diff --git a/cirq-core/cirq/circuits/circuit.py b/cirq-core/cirq/circuits/circuit.py
index b60e1e0b080..5b3d64c9e55 100644
--- a/cirq-core/cirq/circuits/circuit.py
+++ b/cirq-core/cirq/circuits/circuit.py
@@ -67,6 +67,8 @@
 CIRCUIT_TYPE = TypeVar('CIRCUIT_TYPE', bound='AbstractCircuit')
 INT_TYPE = Union[int, np.integer]
 
+_DEVICE_DEP_MESSAGE = 'Attaching devices to circuits will no longer be supported.'
+
 
 class Alignment(enum.Enum):
     # Stop when left ends are lined up.
diff --git a/cirq-google/cirq_google/api/v1/programs.py b/cirq-google/cirq_google/api/v1/programs.py
index 0b5fd34c44d..6ff037007b8 100644
--- a/cirq-google/cirq_google/api/v1/programs.py
+++ b/cirq-google/cirq_google/api/v1/programs.py
@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import json
-from typing import Any, cast, Dict, Iterable, Optional, Sequence, Tuple, TYPE_CHECKING, Iterator
+from typing import Any, cast, Dict, Optional, Sequence, Tuple, TYPE_CHECKING, Iterator
 import numpy as np
 import sympy
 
@@ -166,16 +166,26 @@ def circuit_as_schedule_to_protos(circuit: cirq.Circuit) -> Iterator[operations_
         yield op_proto
 
 
-def circuit_from_schedule_from_protos(
-    device: 'cirq_google.XmonDevice',
-    ops: Iterable[operations_pb2.Operation],
-) -> cirq.Circuit:
-    """Convert protos into a Circuit for the given device."""
+@cirq._compat.deprecated_parameter(
+    deadline='v0.15',
+    fix=cirq.circuits.circuit._DEVICE_DEP_MESSAGE,
+    parameter_desc='device',
+    match=lambda args, kwargs: 'device' in kwargs or len(args) > 1,
+)
+def circuit_from_schedule_from_protos(*args) -> cirq.Circuit:
+    """Convert protos into a Circuit."""
+    if len(args) == 2:
+        device, ops = args[0], args[1]
+    else:
+        ops = args[0]
     result = []
     for op in ops:
         xmon_op = xmon_op_from_proto(op)
         result.append(xmon_op)
-    return cirq.Circuit(result, device=device)
+    ret = cirq.Circuit(result)
+    if len(args) == 2:
+        ret._device = device
+    return ret
 
 
 def pack_results(measurements: Sequence[Tuple[str, np.ndarray]]) -> bytes:
diff --git a/cirq-google/cirq_google/api/v1/programs_test.py b/cirq-google/cirq_google/api/v1/programs_test.py
index 2efabbb50b3..4a1078c525e 100644
--- a/cirq-google/cirq_google/api/v1/programs_test.py
+++ b/cirq-google/cirq_google/api/v1/programs_test.py
@@ -11,7 +11,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-
 import numpy as np
 import pytest
 import sympy
@@ -32,14 +31,29 @@ def test_protobuf_round_trip():
     circuit = cirq.Circuit(
         [cirq.X(q) ** 0.5 for q in device.qubits],
         [cirq.CZ(q, q2) for q in [cirq.GridQubit(0, 0)] for q2 in device.neighbors_of(q)],
-        device=device,
     )
 
     protos = list(programs.circuit_as_schedule_to_protos(circuit))
-    s2 = programs.circuit_from_schedule_from_protos(device, protos)
+    s2 = programs.circuit_from_schedule_from_protos(protos)
     assert s2 == circuit
 
 
+def test_protobuf_round_trip_device_deprecated():
+    device = cg.Foxtail
+    circuit = cirq.Circuit(
+        [cirq.X(q) ** 0.5 for q in device.qubits],
+        [cirq.CZ(q, q2) for q in [cirq.GridQubit(0, 0)] for q2 in device.neighbors_of(q)],
+    )
+    circuit._device = device
+
+    protos = list(programs.circuit_as_schedule_to_protos(circuit))
+    with cirq.testing.assert_deprecated(
+        cirq.circuits.circuit._DEVICE_DEP_MESSAGE, deadline='v0.15'
+    ):
+        s2 = programs.circuit_from_schedule_from_protos(device, protos)
+        assert s2 == circuit
+
+
 def make_bytes(s: str) -> bytes:
     """Helper function to convert a string of digits into packed bytes.
 
diff --git a/cirq-google/cirq_google/engine/engine.py b/cirq-google/cirq_google/engine/engine.py
index 2beec954b0d..68a26278cc2 100644
--- a/cirq-google/cirq_google/engine/engine.py
+++ b/cirq-google/cirq_google/engine/engine.py
@@ -634,7 +634,6 @@ def _serialize_program(
     ) -> any_pb2.Any:
         if not isinstance(program, cirq.AbstractCircuit):
             raise TypeError(f'Unrecognized program type: {type(program)}')
-        program.device.validate_circuit(program)
 
         if self.context.proto_version == ProtoVersion.V2:
             program = gate_set.serialize(program)
diff --git a/cirq-google/cirq_google/engine/engine_test.py b/cirq-google/cirq_google/engine/engine_test.py
index cd848da175e..ecc4e4cdc46 100644
--- a/cirq-google/cirq_google/engine/engine_test.py
+++ b/cirq-google/cirq_google/engine/engine_test.py
@@ -392,21 +392,8 @@ def test_run_circuit(client):
     client.get_job_result.called_once_with()
 
 
-def test_circuit_device_validation_fails():
-    circuit = cirq.Circuit(device=cg.Foxtail)
-
-    # Purposefully create an invalid Circuit by fiddling with internal bits.
-    # This simulates a failure in the incremental checks.
-    circuit._moments.append(cirq.Moment([cirq.Z(cirq.NamedQubit("dorothy"))]))
-    engine = cg.Engine(project_id='project-id')
-    with pytest.raises(ValueError, match='Unsupported qubit type'):
-        engine.run_sweep(program=circuit, gate_set=cg.XMON)
-    with pytest.raises(ValueError, match='Unsupported qubit type'):
-        engine.create_program(circuit, gate_set=cg.XMON)
-
-
 def test_no_gate_set():
-    circuit = cirq.Circuit(device=cg.Sycamore)
+    circuit = cirq.Circuit()
     engine = cg.Engine(project_id='project-id')
     with pytest.raises(ValueError, match='No gate set'):
         engine.run(program=circuit)
diff --git a/cirq-google/cirq_google/optimizers/optimize_for_sycamore.py b/cirq-google/cirq_google/optimizers/optimize_for_sycamore.py
index 032be3c254f..3165ae0ee64 100644
--- a/cirq-google/cirq_google/optimizers/optimize_for_sycamore.py
+++ b/cirq-google/cirq_google/optimizers/optimize_for_sycamore.py
@@ -110,6 +110,12 @@ def _gate_product_tabulation_cached(
         raise NotImplementedError(f"Two qubit gate tabulation not supported for {optimizer_type}")
 
 
+@cirq._compat.deprecated_parameter(
+    deadline='v0.15',
+    fix=cirq.circuits.circuit._DEVICE_DEP_MESSAGE,
+    parameter_desc='new_device',
+    match=lambda args, kwargs: 'new_device' in kwargs,
+)
 def optimized_for_sycamore(
     circuit: cirq.Circuit,
     *,
@@ -161,8 +167,9 @@ def optimized_for_sycamore(
     for optimizer in opts:
         optimizer(copy)
 
-    return cirq.Circuit(
+    ret = cirq.Circuit(
         (op.transform_qubits(qubit_map) for op in copy.all_operations()),
         strategy=cirq.InsertStrategy.EARLIEST,
-        device=new_device or copy.device,
     )
+    ret._device = new_device or copy._device
+    return ret
diff --git a/cirq-google/cirq_google/optimizers/optimize_for_sycamore_test.py b/cirq-google/cirq_google/optimizers/optimize_for_sycamore_test.py
index cc383ca145f..9750ace13d2 100644
--- a/cirq-google/cirq_google/optimizers/optimize_for_sycamore_test.py
+++ b/cirq-google/cirq_google/optimizers/optimize_for_sycamore_test.py
@@ -116,3 +116,14 @@ def test_one_q_matrix_gate():
             assert cg.SYC_GATESET.is_supported_operation(op)
             # single qubit gates shared between gatesets, so:
             assert cg.SQRT_ISWAP_GATESET.is_supported_operation(op)
+
+
+def test_assert_new_device_deprecated():
+    u = cirq.testing.random_special_unitary(2)
+    q = cirq.LineQubit(0)
+    circuit0 = cirq.Circuit(cirq.MatrixGate(u).on(q))
+    _ = cg.optimized_for_sycamore(circuit0, optimizer_type='sqrt_iswap')
+    with cirq.testing.assert_deprecated(
+        cirq.circuits.circuit._DEVICE_DEP_MESSAGE, deadline='v0.15'
+    ):
+        _ = cg.optimized_for_sycamore(circuit0, optimizer_type='sqrt_iswap', new_device=cg.Foxtail)
diff --git a/cirq-google/cirq_google/optimizers/optimize_for_xmon.py b/cirq-google/cirq_google/optimizers/optimize_for_xmon.py
index 3cb0bc1141c..5bd9430622a 100644
--- a/cirq-google/cirq_google/optimizers/optimize_for_xmon.py
+++ b/cirq-google/cirq_google/optimizers/optimize_for_xmon.py
@@ -21,17 +21,19 @@
     import cirq_google
 
 
+@cirq._compat.deprecated_parameter(
+    deadline='v0.15',
+    fix=cirq.circuits.circuit._DEVICE_DEP_MESSAGE,
+    parameter_desc='new_device',
+    match=lambda args, kwargs: 'new_device' in kwargs,
+)
 def optimized_for_xmon(
     circuit: cirq.Circuit,
     new_device: Optional['cirq_google.XmonDevice'] = None,
     qubit_map: Callable[[cirq.Qid], cirq.GridQubit] = lambda e: cast(cirq.GridQubit, e),
     allow_partial_czs: bool = False,
 ) -> cirq.Circuit:
-    if allow_partial_czs:
-        return optimized_for_sycamore(
-            circuit, new_device=new_device, qubit_map=qubit_map, optimizer_type='xmon_partial_cz'
-        )
-    else:
-        return optimized_for_sycamore(
-            circuit, new_device=new_device, qubit_map=qubit_map, optimizer_type='xmon'
-        )
+    optimizer_type = 'xmon_partial_cz' if allow_partial_czs else 'xmon'
+    ret = optimized_for_sycamore(circuit, qubit_map=qubit_map, optimizer_type=optimizer_type)
+    ret._device = new_device or circuit._device
+    return ret
diff --git a/cirq-google/cirq_google/optimizers/optimize_for_xmon_test.py b/cirq-google/cirq_google/optimizers/optimize_for_xmon_test.py
index 1507a8870f9..eed11197768 100644
--- a/cirq-google/cirq_google/optimizers/optimize_for_xmon_test.py
+++ b/cirq-google/cirq_google/optimizers/optimize_for_xmon_test.py
@@ -11,7 +11,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-
 import pytest
 
 import cirq
@@ -57,38 +56,13 @@ def test_ccz():
     assert_circuits_with_terminal_measurements_are_equivalent(before, after, atol=1e-4)
 
 
-def test_adjacent_cz_get_split_apart():
-    before = cirq.Circuit(
-        [
-            cirq.Moment(
-                [
-                    cirq.CZ(cirq.GridQubit(0, 0), cirq.GridQubit(0, 1)),
-                    cirq.CZ(cirq.GridQubit(1, 0), cirq.GridQubit(1, 1)),
-                ]
-            )
-        ]
-    )
-
-    after = cg.optimized_for_xmon(before, new_device=cg.Foxtail)
-
-    assert after == cirq.Circuit(
-        [
-            cirq.Moment([cirq.CZ(cirq.GridQubit(0, 0), cirq.GridQubit(0, 1))]),
-            cirq.Moment([cirq.CZ(cirq.GridQubit(1, 0), cirq.GridQubit(1, 1))]),
-        ],
-        device=cg.Foxtail,
-    )
-
-
 def test_remap_qubits():
     before = cirq.Circuit([cirq.Moment([cirq.CZ(cirq.LineQubit(0), cirq.LineQubit(1))])])
 
-    after = cg.optimized_for_xmon(
-        before, new_device=cg.Foxtail, qubit_map=lambda q: cirq.GridQubit(q.x, 0)
-    )
+    after = cg.optimized_for_xmon(before, qubit_map=lambda q: cirq.GridQubit(q.x, 0))
 
     assert after == cirq.Circuit(
-        [cirq.Moment([cirq.CZ(cirq.GridQubit(0, 0), cirq.GridQubit(1, 0))])], device=cg.Foxtail
+        [cirq.Moment([cirq.CZ(cirq.GridQubit(0, 0), cirq.GridQubit(1, 0))])]
     )
 
 
diff --git a/cirq-google/cirq_google/serialization/gate_sets_test.py b/cirq-google/cirq_google/serialization/gate_sets_test.py
index f7d3b0f865c..8899f599b61 100644
--- a/cirq-google/cirq_google/serialization/gate_sets_test.py
+++ b/cirq-google/cirq_google/serialization/gate_sets_test.py
@@ -11,8 +11,9 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-
+import os
 from typing import Dict
+from unittest import mock
 import numpy as np
 import pytest
 import sympy
@@ -355,7 +356,8 @@ def test_deserialize_circuit():
     assert cg.XMON.deserialize(serialized) == circuit
 
 
-def test_deserialize_schedule():
+@mock.patch.dict(os.environ, clear='CIRQ_TESTING')
+def test_deserialize_schedule_device_deprecated():
     q0 = cirq.GridQubit(4, 4)
     q1 = cirq.GridQubit(4, 5)
     circuit = cirq.Circuit(
@@ -384,6 +386,33 @@ def test_deserialize_schedule():
     assert cg.XMON.deserialize(serialized, cg.Bristlecone) == circuit
 
 
+def test_deserialize_schedule():
+    q0 = cirq.GridQubit(4, 4)
+    q1 = cirq.GridQubit(4, 5)
+    circuit = cirq.Circuit(cirq.CZ(q0, q1), cirq.X(q0), cirq.Z(q1), cirq.measure(q0, key='a'))
+    serialized = v2.program_pb2.Program(
+        language=v2.program_pb2.Language(gate_set='xmon'),
+        schedule=v2.program_pb2.Schedule(
+            scheduled_operations=[
+                v2.program_pb2.ScheduledOperation(
+                    operation=cg.XMON.serialize_op(cirq.CZ(q0, q1)), start_time_picos=0
+                ),
+                v2.program_pb2.ScheduledOperation(
+                    operation=cg.XMON.serialize_op(cirq.X(q0)), start_time_picos=200000
+                ),
+                v2.program_pb2.ScheduledOperation(
+                    operation=cg.XMON.serialize_op(cirq.Z(q1)), start_time_picos=200000
+                ),
+                v2.program_pb2.ScheduledOperation(
+                    operation=cg.XMON.serialize_op(cirq.measure(q0, key='a')),
+                    start_time_picos=400000,
+                ),
+            ]
+        ),
+    )
+    assert cg.XMON.deserialize(serialized) == circuit
+
+
 def test_serialize_deserialize_syc():
     proto = op_proto({'gate': {'id': 'syc'}, 'args': {}, 'qubits': [{'id': '1_2'}, {'id': '1_3'}]})
 
diff --git a/cirq-google/cirq_google/serialization/serializable_gate_set.py b/cirq-google/cirq_google/serialization/serializable_gate_set.py
index 6a9497e2cb1..985098b4eaa 100644
--- a/cirq-google/cirq_google/serialization/serializable_gate_set.py
+++ b/cirq-google/cirq_google/serialization/serializable_gate_set.py
@@ -273,6 +273,12 @@ def serialize_circuit_op(
                 return proto_msg
         raise ValueError(f'Cannot serialize CircuitOperation {op!r}')
 
+    @cirq._compat.deprecated_parameter(
+        deadline='v0.15',
+        fix=cirq.circuits.circuit._DEVICE_DEP_MESSAGE,
+        parameter_desc='device',
+        match=lambda args, kwargs: 'device' in kwargs or len(args) > 2,
+    )
     def deserialize(
         self, proto: v2.program_pb2.Program, device: Optional[cirq.Device] = None
     ) -> cirq.Circuit:
@@ -322,10 +328,10 @@ def deserialize(
                 constants=proto.constants,
                 deserialized_constants=deserialized_constants,
             )
-            return circuit if device is None else circuit.with_device(device)
+            if device is not None:
+                circuit._device = device  # coverage: ignore
+            return circuit
         if which == 'schedule':
-            if device is None:
-                raise ValueError('Deserializing schedule requires a device but None was given.')
             return self._deserialize_schedule(
                 proto.schedule, device, arg_function_language=proto.language.arg_function_language
             )
@@ -497,7 +503,7 @@ def _deserialize_circuit(
     def _deserialize_schedule(
         self,
         schedule_proto: v2.program_pb2.Schedule,
-        device: cirq.Device,
+        device: Optional[cirq.Device],
         *,
         arg_function_language: str,
     ) -> cirq.Circuit:
@@ -510,4 +516,8 @@ def _deserialize_schedule(
                     scheduled_op_proto.operation, arg_function_language=arg_function_language
                 )
             )
-        return cirq.Circuit(result, device=device)
+        ret = cirq.Circuit(result)
+        if device is None:
+            device = cirq.UNCONSTRAINED_DEVICE
+        ret._device = device
+        return ret
diff --git a/cirq-google/cirq_google/serialization/serializable_gate_set_test.py b/cirq-google/cirq_google/serialization/serializable_gate_set_test.py
index 8d7d9689897..a55a50bbca7 100644
--- a/cirq-google/cirq_google/serialization/serializable_gate_set_test.py
+++ b/cirq-google/cirq_google/serialization/serializable_gate_set_test.py
@@ -387,29 +387,6 @@ def test_serialize_unrecognized():
         MY_GATE_SET.serialize("not quite right")
 
 
-def test_serialize_deserialize_schedule_no_device():
-    q0 = cirq.GridQubit(1, 1)
-    q1 = cirq.GridQubit(1, 2)
-    proto = v2.program_pb2.Program(
-        language=v2.program_pb2.Language(arg_function_language='', gate_set='my_gate_set'),
-        schedule=v2.program_pb2.Schedule(
-            scheduled_operations=[
-                v2.program_pb2.ScheduledOperation(
-                    operation=X_SERIALIZER.to_proto(cirq.X(q0)), start_time_picos=0
-                ),
-                v2.program_pb2.ScheduledOperation(
-                    operation=X_SERIALIZER.to_proto(cirq.X(q1)), start_time_picos=200000
-                ),
-                v2.program_pb2.ScheduledOperation(
-                    operation=X_SERIALIZER.to_proto(cirq.X(q0)), start_time_picos=400000
-                ),
-            ]
-        ),
-    )
-    with pytest.raises(ValueError):
-        MY_GATE_SET.deserialize(proto)
-
-
 def test_serialize_deserialize_op():
     q0 = cirq.GridQubit(1, 1)
     proto = op_proto(
@@ -769,15 +746,6 @@ def test_deserialize_invalid_gate_set():
         MY_GATE_SET.deserialize(proto)
 
 
-def test_deserialize_schedule_missing_device():
-    proto = v2.program_pb2.Program(
-        language=v2.program_pb2.Language(gate_set='my_gate_set'),
-        schedule=v2.program_pb2.Schedule(scheduled_operations=[]),
-    )
-    with pytest.raises(ValueError, match='device'):
-        MY_GATE_SET.deserialize(proto)
-
-
 def test_deserialize_no_operation():
     proto = v2.program_pb2.Program(
         language=v2.program_pb2.Language(gate_set='my_gate_set'),
diff --git a/examples/advanced/quantum_volume.py b/examples/advanced/quantum_volume.py
index 9924eed1ec0..b72f42b79fc 100644
--- a/examples/advanced/quantum_volume.py
+++ b/examples/advanced/quantum_volume.py
@@ -40,7 +40,7 @@ def main(*, num_qubits: int, depth: int, num_circuits: int, seed: int, routes: i
     Returns: Pass-through from calculate_quantum_volume.
     """
     device = cirq_google.Bristlecone
-    compiler = lambda circuit: cirq_google.optimized_for_xmon(circuit=circuit, new_device=device)
+    compiler = lambda circuit: cirq_google.optimized_for_xmon(circuit=circuit)
     noisy = cirq.DensityMatrixSimulator(
         noise=cirq.ConstantQubitNoiseModel(qubit_noise_gate=cirq.DepolarizingChannel(p=0.005))
     )
diff --git a/examples/place_on_bristlecone.py b/examples/place_on_bristlecone.py
index 957609197a7..e25b36f0349 100644
--- a/examples/place_on_bristlecone.py
+++ b/examples/place_on_bristlecone.py
@@ -79,9 +79,7 @@ def main():
 
     print()
     print("Xmon circuit:")
-    translated = cirq_google.optimized_for_xmon(
-        circuit=circuit, new_device=cirq_google.Bristlecone, qubit_map=lambda q: line[q.x]
-    )
+    translated = cirq_google.optimized_for_xmon(circuit=circuit, qubit_map=lambda q: line[q.x])
     print(translated)