diff --git a/pyproject.toml b/pyproject.toml
index a5e398e0f..257ca05b3 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -47,6 +47,7 @@ dependencies = [
     "exceptiongroup; python_version<'3.11'",
     "natsort",
     "packaging>=20.0",
+    "pint",
     # array-api-compat 1.5 has https://github.com/scverse/anndata/issues/1410
     "array_api_compat>1.4,!=1.5",
 ]
diff --git a/src/anndata/__init__.py b/src/anndata/__init__.py
index c2006cd72..0b8ea96dc 100644
--- a/src/anndata/__init__.py
+++ b/src/anndata/__init__.py
@@ -21,7 +21,7 @@
     # Backport package for exception groups
     import exceptiongroup  # noqa: F401
 
-from ._core.anndata import AnnData
+from ._core.anndata import AnnData, units
 from ._core.merge import concat
 from ._core.raw import Raw
 from ._io import (
@@ -81,4 +81,5 @@ def read(*args, **kwargs):
     "ExperimentalFeatureWarning",
     "experimental",
     "settings",
+    "units",
 ]
diff --git a/src/anndata/_core/anndata.py b/src/anndata/_core/anndata.py
index 9184fa9c9..c5248ec7d 100644
--- a/src/anndata/_core/anndata.py
+++ b/src/anndata/_core/anndata.py
@@ -25,6 +25,7 @@
 from natsort import natsorted
 from numpy import ma
 from pandas.api.types import infer_dtype
+from pint import UnitRegistry
 from scipy import sparse
 from scipy.sparse import issparse
 
@@ -66,6 +67,8 @@
 if TYPE_CHECKING:
     from os import PathLike
 
+units = UnitRegistry()
+
 
 class StorageType(Enum):
     Array = np.ndarray
diff --git a/src/anndata/_io/specs/methods.py b/src/anndata/_io/specs/methods.py
index 2822154da..a4cb58f35 100644
--- a/src/anndata/_io/specs/methods.py
+++ b/src/anndata/_io/specs/methods.py
@@ -13,7 +13,7 @@
 from scipy import sparse
 
 import anndata as ad
-from anndata import AnnData, Raw
+from anndata import AnnData, Raw, units
 from anndata._core import views
 from anndata._core.index import _normalize_indices
 from anndata._core.merge import intersect_keys
@@ -293,6 +293,27 @@ def write_raw(f, k, raw, _writer, dataset_kwargs=MappingProxyType({})):
     _writer.write_elem(g, "varm", dict(raw.varm), dataset_kwargs=dataset_kwargs)
 
 
+########
+# Pint #
+########
+
+
+@_REGISTRY.register_read(H5Group, IOSpec("pint.Quantity", "0.1.0"))
+@_REGISTRY.register_read(ZarrGroup, IOSpec("pint.Quantity", "0.1.0"))
+def read_quantity(elem, _reader):
+    v_magnitude = _reader.read_elem(elem["magnitude"])
+    v_units = units[_reader.read_elem(elem["units"])]
+    return v_magnitude * v_units
+
+
+@_REGISTRY.register_write(H5Group, units.Quantity, IOSpec("pint.Quantity", "0.1.0"))
+@_REGISTRY.register_write(ZarrGroup, units.Quantity, IOSpec("pint.Quantity", "0.1.0"))
+def write_quantity(f, k, v, _writer, dataset_kwargs=MappingProxyType({})):
+    g = f.require_group(k)
+    _writer.write_elem(g, "magnitude", v.magnitude, dataset_kwargs=dataset_kwargs)
+    _writer.write_elem(g, "units", str(v.units), dataset_kwargs=dataset_kwargs)
+
+
 ############
 # Mappings #
 ############
diff --git a/src/anndata/tests/test_readwrite.py b/src/anndata/tests/test_readwrite.py
index 1c2630f98..5b7954fb5 100644
--- a/src/anndata/tests/test_readwrite.py
+++ b/src/anndata/tests/test_readwrite.py
@@ -821,3 +821,17 @@ def test_io_dtype(tmp_path, diskfmt, dtype):
     curr = read(pth)
 
     assert curr.X.dtype == dtype
+
+
+def test_readwrite_units(read, write, name, tmp_path):
+    X_arr = np.array(X_list)
+    adata = ad.AnnData(
+        X=X_arr,
+        uns={"size": 100 * ad.units["um"]},
+        obsm={"X_spatial": X_arr * ad.units.mm},
+    )
+    write(tmp_path / name, adata)
+    ad_read = read(tmp_path / name)
+
+    assert adata.uns["spot_size"] == ad_read.uns["spot_size"]
+    assert (adata.obsm["X_spatial_units"] == ad_read.obsm["X_spatial_units"]).all()