diff --git a/python/sglang/srt/entrypoints/http_server.py b/python/sglang/srt/entrypoints/http_server.py
index c8b5fcbeb..18ee1f6e0 100644
--- a/python/sglang/srt/entrypoints/http_server.py
+++ b/python/sglang/srt/entrypoints/http_server.py
@@ -72,6 +72,7 @@ from sglang.srt.entrypoints.openai.serving_tokenize import (
     OpenAIServingDetokenize,
     OpenAIServingTokenize,
 )
+from sglang.srt.environ import envs
 from sglang.srt.function_call.function_call_parser import FunctionCallParser
 from sglang.srt.managers.io_struct import (
     AbortReq,
@@ -1528,12 +1529,14 @@ def _execute_server_warmup(
         json_data["sampling_params"]["max_new_tokens"] = 0
 
     try:
+        warmup_timeout = envs.SGLANG_WARMUP_TIMEOUT.get()
         if server_args.disaggregation_mode == "null":
+            logger.info(f"Start of co-locate warmup ...")
             res = requests.post(
                 url + request_name,
                 json=json_data,
                 headers=headers,
-                timeout=600,
+                timeout=warmup_timeout if warmup_timeout > 0 else 600,
             )
             assert res.status_code == 200, f"{res.text}"
             _global_state.tokenizer_manager.server_status = ServerStatus.Up
@@ -1559,7 +1562,9 @@ def _execute_server_warmup(
                 url + request_name,
                 json=json_data,
                 headers=headers,
-                timeout=1800,  # because of deep gemm precache is very long if not precache.
+                timeout=(
+                    warmup_timeout if warmup_timeout > 0 else 1800
+                ),  # because of deep gemm precache is very long if not precache.
             )
             if res.status_code == 200:
                 logger.info(
diff --git a/python/sglang/srt/environ.py b/python/sglang/srt/environ.py
index 8bec1c3c8..a7afa81d3 100644
--- a/python/sglang/srt/environ.py
+++ b/python/sglang/srt/environ.py
@@ -288,6 +288,9 @@ class Envs:
     # Ngram
     SGLANG_NGRAM_FORCE_GREEDY_VERIFY = EnvBool(False)
 
+    # Warmup
+    SGLANG_WARMUP_TIMEOUT = EnvFloat(-1) # in seconds. If a warmup forward batch takes longer than this, the server will crash to prevent hanging. Recommend to increase warmup timeout to 1800 to accommodate some kernel JIT precache e.g. deep gemm
+
     # fmt: on