diff --git a/esphome/components/api/api_connection.cpp b/esphome/components/api/api_connection.cpp
index 4b3a3e2fc8..61178dee5a 100644
--- a/esphome/components/api/api_connection.cpp
+++ b/esphome/components/api/api_connection.cpp
@@ -11,6 +11,7 @@
 #include <utility>
 #include <functional>
 #include <limits>
+#include "api_timing.h"
 #include "esphome/components/network/util.h"
 #include "esphome/core/application.h"
 #include "esphome/core/entity_base.h"
@@ -140,7 +141,9 @@ void APIConnection::loop() {
     return;
   }
 
+  API_TIMING_START(helper);
   APIError err = this->helper_->loop();
+  API_TIMING_END(helper, HELPER_LOOP);
   if (err != APIError::OK) {
     on_fatal_error();
     this->log_socket_operation_failed_(err);
@@ -149,11 +152,16 @@ void APIConnection::loop() {
 
   const uint32_t now = App.get_loop_component_start_time();
   // Check if socket has data ready before attempting to read
-  if (this->helper_->is_socket_ready()) {
+  API_TIMING_START(socket_check);
+  bool socket_ready = this->helper_->is_socket_ready();
+  API_TIMING_END(socket_check, SOCKET_CHECK);
+  if (socket_ready) {
     // Read up to MAX_MESSAGES_PER_LOOP messages per loop to improve throughput
     for (uint8_t message_count = 0; message_count < MAX_MESSAGES_PER_LOOP; message_count++) {
       ReadPacketBuffer buffer;
+      API_TIMING_START(read);
       err = this->helper_->read_packet(&buffer);
+      API_TIMING_END(read, READ_PACKET);
       if (err == APIError::WOULD_BLOCK) {
         // No more data available
         break;
@@ -164,11 +172,13 @@ void APIConnection::loop() {
       } else {
         this->last_traffic_ = now;
         // read a packet
+        API_TIMING_START(process);
         if (buffer.data_len > 0) {
           this->read_message(buffer.data_len, buffer.type, &buffer.container[buffer.data_offset]);
         } else {
           this->read_message(0, buffer.type, nullptr);
         }
+        API_TIMING_END(process, PROCESS_MESSAGE);
         if (this->flags_.remove)
           return;
       }
@@ -177,19 +187,27 @@ void APIConnection::loop() {
 
   // Process deferred batch if scheduled and timer has expired
   if (this->flags_.batch_scheduled && now - this->deferred_batch_.batch_start_time >= this->get_batch_delay_ms_()) {
+    API_TIMING_START(deferred);
     this->process_batch_();
+    API_TIMING_END(deferred, DEFERRED_BATCH);
   }
 
   if (!this->list_entities_iterator_.completed()) {
+    API_TIMING_START(iterator1);
     this->process_iterator_batch_(this->list_entities_iterator_);
+    API_TIMING_END(iterator1, ITERATOR_BATCH);
   } else if (!this->initial_state_iterator_.completed()) {
+    API_TIMING_START(iterator2);
     this->process_iterator_batch_(this->initial_state_iterator_);
+    API_TIMING_END(iterator2, ITERATOR_BATCH);
 
     // If we've completed initial states, process any remaining and clear the flag
     if (this->initial_state_iterator_.completed()) {
       // Process any remaining batched messages immediately
       if (!this->deferred_batch_.empty()) {
+        API_TIMING_START(deferred2);
         this->process_batch_();
+        API_TIMING_END(deferred2, DEFERRED_BATCH);
       }
       // Now that everything is sent, enable immediate sending for future state changes
       this->flags_.should_try_send_immediately = true;
@@ -203,6 +221,7 @@ void APIConnection::loop() {
       ESP_LOGW(TAG, "%s is unresponsive; disconnecting", this->get_client_combined_info().c_str());
     }
   } else if (now - this->last_traffic_ > KEEPALIVE_TIMEOUT_MS && !this->flags_.remove) {
+    API_TIMING_GUARD(KEEPALIVE);
     // Only send ping if we're not disconnecting
     ESP_LOGVV(TAG, "Sending keepalive PING");
     PingRequest req;
@@ -219,6 +238,7 @@ void APIConnection::loop() {
 
 #ifdef USE_CAMERA
   if (this->image_reader_ && this->image_reader_->available() && this->helper_->can_write_without_blocking()) {
+    API_TIMING_GUARD(CAMERA_IMAGE);
     uint32_t to_send = std::min((size_t) MAX_BATCH_PACKET_SIZE, this->image_reader_->available());
     bool done = this->image_reader_->available() == to_send;
 
@@ -273,9 +293,11 @@ uint16_t APIConnection::encode_message_to_buffer(ProtoMessage &msg, uint8_t mess
 #endif
 
   // Calculate size
+  API_TIMING_START(msg_size_calc);
   ProtoSize size_calc;
   msg.calculate_size(size_calc);
   uint32_t calculated_size = size_calc.get_size();
+  API_TIMING_END(msg_size_calc, ENCODE_MESSAGE);
 
   // Cache frame sizes to avoid repeated virtual calls
   const uint8_t header_padding = conn->helper_->frame_header_padding();
@@ -308,7 +330,9 @@ uint16_t APIConnection::encode_message_to_buffer(ProtoMessage &msg, uint8_t mess
 
   // Encode directly into buffer
   size_t size_before_encode = shared_buf.size();
+  API_TIMING_START(msg_encode);
   msg.encode({&shared_buf});
+  API_TIMING_END(msg_encode, ENCODE_MESSAGE);
 
   // Calculate actual encoded size (not including header that was already added)
   size_t actual_payload_size = shared_buf.size() - size_before_encode;
@@ -1560,7 +1584,9 @@ bool APIConnection::send_buffer(ProtoWriteBuffer buffer, uint8_t message_type) {
     return false;
   }
 
+  API_TIMING_START(write);
   APIError err = this->helper_->write_protobuf_packet(message_type, buffer);
+  API_TIMING_END(write, WRITE_PACKET);
   if (err == APIError::WOULD_BLOCK)
     return false;
   if (err != APIError::OK) {
@@ -1636,11 +1662,14 @@ void APIConnection::process_batch_() {
     return;
   }
 
+  API_TIMING_START(buffer_clear);
   // Try to clear buffer first
   if (!this->try_to_clear_buffer(true)) {
     // Can't write now, we'll try again later
+    API_TIMING_END(buffer_clear, BUFFER_MANAGEMENT);
     return;
   }
+  API_TIMING_END(buffer_clear, BUFFER_MANAGEMENT);
 
   // Get shared buffer reference once to avoid multiple calls
   auto &shared_buf = this->parent_->get_shared_buffer_ref();
@@ -1651,8 +1680,10 @@ void APIConnection::process_batch_() {
     const auto &item = this->deferred_batch_[0];
 
     // Let the creator calculate size and encode if it fits
+    API_TIMING_START(single_encode);
     uint16_t payload_size =
         item.creator(item.entity, this, std::numeric_limits<uint16_t>::max(), true, item.message_type);
+    API_TIMING_END(single_encode, BATCH_ENCODE);
 
     if (payload_size > 0 && this->send_buffer(ProtoWriteBuffer{&shared_buf}, item.message_type)) {
 #ifdef HAS_PROTO_MESSAGE_DUMP
@@ -1705,11 +1736,14 @@ void APIConnection::process_batch_() {
   uint32_t current_offset = 0;
 
   // Process items and encode directly to buffer (up to our limit)
+  API_TIMING_START(batch_encode_loop);
   for (size_t i = 0; i < packets_to_process; i++) {
     const auto &item = this->deferred_batch_[i];
     // Try to encode message
     // The creator will calculate overhead to determine if the message fits
+    API_TIMING_START(item_encode);
     uint16_t payload_size = item.creator(item.entity, this, remaining_size, false, item.message_type);
+    API_TIMING_END(item_encode, BATCH_ENCODE);
 
     if (payload_size == 0) {
       // Message won't fit, stop processing
@@ -1736,6 +1770,7 @@ void APIConnection::process_batch_() {
     // Current buffer size + footer space for this message
     current_offset = shared_buf.size() + footer_size;
   }
+  API_TIMING_END(batch_encode_loop, BATCH_ENCODE);
 
   if (items_processed == 0) {
     this->deferred_batch_.clear();
@@ -1748,8 +1783,10 @@ void APIConnection::process_batch_() {
   }
 
   // Send all collected packets
+  API_TIMING_START(batch_write);
   APIError err = this->helper_->write_protobuf_packets(ProtoWriteBuffer{&shared_buf},
                                                        std::span<const PacketInfo>(packet_info, packet_count));
+  API_TIMING_END(batch_write, PROTO_WRITE);
   if (err != APIError::OK && err != APIError::WOULD_BLOCK) {
     on_fatal_error();
     this->log_warning_("Batch write failed", err);
diff --git a/esphome/components/api/api_connection.h b/esphome/components/api/api_connection.h
index 6254854238..3c8cb04f96 100644
--- a/esphome/components/api/api_connection.h
+++ b/esphome/components/api/api_connection.h
@@ -6,6 +6,7 @@
 #include "api_pb2.h"
 #include "api_pb2_service.h"
 #include "api_server.h"
+#include "api_timing.h"
 #include "esphome/core/application.h"
 #include "esphome/core/component.h"
 #include "esphome/core/entity_base.h"
@@ -335,9 +336,11 @@ class APIConnection final : public APIServerConnection {
   // Helper method to process multiple entities from an iterator in a batch
   template<typename Iterator> void process_iterator_batch_(Iterator &iterator) {
     size_t initial_size = this->deferred_batch_.size();
+    API_TIMING_START(iter_advance);
     while (!iterator.completed() && (this->deferred_batch_.size() - initial_size) < MAX_INITIAL_PER_BATCH) {
       iterator.advance();
     }
+    API_TIMING_END(iter_advance, ITERATOR_ENCODE);
 
     // If the batch is full, process it immediately
     // Note: iterator.advance() already calls schedule_batch_() via schedule_message_()
@@ -681,6 +684,7 @@ class APIConnection final : public APIServerConnection {
   // Helper method to send a message either immediately or via batching
   bool send_message_smart_(EntityBase *entity, MessageCreatorPtr creator, uint8_t message_type,
                            uint8_t estimated_size) {
+    API_TIMING_GUARD(SEND_STATE);
     // Try to send immediately if:
     // 1. It's an UpdateStateResponse (always send immediately to handle cases where
     //    the main loop is blocked, e.g., during OTA updates)
diff --git a/esphome/components/api/api_pb2_service.cpp b/esphome/components/api/api_pb2_service.cpp
index 6b7b8b9ebd..27a3a5eff9 100644
--- a/esphome/components/api/api_pb2_service.cpp
+++ b/esphome/components/api/api_pb2_service.cpp
@@ -1,6 +1,7 @@
 // This file was automatically generated with a tool.
 // See script/api_protobuf/api_protobuf.py
 #include "api_pb2_service.h"
+#include "api_timing.h"
 #include "esphome/core/log.h"
 
 namespace esphome::api {
@@ -14,14 +15,18 @@ void APIServerConnectionBase::log_send_message_(const char *name, const std::str
 #endif
 
 void APIServerConnectionBase::read_message(uint32_t msg_size, uint32_t msg_type, uint8_t *msg_data) {
+  API_TIMING_START(msg_decode);
   switch (msg_type) {
     case HelloRequest::MESSAGE_TYPE: {
       HelloRequest msg;
       msg.decode(msg_data, msg_size);
+      API_TIMING_END(msg_decode, PROTO_READ);
 #ifdef HAS_PROTO_MESSAGE_DUMP
       ESP_LOGVV(TAG, "on_hello_request: %s", msg.dump().c_str());
 #endif
+      API_TIMING_START(msg_handle);
       this->on_hello_request(msg);
+      API_TIMING_END(msg_handle, MESSAGE_DISPATCH);
       break;
     }
     case ConnectRequest::MESSAGE_TYPE: {
diff --git a/esphome/components/api/api_server.cpp b/esphome/components/api/api_server.cpp
index 1f38f4a31a..8bee564568 100644
--- a/esphome/components/api/api_server.cpp
+++ b/esphome/components/api/api_server.cpp
@@ -2,6 +2,7 @@
 #ifdef USE_API
 #include <cerrno>
 #include "api_connection.h"
+#include "api_timing.h"
 #include "esphome/components/network/util.h"
 #include "esphome/core/application.h"
 #include "esphome/core/defines.h"
@@ -135,6 +136,7 @@ void APIServer::schedule_reboot_timeout_() {
 void APIServer::loop() {
   // Accept new clients only if the socket exists and has incoming connections
   if (this->socket_ && this->socket_->ready()) {
+    API_TIMING_GUARD(ACCEPT_CLIENT);
     while (true) {
       struct sockaddr_storage source_addr;
       socklen_t addr_len = sizeof(source_addr);
@@ -161,6 +163,7 @@ void APIServer::loop() {
 
   // Process clients and remove disconnected ones in a single pass
   // Check network connectivity once for all clients
+  API_TIMING_START(network_check);
   if (!network::is_connected()) {
     // Network is down - disconnect all clients
     for (auto &client : this->clients_) {
@@ -169,6 +172,7 @@ void APIServer::loop() {
     }
     // Continue to process and clean up the clients below
   }
+  API_TIMING_END(network_check, NETWORK_CHECK);
 
   size_t client_index = 0;
   while (client_index < this->clients_.size()) {
@@ -176,12 +180,15 @@ void APIServer::loop() {
 
     if (!client->flags_.remove) {
       // Common case: process active client
+      API_TIMING_START(client_loop);
       client->loop();
+      API_TIMING_END(client_loop, CLIENT_LOOP);
       client_index++;
       continue;
     }
 
     // Rare case: handle disconnection
+    API_TIMING_GUARD(CLIENT_REMOVAL);
 #ifdef USE_API_CLIENT_DISCONNECTED_TRIGGER
     this->client_disconnected_trigger_->trigger(client->client_info_.name, client->client_info_.peername);
 #endif
diff --git a/esphome/components/runtime_stats/runtime_stats.cpp b/esphome/components/runtime_stats/runtime_stats.cpp
index 8f5d5daf01..4570fe3919 100644
--- a/esphome/components/runtime_stats/runtime_stats.cpp
+++ b/esphome/components/runtime_stats/runtime_stats.cpp
@@ -4,6 +4,10 @@
 
 #include "esphome/core/component.h"
 #include <algorithm>
+#include <cinttypes>
+#ifdef USE_API
+#include "esphome/components/api/api_timing.h"
+#endif
 
 namespace esphome {
 
@@ -62,6 +66,29 @@ void RuntimeStatsCollector::log_stats_() {
              stats->get_period_time_ms());
   }
 
+#ifdef USE_API
+  // Log detailed API timing breakdown if available
+  bool has_api_timing = false;
+  for (int i = 0; i < static_cast<int>(api::APITimingCategory::MAX_CATEGORIES); i++) {
+    if (api::api_timing_stats[i].count > 0) {
+      has_api_timing = true;
+      break;
+    }
+  }
+
+  if (has_api_timing) {
+    ESP_LOGI(TAG, "API Component Breakdown (microseconds):");
+    for (int i = 0; i < static_cast<int>(api::APITimingCategory::MAX_CATEGORIES); i++) {
+      const api::APITimingStats &stats = api::api_timing_stats[i];
+      if (stats.count > 0) {
+        const char *category_name = api::api_timing_category_to_str(static_cast<api::APITimingCategory>(i));
+        ESP_LOGI(TAG, "    %s: count=%" PRIu32 ", avg=%.1fus, max=%" PRIu32 "us, total=%" PRIu64 "us", category_name,
+                 stats.count, stats.get_avg_us(), stats.max_us, stats.total_us);
+      }
+    }
+  }
+#endif
+
   // Log total stats since boot
   ESP_LOGI(TAG, "Total stats (since boot):");
 
@@ -81,6 +108,16 @@ void RuntimeStatsCollector::log_stats_() {
   }
 }
 
+void RuntimeStatsCollector::reset_stats_() {
+  for (auto &it : this->component_stats_) {
+    it.second.reset_period_stats();
+  }
+#ifdef USE_API
+  // Reset API timing stats for next period
+  api::reset_api_timing_stats();
+#endif
+}
+
 void RuntimeStatsCollector::process_pending_stats(uint32_t current_time) {
   if (this->next_log_time_ == 0)
     return;
diff --git a/esphome/components/runtime_stats/runtime_stats.h b/esphome/components/runtime_stats/runtime_stats.h
index e2f8bee563..3dcf289d01 100644
--- a/esphome/components/runtime_stats/runtime_stats.h
+++ b/esphome/components/runtime_stats/runtime_stats.h
@@ -103,11 +103,7 @@ class RuntimeStatsCollector {
  protected:
   void log_stats_();
 
-  void reset_stats_() {
-    for (auto &it : this->component_stats_) {
-      it.second.reset_period_stats();
-    }
-  }
+  void reset_stats_();
 
   // Use const char* keys for efficiency
   // Custom comparator for const char* keys in map