diff --git a/compiler/rustc_hir_analysis/src/hir_ty_lowering/generics.rs b/compiler/rustc_hir_analysis/src/hir_ty_lowering/generics.rs
index f5a64ede398ea..0a2946323c7b3 100644
--- a/compiler/rustc_hir_analysis/src/hir_ty_lowering/generics.rs
+++ b/compiler/rustc_hir_analysis/src/hir_ty_lowering/generics.rs
@@ -9,6 +9,7 @@ use rustc_middle::ty::{
 };
 use rustc_session::lint::builtin::LATE_BOUND_LIFETIME_ARGUMENTS;
 use rustc_span::kw;
+use rustc_trait_selection::traits;
 use smallvec::SmallVec;
 use tracing::{debug, instrument};
 
@@ -535,9 +536,26 @@ pub(crate) fn check_generic_arg_count(
                     .map(|param| param.name)
                     .collect();
                 if constraint_names == param_names {
+                    let has_assoc_ty_with_same_name =
+                        if let DefKind::Trait = cx.tcx().def_kind(def_id) {
+                            gen_args.constraints.iter().any(|constraint| {
+                                traits::supertrait_def_ids(cx.tcx(), def_id).any(|trait_did| {
+                                    cx.probe_trait_that_defines_assoc_item(
+                                        trait_did,
+                                        ty::AssocTag::Type,
+                                        constraint.ident,
+                                    )
+                                })
+                            })
+                        } else {
+                            false
+                        };
                     // We set this to true and delay emitting `WrongNumberOfGenericArgs`
-                    // to provide a succinct error for cases like issue #113073
-                    all_params_are_binded = true;
+                    // to provide a succinct error for cases like issue #113073,
+                    // but only if when we don't have any assoc type with the same name with a
+                    // generic arg. Otherwise it will cause an ICE due to a delayed error because we
+                    // don't have any error other than `WrongNumberOfGenericArgs`.
+                    all_params_are_binded = !has_assoc_ty_with_same_name;
                 };
             }
 
diff --git a/tests/run-make/dump-ice-to-disk/rmake.rs b/tests/run-make/dump-ice-to-disk/rmake.rs
index 09a34cdeb5e70..319cbc6e0ca89 100644
--- a/tests/run-make/dump-ice-to-disk/rmake.rs
+++ b/tests/run-make/dump-ice-to-disk/rmake.rs
@@ -18,13 +18,16 @@
 //! # Test history
 //!
 //! The previous rmake.rs iteration of this test was flaky for unknown reason on
-//! `i686-pc-windows-gnu` *specifically*, so assertion failures in this test was made extremely
-//! verbose to help diagnose why the ICE messages was different. It appears that backtraces on
-//! `i686-pc-windows-gnu` specifically are quite unpredictable in how many backtrace frames are
-//! involved.
+//! `i686-pc-windows-gnu`, so assertion failures in this test was made extremely verbose to help
+//! diagnose why the ICE messages was different. It appears that backtraces on `i686-pc-windows-gnu`
+//! specifically are quite unpredictable in how many backtrace frames are involved.
+//!
+//! Disabled on `i686-pc-windows-msvc` as well, because sometimes the middle portion of the ICE
+//! backtrace becomes `<unknown>`.
 
 //@ ignore-cross-compile (exercising ICE dump on host)
 //@ ignore-i686-pc-windows-gnu (unwind mechanism produces unpredictable backtraces)
+//@ ignore-i686-pc-windows-msvc (sometimes partial backtrace becomes `<unknown>`)
 
 use std::cell::OnceCell;
 use std::path::{Path, PathBuf};
diff --git a/tests/ui/traits/associated_type_bound/assoc-type-bounds-with-the-same-name-with-lacking-generic-arg-148121.rs b/tests/ui/traits/associated_type_bound/assoc-type-bounds-with-the-same-name-with-lacking-generic-arg-148121.rs
new file mode 100644
index 0000000000000..f1dffc0ff6b4e
--- /dev/null
+++ b/tests/ui/traits/associated_type_bound/assoc-type-bounds-with-the-same-name-with-lacking-generic-arg-148121.rs
@@ -0,0 +1,17 @@
+// A regression test for https://github.com/rust-lang/rust/issues/148121
+
+pub trait Super<X> {
+    type X;
+}
+
+pub trait Zelf<X>: Super<X> {}
+
+pub trait A {}
+
+impl A for dyn Super<X = ()> {}
+//~^ ERROR: trait takes 1 generic argument but 0 generic arguments were supplied
+
+impl A for dyn Zelf<X = ()> {}
+//~^ ERROR: trait takes 1 generic argument but 0 generic arguments were supplied
+
+fn main() {}
diff --git a/tests/ui/traits/associated_type_bound/assoc-type-bounds-with-the-same-name-with-lacking-generic-arg-148121.stderr b/tests/ui/traits/associated_type_bound/assoc-type-bounds-with-the-same-name-with-lacking-generic-arg-148121.stderr
new file mode 100644
index 0000000000000..5a7969193b553
--- /dev/null
+++ b/tests/ui/traits/associated_type_bound/assoc-type-bounds-with-the-same-name-with-lacking-generic-arg-148121.stderr
@@ -0,0 +1,35 @@
+error[E0107]: trait takes 1 generic argument but 0 generic arguments were supplied
+  --> $DIR/assoc-type-bounds-with-the-same-name-with-lacking-generic-arg-148121.rs:11:16
+   |
+LL | impl A for dyn Super<X = ()> {}
+   |                ^^^^^ expected 1 generic argument
+   |
+note: trait defined here, with 1 generic parameter: `X`
+  --> $DIR/assoc-type-bounds-with-the-same-name-with-lacking-generic-arg-148121.rs:3:11
+   |
+LL | pub trait Super<X> {
+   |           ^^^^^ -
+help: add missing generic argument
+   |
+LL | impl A for dyn Super<X, X = ()> {}
+   |                      ++
+
+error[E0107]: trait takes 1 generic argument but 0 generic arguments were supplied
+  --> $DIR/assoc-type-bounds-with-the-same-name-with-lacking-generic-arg-148121.rs:14:16
+   |
+LL | impl A for dyn Zelf<X = ()> {}
+   |                ^^^^ expected 1 generic argument
+   |
+note: trait defined here, with 1 generic parameter: `X`
+  --> $DIR/assoc-type-bounds-with-the-same-name-with-lacking-generic-arg-148121.rs:7:11
+   |
+LL | pub trait Zelf<X>: Super<X> {}
+   |           ^^^^ -
+help: add missing generic argument
+   |
+LL | impl A for dyn Zelf<X, X = ()> {}
+   |                     ++
+
+error: aborting due to 2 previous errors
+
+For more information about this error, try `rustc --explain E0107`.