diff --git a/src/proto_alpha/lib_parameters/default_parameters.ml b/src/proto_alpha/lib_parameters/default_parameters.ml
index 545470e06e4044b433f634ef91ef9b9b53061a79..80b8a541f77fefd45b44110c862b89edc3e1742f 100644
--- a/src/proto_alpha/lib_parameters/default_parameters.ml
+++ b/src/proto_alpha/lib_parameters/default_parameters.ml
@@ -293,6 +293,7 @@ let constants_mainnet =
         staking_over_baking_limit = 5;
         max_costaking_baker_count = 5;
         staking_over_delegation_edge = 2;
+        launch_ema_threshold = 1_600_000_000L;
       };
   }
 
diff --git a/src/proto_alpha/lib_protocol/adaptive_inflation_storage.ml b/src/proto_alpha/lib_protocol/adaptive_inflation_storage.ml
index 0b6940d510ab383178f39fb55bed9a6241918df0..6f0b3231ab06459de9711bdb974eaf382a9a5900 100644
--- a/src/proto_alpha/lib_protocol/adaptive_inflation_storage.ml
+++ b/src/proto_alpha/lib_protocol/adaptive_inflation_storage.ml
@@ -203,6 +203,10 @@ let load_reward_coeff ctxt =
 
 let init_ema ctxt = Storage.Adaptive_inflation.Launch_ema.init ctxt 0L
 
+let activate ctxt ~cycle = Storage.Adaptive_inflation.Activation.init ctxt cycle
+
+let launch_cycle ctxt = Storage.Adaptive_inflation.Activation.find ctxt
+
 let update_ema ctxt ~vote =
   Storage.Adaptive_inflation.Launch_ema.get ctxt >>=? fun old_ema ->
   Toggle_votes_repr.Adaptive_inflation_launch_EMA.of_int64 old_ema
@@ -215,12 +219,24 @@ let update_ema ctxt ~vote =
   Storage.Adaptive_inflation.Launch_ema.update
     ctxt
     (Toggle_votes_repr.Adaptive_inflation_launch_EMA.to_int64 new_ema)
+  >>=? fun ctxt ->
+  let open Constants_storage in
+  (if
+   Toggle_votes_repr.Adaptive_inflation_launch_EMA.(
+     new_ema < adaptive_inflation_launch_ema_threshold ctxt)
+  then return ctxt
+  else
+    launch_cycle ctxt >>=? function
+    | Some _ ->
+        (* the feature is already set to launch, do nothing to avoid postponing it. *)
+        return ctxt
+    | None ->
+        (* set the feature to activate in a few cycles *)
+        let current_cycle = (Level_storage.current ctxt).cycle in
+        let delay = 1 + preserved_cycles ctxt + max_slashing_period ctxt in
+        activate ctxt ~cycle:(Cycle_repr.add current_cycle delay))
   >|=? fun ctxt -> (ctxt, new_ema)
 
-let activate ctxt ~cycle = Storage.Adaptive_inflation.Activation.init ctxt cycle
-
-let launch_cycle ctxt = Storage.Adaptive_inflation.Activation.find ctxt
-
 module For_RPC = struct
   let get_reward_coeff = get_reward_coeff
 end
diff --git a/src/proto_alpha/lib_protocol/adaptive_inflation_storage.mli b/src/proto_alpha/lib_protocol/adaptive_inflation_storage.mli
index 2551ad0c88bee7ff769d39932c541b0e48c0ac1d..cd57bf18aaeb91b373ab7ee22752d971fc6a8899 100644
--- a/src/proto_alpha/lib_protocol/adaptive_inflation_storage.mli
+++ b/src/proto_alpha/lib_protocol/adaptive_inflation_storage.mli
@@ -49,14 +49,6 @@ val update_ema :
   (Raw_context.t * Toggle_votes_repr.Adaptive_inflation_launch_EMA.t) tzresult
   Lwt.t
 
-(** [activate ctxt ~cycle] adds into the context the cycle at which
-    the adaptive inflation feature gets activated. If this function is
-    never called, then the context does not contain the cycle for the
-    feature activation, which implies the feature is inactive.
-*)
-val activate :
-  Raw_context.t -> cycle:Cycle_repr.t -> Raw_context.t tzresult Lwt.t
-
 (** [launch_cycle ctxt] reads from the context the cycle at which
     the adaptive inflation feature is set to activate.
 
diff --git a/src/proto_alpha/lib_protocol/alpha_context.mli b/src/proto_alpha/lib_protocol/alpha_context.mli
index 1bb9a3f539ff9b8447ef3fcd49adc0f76f650884..4ec828986a1bac6118345ec3d0a068a9c6790ff9 100644
--- a/src/proto_alpha/lib_protocol/alpha_context.mli
+++ b/src/proto_alpha/lib_protocol/alpha_context.mli
@@ -847,6 +847,7 @@ module Constants : sig
       staking_over_baking_limit : int;
       max_costaking_baker_count : int;
       staking_over_delegation_edge : int;
+      launch_ema_threshold : int64;
     }
 
     type reward_weights = {
diff --git a/src/proto_alpha/lib_protocol/constants_parametric_repr.ml b/src/proto_alpha/lib_protocol/constants_parametric_repr.ml
index 27e76de5c5e8fb26400d87c9b9ede1b49a5f2d58..f9e9dd33bf36bda40eb1b712c0d2c564112be407 100644
--- a/src/proto_alpha/lib_protocol/constants_parametric_repr.ml
+++ b/src/proto_alpha/lib_protocol/constants_parametric_repr.ml
@@ -128,6 +128,7 @@ type adaptive_inflation = {
   staking_over_baking_limit : int;
   max_costaking_baker_count : int;
   staking_over_delegation_edge : int;
+  launch_ema_threshold : int64;
 }
 
 type reward_weights = {
@@ -344,26 +345,31 @@ let adaptive_inflation_encoding =
            staking_over_baking_limit;
            max_costaking_baker_count;
            staking_over_delegation_edge;
+           launch_ema_threshold;
          } ->
       ( enable,
         staking_over_baking_limit,
         max_costaking_baker_count,
-        staking_over_delegation_edge ))
+        staking_over_delegation_edge,
+        launch_ema_threshold ))
     (fun ( adaptive_inflation_enable,
            staking_over_baking_limit,
            max_costaking_baker_count,
-           staking_over_delegation_edge ) ->
+           staking_over_delegation_edge,
+           launch_ema_threshold ) ->
       {
         enable = adaptive_inflation_enable;
         staking_over_baking_limit;
         max_costaking_baker_count;
         staking_over_delegation_edge;
+        launch_ema_threshold;
       })
-    (obj4
+    (obj5
        (req "adaptive_inflation_enable" bool)
        (req "staking_over_baking_limit" uint8)
        (req "max_costaking_baker_count" uint16)
-       (req "staking_over_delegation_edge" uint8))
+       (req "staking_over_delegation_edge" uint8)
+       (req "adaptive_inflation_launch_ema_threshold" int64))
 
 let reward_weights_encoding =
   let open Data_encoding in
diff --git a/src/proto_alpha/lib_protocol/constants_parametric_repr.mli b/src/proto_alpha/lib_protocol/constants_parametric_repr.mli
index 620a025d7520103546850e83f22aa416d792f39a..04c8b5b47d2f28dd85926d486c1a468d70b20d31 100644
--- a/src/proto_alpha/lib_protocol/constants_parametric_repr.mli
+++ b/src/proto_alpha/lib_protocol/constants_parametric_repr.mli
@@ -146,6 +146,7 @@ type adaptive_inflation = {
   max_costaking_baker_count :
     (* Maximal number of bakers an account can costake to. *) int;
   staking_over_delegation_edge : (* Weight of staking over delegation. *) int;
+  launch_ema_threshold : (* Threshold of the activation vote *) int64;
 }
 
 type reward_weights = {
diff --git a/src/proto_alpha/lib_protocol/constants_storage.ml b/src/proto_alpha/lib_protocol/constants_storage.ml
index 39b7c4439d4a383c4134dec39a1879b6f484184b..87df7d47928eff0471480bfb1d0293080e3dabbd 100644
--- a/src/proto_alpha/lib_protocol/constants_storage.ml
+++ b/src/proto_alpha/lib_protocol/constants_storage.ml
@@ -288,4 +288,7 @@ let adaptive_inflation_staking_over_baking_limit c =
 let adaptive_inflation_staking_over_delegation_edge c =
   (adaptive_inflation c).staking_over_delegation_edge
 
+let adaptive_inflation_launch_ema_threshold c =
+  (adaptive_inflation c).launch_ema_threshold
+
 let freeze_rewards = adaptive_inflation_enable
diff --git a/src/proto_alpha/lib_protocol/constants_storage.mli b/src/proto_alpha/lib_protocol/constants_storage.mli
index 0c5204a09ee383eca842d13b26279c1fba191cf9..e74570d6468916c4dd5f41243451b61909c396fb 100644
--- a/src/proto_alpha/lib_protocol/constants_storage.mli
+++ b/src/proto_alpha/lib_protocol/constants_storage.mli
@@ -168,4 +168,6 @@ val adaptive_inflation_staking_over_baking_limit : Raw_context.t -> int
 
 val adaptive_inflation_staking_over_delegation_edge : Raw_context.t -> int
 
+val adaptive_inflation_launch_ema_threshold : Raw_context.t -> int64
+
 val freeze_rewards : Raw_context.t -> bool
diff --git a/src/proto_alpha/lib_protocol/raw_context.ml b/src/proto_alpha/lib_protocol/raw_context.ml
index 84ba6d37a825167bf62a9b9d8b764a74188a3d43..a6672ffe1ac60dec7b7fa5a59100eabeca2e440b 100644
--- a/src/proto_alpha/lib_protocol/raw_context.ml
+++ b/src/proto_alpha/lib_protocol/raw_context.ml
@@ -999,6 +999,8 @@ let prepare_first_block ~level ~timestamp ctxt =
             staking_over_baking_limit = 5;
             max_costaking_baker_count = 5;
             staking_over_delegation_edge = 2;
+            launch_ema_threshold =
+              (* 80% of the max ema (which is 2 billion) *) 1_600_000_000L;
           }
       in
 
diff --git a/tezt/tests/expected/RPC_test.ml/Alpha- (mode client) RPC regression tests- misc_protocol.out b/tezt/tests/expected/RPC_test.ml/Alpha- (mode client) RPC regression tests- misc_protocol.out
index b1f8e8278ba3c97a9d6c2a93b5b886988d8f1e49..f871f256776ce35d7ec367112dc7d0567c2532cc 100644
--- a/tezt/tests/expected/RPC_test.ml/Alpha- (mode client) RPC regression tests- misc_protocol.out	
+++ b/tezt/tests/expected/RPC_test.ml/Alpha- (mode client) RPC regression tests- misc_protocol.out	
@@ -66,7 +66,8 @@
   "smart_rollup_max_number_of_parallel_games": 32, "zk_rollup_enable": false,
   "zk_rollup_origination_size": 4000, "zk_rollup_min_pending_to_process": 10,
   "adaptive_inflation_enable": false, "staking_over_baking_limit": 5,
-  "max_costaking_baker_count": 5, "staking_over_delegation_edge": 2 }
+  "max_costaking_baker_count": 5, "staking_over_delegation_edge": 2,
+  "adaptive_inflation_launch_ema_threshold": "1600000000" }
 
 ./octez-client rpc get /chains/main/blocks/head/helpers/baking_rights
 [ { "level": 2, "delegate": "[PUBLIC_KEY_HASH]",
diff --git a/tezt/tests/expected/RPC_test.ml/Alpha- (mode light) RPC regression tests- misc_protocol.out b/tezt/tests/expected/RPC_test.ml/Alpha- (mode light) RPC regression tests- misc_protocol.out
index bfe7e0088d6f338955fffc50ba3f758bee99e201..e78ae19711bc121241ac7d6dbf819fd3fc752d26 100644
--- a/tezt/tests/expected/RPC_test.ml/Alpha- (mode light) RPC regression tests- misc_protocol.out	
+++ b/tezt/tests/expected/RPC_test.ml/Alpha- (mode light) RPC regression tests- misc_protocol.out	
@@ -66,7 +66,8 @@
   "smart_rollup_max_number_of_parallel_games": 32, "zk_rollup_enable": false,
   "zk_rollup_origination_size": 4000, "zk_rollup_min_pending_to_process": 10,
   "adaptive_inflation_enable": false, "staking_over_baking_limit": 5,
-  "max_costaking_baker_count": 5, "staking_over_delegation_edge": 2 }
+  "max_costaking_baker_count": 5, "staking_over_delegation_edge": 2,
+  "adaptive_inflation_launch_ema_threshold": "1600000000" }
 
 ./octez-client --mode light rpc get /chains/main/blocks/head/helpers/baking_rights
 [ { "level": 2, "delegate": "[PUBLIC_KEY_HASH]",
diff --git a/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy) RPC regression tests- misc_protocol.out b/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy) RPC regression tests- misc_protocol.out
index 8ea4205178e9d442ea2b68f1192bbdcd60176440..0fc0ccbd2b9d48437e41bbc1c1e3dd139608f0df 100644
--- a/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy) RPC regression tests- misc_protocol.out	
+++ b/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy) RPC regression tests- misc_protocol.out	
@@ -66,7 +66,8 @@
   "smart_rollup_max_number_of_parallel_games": 32, "zk_rollup_enable": false,
   "zk_rollup_origination_size": 4000, "zk_rollup_min_pending_to_process": 10,
   "adaptive_inflation_enable": false, "staking_over_baking_limit": 5,
-  "max_costaking_baker_count": 5, "staking_over_delegation_edge": 2 }
+  "max_costaking_baker_count": 5, "staking_over_delegation_edge": 2,
+  "adaptive_inflation_launch_ema_threshold": "1600000000" }
 
 ./octez-client --mode proxy rpc get /chains/main/blocks/head/helpers/baking_rights
 [ { "level": 2, "delegate": "[PUBLIC_KEY_HASH]",
diff --git a/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy_server_data_dir) RPC regression tests- misc_protocol.out b/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy_server_data_dir) RPC regression tests- misc_protocol.out
index 52f185b881774b218729e98adb310e333d3aa794..ce9a02becea58f53aeab369d2d1cca5dcfb63d45 100644
--- a/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy_server_data_dir) RPC regression tests- misc_protocol.out	
+++ b/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy_server_data_dir) RPC regression tests- misc_protocol.out	
@@ -66,7 +66,8 @@
   "smart_rollup_max_number_of_parallel_games": 32, "zk_rollup_enable": false,
   "zk_rollup_origination_size": 4000, "zk_rollup_min_pending_to_process": 10,
   "adaptive_inflation_enable": false, "staking_over_baking_limit": 5,
-  "max_costaking_baker_count": 5, "staking_over_delegation_edge": 2 }
+  "max_costaking_baker_count": 5, "staking_over_delegation_edge": 2,
+  "adaptive_inflation_launch_ema_threshold": "1600000000" }
 
 ./octez-client rpc get /chains/main/blocks/head/helpers/baking_rights
 [ { "level": 3, "delegate": "[PUBLIC_KEY_HASH]",
diff --git a/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy_server_rpc) RPC regression tests- misc_protocol.out b/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy_server_rpc) RPC regression tests- misc_protocol.out
index 52f185b881774b218729e98adb310e333d3aa794..ce9a02becea58f53aeab369d2d1cca5dcfb63d45 100644
--- a/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy_server_rpc) RPC regression tests- misc_protocol.out	
+++ b/tezt/tests/expected/RPC_test.ml/Alpha- (mode proxy_server_rpc) RPC regression tests- misc_protocol.out	
@@ -66,7 +66,8 @@
   "smart_rollup_max_number_of_parallel_games": 32, "zk_rollup_enable": false,
   "zk_rollup_origination_size": 4000, "zk_rollup_min_pending_to_process": 10,
   "adaptive_inflation_enable": false, "staking_over_baking_limit": 5,
-  "max_costaking_baker_count": 5, "staking_over_delegation_edge": 2 }
+  "max_costaking_baker_count": 5, "staking_over_delegation_edge": 2,
+  "adaptive_inflation_launch_ema_threshold": "1600000000" }
 
 ./octez-client rpc get /chains/main/blocks/head/helpers/baking_rights
 [ { "level": 3, "delegate": "[PUBLIC_KEY_HASH]",