diff --git a/cpp/src/arrow/pretty_print.cc b/cpp/src/arrow/pretty_print.cc
index e666ec70f94..53a69536816 100644
--- a/cpp/src/arrow/pretty_print.cc
+++ b/cpp/src/arrow/pretty_print.cc
@@ -391,7 +391,14 @@ class ArrayPrinter : public PrettyPrinter {
   }
 
   Status Print(const Array& array) {
-    RETURN_NOT_OK(VisitArrayInline(array, this));
+    if (array.device_type() != DeviceAllocationType::kCPU) {
+      // GH-43055: ideally we only copy start/end slices from non-CPU memory
+      // based on the window size that is being printed
+      ARROW_ASSIGN_OR_RAISE(auto array_cpu, array.CopyTo(default_cpu_memory_manager()));
+      RETURN_NOT_OK(VisitArrayInline(*array_cpu, this));
+    } else {
+      RETURN_NOT_OK(VisitArrayInline(array, this));
+    }
     Flush();
     return Status::OK();
   }
diff --git a/python/pyarrow/array.pxi b/python/pyarrow/array.pxi
index a9af1419ae4..daf4adc33e5 100644
--- a/python/pyarrow/array.pxi
+++ b/python/pyarrow/array.pxi
@@ -1241,6 +1241,9 @@ cdef class Array(_PandasConvertible):
         """
         Render a "pretty-printed" string representation of the Array.
 
+        Note: for data on a non-CPU device, the full array is copied to CPU
+        memory.
+
         Parameters
         ----------
         indent : int, default 2
@@ -1261,8 +1264,6 @@ cdef class Array(_PandasConvertible):
             If the array should be rendered as a single line of text
             or if each element should be on its own line.
         """
-        self._assert_cpu()
-
         cdef:
             c_string result
             PrettyPrintOptions options
diff --git a/python/pyarrow/tests/test_array.py b/python/pyarrow/tests/test_array.py
index 62a3eebb8c0..27e78e3396e 100644
--- a/python/pyarrow/tests/test_array.py
+++ b/python/pyarrow/tests/test_array.py
@@ -4035,6 +4035,8 @@ def test_non_cpu_array():
     assert arr.is_cpu is False
     assert len(arr) == 4
     assert arr.slice(2, 2).offset == 2
+    assert repr(arr)
+    assert str(arr)
 
     # TODO support DLPack for CUDA
     with pytest.raises(NotImplementedError):
@@ -4065,10 +4067,6 @@ def test_non_cpu_array():
         arr.get_total_buffer_size()
     with pytest.raises(NotImplementedError):
         [i for i in iter(arr)]
-    with pytest.raises(NotImplementedError):
-        repr(arr)
-    with pytest.raises(NotImplementedError):
-        str(arr)
     with pytest.raises(NotImplementedError):
         arr == arr2
     with pytest.raises(NotImplementedError):
diff --git a/python/pyarrow/tests/test_cuda.py b/python/pyarrow/tests/test_cuda.py
index 400db2643bd..5e833d5371d 100644
--- a/python/pyarrow/tests/test_cuda.py
+++ b/python/pyarrow/tests/test_cuda.py
@@ -944,3 +944,57 @@ def test_device_interface_batch_array():
     with pytest.raises(ValueError,
                        match="ArrowSchema describes non-struct type"):
         pa.RecordBatch._import_from_c_device(ptr_array, ptr_schema)
+
+
+def test_print_array():
+    batch = make_recordbatch(10)
+    cbuf = cuda.serialize_record_batch(batch, global_context)
+    cbatch = cuda.read_record_batch(cbuf, batch.schema)
+    arr = batch["f0"]
+    carr = cbatch["f0"]
+    assert str(carr) == str(arr)
+
+    batch = make_recordbatch(100)
+    cbuf = cuda.serialize_record_batch(batch, global_context)
+    cbatch = cuda.read_record_batch(cbuf, batch.schema)
+    arr = batch["f0"]
+    carr = cbatch["f0"]
+    assert str(carr) == str(arr)
+
+
+def make_chunked_array(n_elements_per_chunk, n_chunks):
+    arrs = []
+    carrs = []
+    for _ in range(n_chunks):
+        batch = make_recordbatch(n_elements_per_chunk)
+        cbuf = cuda.serialize_record_batch(batch, global_context)
+        cbatch = cuda.read_record_batch(cbuf, batch.schema)
+        arrs.append(batch["f0"])
+        carrs.append(cbatch["f0"])
+
+    return pa.chunked_array(arrs), pa.chunked_array(carrs)
+
+
+def test_print_chunked_array():
+    arr, carr = make_chunked_array(10, 3)
+    assert str(carr) == str(arr)
+
+    arr, carr = make_chunked_array(100, 20)
+    assert str(carr) == str(arr)
+
+
+def test_print_record_batch():
+    batch = make_recordbatch(10)
+    cbuf = cuda.serialize_record_batch(batch, global_context)
+    cbatch = cuda.read_record_batch(cbuf, batch.schema)
+    assert str(cbatch) == str(batch)
+
+    batch = make_recordbatch(100)
+    cbuf = cuda.serialize_record_batch(batch, global_context)
+    cbatch = cuda.read_record_batch(cbuf, batch.schema)
+    assert str(cbatch) == str(batch)
+
+
+def test_print_table():
+    _, table, _, ctable = make_table_cuda()
+    assert str(ctable) == str(table)