diff --git a/compiler/rustc_codegen_ssa/src/back/write.rs b/compiler/rustc_codegen_ssa/src/back/write.rs
index 3e36bd8552b18..8409c927d1564 100644
--- a/compiler/rustc_codegen_ssa/src/back/write.rs
+++ b/compiler/rustc_codegen_ssa/src/back/write.rs
@@ -156,7 +156,7 @@ impl ModuleConfig {
             // `#![no_builtins]` is assumed to not participate in LTO and
             // instead goes on to generate object code.
             EmitObj::Bitcode
-        } else if need_bitcode_in_object(tcx) {
+        } else if need_bitcode_in_object(tcx) || sess.target.requires_lto {
             EmitObj::ObjectCode(BitcodeSection::Full)
         } else {
             EmitObj::ObjectCode(BitcodeSection::None)
diff --git a/src/doc/rustc/src/platform-support/amdgcn-amd-amdhsa.md b/src/doc/rustc/src/platform-support/amdgcn-amd-amdhsa.md
index dbdb96283a5fa..8934e7085b8d7 100644
--- a/src/doc/rustc/src/platform-support/amdgcn-amd-amdhsa.md
+++ b/src/doc/rustc/src/platform-support/amdgcn-amd-amdhsa.md
@@ -59,11 +59,6 @@ Build the library as `cdylib`:
 # Cargo.toml
 [lib]
 crate-type = ["cdylib"]
-
-[profile.dev]
-lto = true # LTO must be explicitly enabled for now
-[profile.release]
-lto = true
 ```
 
 The target-cpu must be from the list [supported by LLVM] (or printed with `rustc --target amdgcn-amd-amdhsa --print target-cpus`).
diff --git a/tests/run-make-cargo/amdgpu-lto/Cargo.toml b/tests/run-make-cargo/amdgpu-lto/Cargo.toml
new file mode 100644
index 0000000000000..b2607b747fd7f
--- /dev/null
+++ b/tests/run-make-cargo/amdgpu-lto/Cargo.toml
@@ -0,0 +1,8 @@
+[package]
+name = "amdgpu_lto"
+version = "0.1.0"
+edition = "2024"
+
+[lib]
+path = "lib.rs"
+crate-type = ["cdylib"]
diff --git a/tests/run-make-cargo/amdgpu-lto/lib.rs b/tests/run-make-cargo/amdgpu-lto/lib.rs
new file mode 100644
index 0000000000000..d17cf5a8316ca
--- /dev/null
+++ b/tests/run-make-cargo/amdgpu-lto/lib.rs
@@ -0,0 +1,15 @@
+#![feature(abi_gpu_kernel)]
+#![no_std]
+
+#[panic_handler]
+fn panic_handler(_info: &core::panic::PanicInfo) -> ! {
+    loop {}
+}
+
+#[unsafe(no_mangle)]
+fn foo(a: i32, b: i32) -> i32 {
+    a + b
+}
+
+#[unsafe(no_mangle)]
+extern "gpu-kernel" fn kernel() {}
diff --git a/tests/run-make-cargo/amdgpu-lto/rmake.rs b/tests/run-make-cargo/amdgpu-lto/rmake.rs
new file mode 100644
index 0000000000000..cb3dc81f34d15
--- /dev/null
+++ b/tests/run-make-cargo/amdgpu-lto/rmake.rs
@@ -0,0 +1,28 @@
+// Check that compiling for the amdgpu target which needs LTO works with a default
+// cargo configuration.
+
+//@ needs-llvm-components: amdgpu
+//@ needs-rust-lld
+
+#![deny(warnings)]
+
+use run_make_support::{cargo, path};
+
+fn main() {
+    let target_dir = path("target");
+
+    cargo()
+        .args(&[
+            "build",
+            "--release",
+            "--lib",
+            "--manifest-path",
+            "Cargo.toml",
+            "-Zbuild-std=core",
+            "--target",
+            "amdgcn-amd-amdhsa",
+        ])
+        .env("RUSTFLAGS", "-Ctarget-cpu=gfx900")
+        .env("CARGO_TARGET_DIR", &target_dir)
+        .run();
+}