 
    Each entry is a dictionary with the following keys.  Any other keys
    are interpreted as driver-specific resource limits (otherwise
diff --git a/nodepool/driver/aws/adapter.py b/nodepool/driver/aws/adapter.py
index 318961109..bce81df6a 100644
--- a/nodepool/driver/aws/adapter.py
+++ b/nodepool/driver/aws/adapter.py
@@ -629,15 +629,18 @@ class AwsAdapter(statemachine.Adapter):
     def _getQuotaForInstanceType(self, instance_type):
         itype = self._getInstanceType(instance_type)
         cores = itype['InstanceTypes'][0]['VCpuInfo']['DefaultCores']
+        vcpus = itype['InstanceTypes'][0]['VCpuInfo']['DefaultVCpus']
         ram = itype['InstanceTypes'][0]['MemoryInfo']['SizeInMiB']
         code = self._getQuotaCodeForInstanceType(instance_type)
-        # We include cores twice: one to match the overall cores quota
-        # (which may be set as a tenant resource limit), and a second
-        # time as the specific AWS quota code which in for a specific
-        # instance type.
+        # We include cores to match the overall cores quota (which may
+        # be set as a tenant resource limit), and include vCPUs for the
+        # specific AWS quota code which in for a specific instance
+        # type. With two threads per core, the vCPU number is
+        # typically twice the number of cores. AWS service quotas are
+        # implemented in terms of vCPUs.
         args = dict(cores=cores, ram=ram, instances=1)
         if code:
-            args[code] = cores
+            args[code] = vcpus
         return QuotaInformation(**args)
 
     # This method is wrapped with an LRU cache in the constructor.
diff --git a/nodepool/tests/fixtures/aws/aws-limits.yaml b/nodepool/tests/fixtures/aws/aws-limits.yaml
index 089d068cc..2d00d00d2 100644
--- a/nodepool/tests/fixtures/aws/aws-limits.yaml
+++ b/nodepool/tests/fixtures/aws/aws-limits.yaml
@@ -11,7 +11,7 @@ zookeeper-tls:
 tenant-resource-limits:
   - tenant-name: tenant-1
     max-cores: 1024
-    'L-43DA4232': 224  # high mem cores
+    'L-43DA4232': 448  # high mem vCPUs
 
 labels:
   - name: standard
@@ -34,7 +34,7 @@ providers:
           key1: value1
           key2: value2
         max-resources:
-          'L-1216C47A': 1  # standard cores
+          'L-1216C47A': 2  # standard vCPUs
         labels:
           - name: standard
             cloud-image: ubuntu1404
diff --git a/nodepool/tests/unit/test_driver_aws.py b/nodepool/tests/unit/test_driver_aws.py
index e17e9a832..d88bb3551 100644
--- a/nodepool/tests/unit/test_driver_aws.py
+++ b/nodepool/tests/unit/test_driver_aws.py
@@ -240,8 +240,8 @@ class TestDriverAws(tests.DBTestCase):
             self.waitForNodeDeletion(node)
 
     @aws_quotas({
-        'L-1216C47A': 1,
-        'L-43DA4232': 224,
+        'L-1216C47A': 2,
+        'L-43DA4232': 448,
     })
     def test_aws_multi_quota(self):
         # Test multiple instance type quotas (standard and high-mem)