diff --git a/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/long.h b/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/long.h
index 0f66203dfd1..512fd81fb3e 100644
--- a/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/long.h
+++ b/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/long.h
@@ -5,6 +5,7 @@
 
 #include <Python.h>
 #if PY_VERSION_HEX >= 0x030c0000
+#define Py_BUILD_CORE
 #include <internal/pycore_long.h>
 // Note: Even if use the right PYLONG_BITS_IN_DIGIT that is specified in the
 // Python we use to build echion, it can be different from the Python that is
diff --git a/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/tasks.h b/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/tasks.h
index 87d86c470e7..848af75aef2 100644
--- a/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/tasks.h
+++ b/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/tasks.h
@@ -4,6 +4,8 @@
 
 #pragma once
 
+#include <optional>
+
 #define PY_SSIZE_T_CLEAN
 #include <Python.h>
 #include <weakrefobject.h>
@@ -111,8 +113,8 @@ GenInfo::create(PyObject* gen_addr)
         // Type-pun the PyGenObject to a PyAsyncGenASend. *gen_addr was actually never a PyGenObject to begin with,
         // but we do not care as the only thing we will use from it is the ags_gen field.
         PyAsyncGenASend* asend = reinterpret_cast<PyAsyncGenASend*>(&gen);
-        PyAsyncGenObject* gen = asend->ags_gen;
-        auto asend_yf = reinterpret_cast<PyObject*>(gen);
+        PyAsyncGenObject* gen_ptr = asend->ags_gen;
+        auto asend_yf = reinterpret_cast<PyObject*>(gen_ptr);
         auto result = GenInfo::create(asend_yf);
         recursion_depth--;
         return result;
@@ -176,6 +178,7 @@ class TaskInfo
 
     // Information to reconstruct the async stack as best as we can
     TaskInfo::Ptr waiter = nullptr;
+    std::optional<bool> is_on_cpu_ = std::nullopt;
 
     [[nodiscard]] static Result<TaskInfo::Ptr> create(TaskObj*);
     TaskInfo(PyObject* origin, PyObject* loop, GenInfo::Ptr coro, StringTable::Key name, TaskInfo::Ptr waiter)
@@ -189,6 +192,27 @@ class TaskInfo
 
     [[nodiscard]] static Result<TaskInfo::Ptr> current(PyObject*);
     inline size_t unwind(FrameStack&);
+
+    // Check if any coroutine in the chain is currently running (on CPU)
+    inline bool is_on_cpu()
+    {
+        if (is_on_cpu_.has_value()) {
+            return *is_on_cpu_;
+        }
+
+        auto* last_coro = static_cast<GenInfo*>(nullptr);
+        auto* current_coro = this->coro.get();
+
+        // Check if the innermost coroutine is running.
+        // We don't need to test all the other coroutines as they are awaiting, by definition.
+        while (current_coro) {
+            last_coro = current_coro;
+            current_coro = current_coro->await.get();
+        }
+
+        is_on_cpu_ = last_coro && last_coro->is_running;
+        return *is_on_cpu_;
+    }
 };
 
 inline std::unordered_map<PyObject*, PyObject*> task_link_map;
diff --git a/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/threads.h b/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/threads.h
index 4bffbcd0dc5..7eed87f849a 100644
--- a/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/threads.h
+++ b/ddtrace/internal/datadog/profiling/stack_v2/echion/echion/threads.h
@@ -252,8 +252,17 @@ ThreadInfo::unwind_tasks()
         }
     }
 
+    // Only one Task can be on CPU at a time.
+    // Since determining if a task is on CPU is somewhat costly, we
+    // stop checking if Tasks are on CPU after seeing the first one.
+    bool on_cpu_task_seen = false;
     for (auto& leaf_task : leaf_tasks) {
-        bool on_cpu = leaf_task.get().coro->is_running;
+        bool on_cpu = false;
+        if (!on_cpu_task_seen) {
+            on_cpu = leaf_task.get().is_on_cpu();
+            on_cpu_task_seen = on_cpu;
+        }
+
         auto stack_info = std::make_unique<StackInfo>(leaf_task.get().name, on_cpu);
         auto& stack = stack_info->stack;
         for (auto current_task = leaf_task;;) {
diff --git a/releasenotes/notes/profiling-fix-detection-of-on-cpu-tasks-a2eb47163c3df850.yaml b/releasenotes/notes/profiling-fix-detection-of-on-cpu-tasks-a2eb47163c3df850.yaml
new file mode 100644
index 00000000000..c7e066d8893
--- /dev/null
+++ b/releasenotes/notes/profiling-fix-detection-of-on-cpu-tasks-a2eb47163c3df850.yaml
@@ -0,0 +1,5 @@
+fixes:
+  - |
+    profiling: This fix improves the detection of on-CPU asyncio Tasks. Previously, the Profiler would only
+    consider a Task as running if its coroutine was running. The Profiler now recursively checks if any coroutine in the
+    await chain of the Task's coroutine is running.