diff --git a/src/gpuhunt/providers/hotaisle.py b/src/gpuhunt/providers/hotaisle.py
index 22c4893..34a768a 100644
--- a/src/gpuhunt/providers/hotaisle.py
+++ b/src/gpuhunt/providers/hotaisle.py
@@ -1,12 +1,12 @@
 import logging
 import os
-from typing import Optional
+from typing import Optional, TypedDict, cast
 
 import requests
 from requests import Response
 
 from gpuhunt._internal.constraints import find_accelerators
-from gpuhunt._internal.models import AcceleratorVendor, QueryFilter, RawCatalogItem
+from gpuhunt._internal.models import AcceleratorVendor, JSONObject, QueryFilter, RawCatalogItem
 from gpuhunt.providers import AbstractProvider
 
 logger = logging.getLogger(__name__)
@@ -53,6 +53,10 @@ def _make_request(self, method: str, url: str) -> Response:
         return response
 
 
+class HotAisleCatalogItemProviderData(TypedDict):
+    vm_specs: JSONObject
+
+
 def get_gpu_memory(gpu_name: str) -> Optional[float]:
     if accelerators := find_accelerators(names=[gpu_name], vendors=[AcceleratorVendor.AMD]):
         return float(accelerators[0].memory)
@@ -96,6 +100,14 @@ def convert_response_to_raw_catalog_items(response: Response) -> list[RawCatalog
             gpu_vendor=gpu_vendor,
             spot=False,
             disk_size=disk_gb,
+            provider_data=cast(
+                JSONObject,
+                HotAisleCatalogItemProviderData(
+                    # The specs object may duplicate some RawCatalogItem fields, but we store it in
+                    # full because we need to pass it back to the API when creating VMs.
+                    vm_specs=specs,
+                ),
+            ),
         )
         offers.append(offer)