diff --git a/rust/Cargo.lock b/rust/Cargo.lock
index 6acc0e7d..342d5610 100644
--- a/rust/Cargo.lock
+++ b/rust/Cargo.lock
@@ -1011,8 +1011,18 @@ version = "0.14.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7b750cb3417fd1b327431a470f388520309479ab0bf5e323505daf0290cd3850"
 dependencies = [
- "darling_core",
- "darling_macro",
+ "darling_core 0.14.4",
+ "darling_macro 0.14.4",
+]
+
+[[package]]
+name = "darling"
+version = "0.20.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fc7f46116c46ff9ab3eb1597a45688b6715c6e628b5c133e288e709a29bcb4ee"
+dependencies = [
+ "darling_core 0.20.11",
+ "darling_macro 0.20.11",
 ]
 
 [[package]]
@@ -1029,17 +1039,42 @@ dependencies = [
  "syn 1.0.109",
 ]
 
+[[package]]
+name = "darling_core"
+version = "0.20.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0d00b9596d185e565c2207a0b01f8bd1a135483d02d9b7b0a54b11da8d53412e"
+dependencies = [
+ "fnv",
+ "ident_case",
+ "proc-macro2",
+ "quote",
+ "strsim 0.11.1",
+ "syn 2.0.100",
+]
+
 [[package]]
 name = "darling_macro"
 version = "0.14.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a4aab4dbc9f7611d8b55048a3a16d2d010c2c8334e46304b40ac1cc14bf3b48e"
 dependencies = [
- "darling_core",
+ "darling_core 0.14.4",
  "quote",
  "syn 1.0.109",
 ]
 
+[[package]]
+name = "darling_macro"
+version = "0.20.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fc34b93ccb385b40dc71c6fceac4b2ad23662c7eeb248cf10d529b7e055b6ead"
+dependencies = [
+ "darling_core 0.20.11",
+ "quote",
+ "syn 2.0.100",
+]
+
 [[package]]
 name = "dasp_sample"
 version = "0.11.0"
@@ -1087,7 +1122,7 @@ version = "0.12.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c11bdc11a0c47bc7d37d582b5285da6849c96681023680b906673c5707af7b0f"
 dependencies = [
- "darling",
+ "darling 0.14.4",
  "proc-macro2",
  "quote",
  "syn 1.0.109",
@@ -1131,7 +1166,7 @@ dependencies = [
  "libc",
  "option-ext",
  "redox_users",
- "windows-sys 0.59.0",
+ "windows-sys 0.60.2",
 ]
 
 [[package]]
@@ -1971,7 +2006,7 @@ dependencies = [
  "js-sys",
  "log",
  "wasm-bindgen",
- "windows-core 0.61.0",
+ "windows-core 0.61.2",
 ]
 
 [[package]]
@@ -2333,7 +2368,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "fc2f4eb4bc735547cfed7c0a4922cbd04a4655978c09b54f1f7b228750664c34"
 dependencies = [
  "cfg-if",
- "windows-targets 0.52.6",
+ "windows-targets 0.48.5",
 ]
 
 [[package]]
@@ -2709,6 +2744,7 @@ dependencies = [
  "log",
  "moshi",
  "numpy",
+ "nvml-wrapper",
  "ogg",
  "opus",
  "prometheus",
@@ -2722,6 +2758,7 @@ dependencies = [
  "serde",
  "serde_json",
  "symphonia",
+ "sysinfo 0.35.2",
  "tokio",
  "toml",
  "tower 0.4.13",
@@ -2986,6 +3023,29 @@ dependencies = [
  "rustc-hash 2.1.1",
 ]
 
+[[package]]
+name = "nvml-wrapper"
+version = "0.11.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0d5c6c0ef9702176a570f06ad94f3198bc29c524c8b498f1b9346e1b1bdcbb3a"
+dependencies = [
+ "bitflags 2.9.0",
+ "libloading",
+ "nvml-wrapper-sys",
+ "static_assertions",
+ "thiserror 1.0.69",
+ "wrapcenum-derive",
+]
+
+[[package]]
+name = "nvml-wrapper-sys"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dd23dbe2eb8d8335d2bce0299e0a07d6a63c089243d626ca75b770a962ff49e6"
+dependencies = [
+ "libloading",
+]
+
 [[package]]
 name = "objc"
 version = "0.2.7"
@@ -2996,6 +3056,25 @@ dependencies = [
  "objc_exception",
 ]
 
+[[package]]
+name = "objc2-core-foundation"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1c10c2894a6fed806ade6027bcd50662746363a9589d3ec9d9bef30a4e4bc166"
+dependencies = [
+ "bitflags 2.9.0",
+]
+
+[[package]]
+name = "objc2-io-kit"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "71c1c64d6120e51cd86033f67176b1cb66780c2efe34dec55176f77befd93c0a"
+dependencies = [
+ "libc",
+ "objc2-core-foundation",
+]
+
 [[package]]
 name = "objc_exception"
 version = "0.1.2"
@@ -4578,6 +4657,20 @@ dependencies = [
  "windows 0.52.0",
 ]
 
+[[package]]
+name = "sysinfo"
+version = "0.35.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3c3ffa3e4ff2b324a57f7aeb3c349656c7b127c3c189520251a648102a92496e"
+dependencies = [
+ "libc",
+ "memchr",
+ "ntapi",
+ "objc2-core-foundation",
+ "objc2-io-kit",
+ "windows 0.61.3",
+]
+
 [[package]]
 name = "system-configuration"
 version = "0.6.1"
@@ -5343,7 +5436,7 @@ dependencies = [
  "regex",
  "rustc_version",
  "rustversion",
- "sysinfo",
+ "sysinfo 0.30.13",
  "time",
 ]
 
@@ -5563,6 +5656,28 @@ dependencies = [
  "windows-targets 0.52.6",
 ]
 
+[[package]]
+name = "windows"
+version = "0.61.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9babd3a767a4c1aef6900409f85f5d53ce2544ccdfaa86dad48c91782c6d6893"
+dependencies = [
+ "windows-collections",
+ "windows-core 0.61.2",
+ "windows-future",
+ "windows-link",
+ "windows-numerics",
+]
+
+[[package]]
+name = "windows-collections"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3beeceb5e5cfd9eb1d76b381630e82c4241ccd0d27f1a39ed41b2760b255c5e8"
+dependencies = [
+ "windows-core 0.61.2",
+]
+
 [[package]]
 name = "windows-core"
 version = "0.52.0"
@@ -5584,15 +5699,26 @@ dependencies = [
 
 [[package]]
 name = "windows-core"
-version = "0.61.0"
+version = "0.61.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4763c1de310c86d75a878046489e2e5ba02c649d185f21c67d4cf8a56d098980"
+checksum = "c0fdd3ddb90610c7638aa2b3a3ab2904fb9e5cdbecc643ddb3647212781c4ae3"
 dependencies = [
  "windows-implement",
  "windows-interface",
  "windows-link",
- "windows-result 0.3.2",
- "windows-strings 0.4.0",
+ "windows-result 0.3.4",
+ "windows-strings 0.4.2",
+]
+
+[[package]]
+name = "windows-future"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fc6a41e98427b19fe4b73c550f060b59fa592d7d686537eebf9385621bfbad8e"
+dependencies = [
+ "windows-core 0.61.2",
+ "windows-link",
+ "windows-threading",
 ]
 
 [[package]]
@@ -5619,9 +5745,19 @@ dependencies = [
 
 [[package]]
 name = "windows-link"
-version = "0.1.1"
+version = "0.1.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "76840935b766e1b0a05c0066835fb9ec80071d4c09a16f6bd5f7e655e3c14c38"
+checksum = "5e6ad25900d524eaabdbbb96d20b4311e1e7ae1699af4fb28c17ae66c80d798a"
+
+[[package]]
+name = "windows-numerics"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9150af68066c4c5c07ddc0ce30421554771e528bde427614c61038bc2c92c2b1"
+dependencies = [
+ "windows-core 0.61.2",
+ "windows-link",
+]
 
 [[package]]
 name = "windows-registry"
@@ -5629,7 +5765,7 @@ version = "0.4.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4286ad90ddb45071efd1a66dfa43eb02dd0dfbae1545ad6cc3c51cf34d7e8ba3"
 dependencies = [
- "windows-result 0.3.2",
+ "windows-result 0.3.4",
  "windows-strings 0.3.1",
  "windows-targets 0.53.2",
 ]
@@ -5645,9 +5781,9 @@ dependencies = [
 
 [[package]]
 name = "windows-result"
-version = "0.3.2"
+version = "0.3.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c64fd11a4fd95df68efcfee5f44a294fe71b8bc6a91993e2791938abcc712252"
+checksum = "56f42bd332cc6c8eac5af113fc0c1fd6a8fd2aa08a0119358686e5160d0586c6"
 dependencies = [
  "windows-link",
 ]
@@ -5663,9 +5799,9 @@ dependencies = [
 
 [[package]]
 name = "windows-strings"
-version = "0.4.0"
+version = "0.4.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7a2ba9642430ee452d5a7aa78d72907ebe8cfda358e8cb7918a2050581322f97"
+checksum = "56e6c93f3a0c3b36176cb1327a4958a0353d5d166c2a35cb268ace15e91d3b57"
 dependencies = [
  "windows-link",
 ]
@@ -5777,6 +5913,15 @@ dependencies = [
  "windows_x86_64_msvc 0.53.0",
 ]
 
+[[package]]
+name = "windows-threading"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b66463ad2e0ea3bbf808b7f1d371311c80e115c0b71d60efc142cafbcfb057a6"
+dependencies = [
+ "windows-link",
+]
+
 [[package]]
 name = "windows_aarch64_gnullvm"
 version = "0.42.2"
@@ -5975,6 +6120,18 @@ dependencies = [
  "bitflags 2.9.0",
 ]
 
+[[package]]
+name = "wrapcenum-derive"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a76ff259533532054cfbaefb115c613203c73707017459206380f03b3b3f266e"
+dependencies = [
+ "darling 0.20.11",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.100",
+]
+
 [[package]]
 name = "write16"
 version = "1.0.0"
diff --git a/rust/Cargo.toml b/rust/Cargo.toml
index a828ec94..f5a9961d 100644
--- a/rust/Cargo.toml
+++ b/rust/Cargo.toml
@@ -44,6 +44,7 @@ log = "0.4.20"
 moshi = { path = "./moshi-core", version = "0.6.3" }
 native-tls = "0.2.11"
 numpy = "0.23.0"
+nvml-wrapper = "0.11.0"
 ogg = { version = "0.9.1", features = ["async"] }
 opus = "0.3.0"
 prometheus = "0.13.4"
@@ -63,6 +64,7 @@ serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0.115"
 sha3 = "0.10.8"
 symphonia = { version = "0.5.3", features = ["all"] }
+sysinfo = "0.35.2"
 tokenizers = "0.15.2"
 tokio = { version = "1.35.1", features = ["full"] }
 tokio-rustls = "0.24.1"
diff --git a/rust/moshi-server/Cargo.toml b/rust/moshi-server/Cargo.toml
index e02403a0..103c3704 100644
--- a/rust/moshi-server/Cargo.toml
+++ b/rust/moshi-server/Cargo.toml
@@ -25,6 +25,7 @@ lazy_static = { workspace = true }
 log = { workspace = true }
 moshi = { workspace = true }
 numpy = { workspace = true }
+nvml-wrapper = { workspace = true, optional = true }
 ogg = { workspace = true }
 opus = { workspace = true }
 prometheus = { workspace = true }
@@ -38,6 +39,7 @@ sentencepiece = { workspace = true }
 serde = { workspace = true }
 serde_json = { workspace = true }
 symphonia = { workspace = true }
+sysinfo = { workspace = true }
 tokio = { workspace = true }
 toml = { workspace = true }
 tower = { workspace = true }
@@ -52,5 +54,5 @@ vergen = { workspace = true }
 
 [features]
 default = []
-cuda = ["moshi/cuda", "candle/cuda", "candle-nn/cuda", "candle-transformers/cuda"]
+cuda = ["moshi/cuda", "candle/cuda", "candle-nn/cuda", "candle-transformers/cuda", "dep:nvml-wrapper"]
 metal = ["moshi/metal", "candle/metal", "candle-nn/metal", "candle-transformers/metal"]
diff --git a/rust/moshi-server/src/main.rs b/rust/moshi-server/src/main.rs
index d558d52c..b0a75eeb 100644
--- a/rust/moshi-server/src/main.rs
+++ b/rust/moshi-server/src/main.rs
@@ -2,6 +2,7 @@
 // This source code is licensed under the license found in the
 // LICENSE file in the root directory of this source tree.
 
+use crate::metrics::system::update_system_metrics;
 use anyhow::Result;
 use axum::{http::StatusCode, response::IntoResponse, response::Response};
 use candle::Device;
@@ -521,6 +522,14 @@ async fn main_() -> Result<()> {
                 app = app.merge(module.router(&shared_state)?)
             }
 
+            // Starts updating system metrics
+            tokio::task::spawn(async {
+                loop {
+                    update_system_metrics().await;
+                    tokio::time::sleep(tokio::time::Duration::from_millis(100)).await;
+                }
+            });
+
             let sock_addr = std::net::SocketAddr::from((
                 std::net::IpAddr::from_str(args.addr.as_str())
                     .unwrap_or(std::net::IpAddr::V6(std::net::Ipv6Addr::LOCALHOST)),
diff --git a/rust/moshi-server/src/metrics.rs b/rust/moshi-server/src/metrics.rs
index a6871940..9b901c7b 100644
--- a/rust/moshi-server/src/metrics.rs
+++ b/rust/moshi-server/src/metrics.rs
@@ -3,10 +3,133 @@
 // LICENSE file in the root directory of this source tree.
 
 use lazy_static::lazy_static;
+#[cfg(feature = "cuda")]
+use nvml_wrapper::{enum_wrappers::device::TemperatureSensor, error::NvmlError, Nvml};
 use prometheus::{
     histogram_opts, labels, opts, register_counter, register_gauge, register_histogram,
 };
 use prometheus::{Counter, Gauge, Histogram};
+use sysinfo::System;
+
+pub mod system {
+    use super::*;
+    #[cfg(feature = "cuda")]
+    lazy_static! {
+        pub static ref GPU_UTILIZATION: Gauge = register_gauge!(opts!(
+            "system_gpu_utilization",
+            "Utilization of the GPU. (percentage)",
+            labels! {"handler" => "all",}
+        ))
+        .unwrap();
+        pub static ref GPU_MEMORY_UTILIZATION: Gauge = register_gauge!(opts!(
+            "system_gpu_memory_utilization",
+            "Utilization of the GPU's memory. (percentage)",
+            labels! {"handler" => "all",}
+        ))
+        .unwrap();
+        pub static ref GPU_TEMPERATURE: Gauge = register_gauge!(opts!(
+            "system_gpu_temperature",
+            "Temperature of the GPU.",
+            labels! {"handler" => "all",}
+        ))
+        .unwrap();
+    }
+
+    lazy_static! {
+        pub static ref CPU_UTILIZATION: Gauge = register_gauge!(opts!(
+            "system_cpu_utilization",
+            "Utilization of the CPU. (percentage)",
+            labels! {"handler" => "all",}
+        ))
+        .unwrap();
+        pub static ref MEMORY_UTILIZATION: Gauge = register_gauge!(opts!(
+            "system_memory_utilization",
+            "Utilization of the system memory. (percentage)",
+            labels! {"handler" => "all",}
+        ))
+        .unwrap();
+    }
+
+    #[cfg(feature = "cuda")]
+    struct GpuMetrics {
+        utilization: f64,
+        total_memory: f64,
+        used_memory: f64,
+        temperature: f64,
+    }
+
+    #[cfg(feature = "cuda")]
+    fn get_gpu_metrics(nvml: &Nvml, device_idx: u32) -> Result<GpuMetrics, NvmlError> {
+        let dev = nvml.device_by_index(device_idx)?;
+        let memory_info = dev.memory_info()?;
+
+        Ok(GpuMetrics {
+            utilization: dev.utilization_rates()?.gpu as f64,
+            total_memory: memory_info.total as f64,
+            used_memory: memory_info.used as f64,
+            temperature: dev.temperature(TemperatureSensor::Gpu).map_or(f64::NAN, |x| x as f64),
+        })
+    }
+
+    #[cfg(feature = "cuda")]
+    fn get_gpus_metrics() -> Result<Option<GpuMetrics>, NvmlError> {
+        let nvml = Nvml::init()?;
+
+        let mut global_metrics = GpuMetrics {
+            utilization: 0.0,
+            total_memory: 0.0,
+            used_memory: 0.0,
+            temperature: f64::NAN,
+        };
+        let mut gpu_count: usize = 0;
+
+        for device_idx in 0..nvml.device_count()? {
+            match get_gpu_metrics(&nvml, device_idx) {
+                Ok(metrics) => {
+                    global_metrics.utilization += metrics.utilization;
+                    global_metrics.total_memory += metrics.total_memory;
+                    global_metrics.used_memory += metrics.used_memory;
+                    global_metrics.temperature = global_metrics.temperature.max(metrics.temperature);
+
+                    gpu_count += 1;
+                }
+                Err(err) => {
+                    tracing::debug!(?err, "couldn't get statistics on one gpu");
+                }
+            }
+        }
+
+        if gpu_count == 0 {
+            tracing::debug!("no gpu was found while collecting metrics");
+
+            Ok(None)
+        } else {
+            global_metrics.utilization /= (gpu_count as f64);
+            Ok(Some(global_metrics))
+        }
+    }
+
+    pub(crate) async fn update_system_metrics() {
+        #[cfg(feature = "cuda")]
+        {
+            match get_gpus_metrics() {
+                Ok(Some(metrics)) => {
+                    GPU_UTILIZATION.set(metrics.utilization);
+                    GPU_MEMORY_UTILIZATION.set(metrics.used_memory / metrics.total_memory);
+                    GPU_TEMPERATURE.set(metrics.temperature);
+                }
+                Err(err) => {
+                    tracing::debug!(?err, "error while collecting gpu statistics");
+                }
+                _ => ()
+            };
+        }
+
+        let sys = System::new_all();
+        CPU_UTILIZATION.set((sys.global_cpu_usage() as f64) / 100.0);
+        MEMORY_UTILIZATION.set((sys.used_memory() as f64) / (sys.total_memory() as f64));
+    }
+}
 
 pub mod asr {
     use super::*;