diff --git a/CHANGELOG.md b/CHANGELOG.md
index 2c671913f26..0419f9b91ee 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -27,6 +27,8 @@ release.
   ([#2154](https://github.com/open-telemetry/opentelemetry-specification/pull/2154))
 - Mark In-memory, OTLP and Stdout exporter specs as Stable.
   ([#2175](https://github.com/open-telemetry/opentelemetry-specification/pull/2175))
+- Use UCUM units in Metrics Semantic Conventions.
+  ([#2199](https://github.com/open-telemetry/opentelemetry-specification/pull/2199))
 
 ### Logs
 
diff --git a/specification/metrics/semantic_conventions/faas-metrics.md b/specification/metrics/semantic_conventions/faas-metrics.md
index 3ff5b56d3f3..67dcf534e2e 100644
--- a/specification/metrics/semantic_conventions/faas-metrics.md
+++ b/specification/metrics/semantic_conventions/faas-metrics.md
@@ -32,22 +32,22 @@ type and units.
 
 Below is a table of FaaS invocation metric instruments.
 
-| Name | Instrument | Units | Description |
-|------|------------|----|-------------|
-| `faas.invoke_duration` | Histogram | milliseconds | Measures the duration of the invocation |
-| `faas.init_duration` | Histogram | milliseconds | Measures the duration of the function's initialization, such as a cold start |
-| `faas.coldstarts` | Counter | default unit | Number of invocation cold starts. |
-| `faas.errors` | Counter | default unit | Number of invocation errors. |
-| `faas.executions` | Counter | default unit  | Number of successful invocations. |
-| `faas.timeouts` | Counter | default unit | Number of invocation timeouts. |
+| Name | Instrument | Unit | Unit ([UCUM](README.md#instrument-units)) | Description |
+|------|------------|------|-------------------------------------------|-------------|
+| `faas.invoke_duration` | Histogram | milliseconds | `ms` | Measures the duration of the invocation |
+| `faas.init_duration` | Histogram | milliseconds | `ms` | Measures the duration of the function's initialization, such as a cold start |
+| `faas.coldstarts` | Counter | default unit | `{coldstarts}` | Number of invocation cold starts. |
+| `faas.errors` | Counter | default unit | `{errors}` | Number of invocation errors. |
+| `faas.executions` | Counter | default unit | `{executions}` | Number of successful invocations. |
+| `faas.timeouts` | Counter | default unit | `{timeouts}` | Number of invocation timeouts. |
 
 Optionally, when applicable:
 
-| Name | Instrument | Units | Description |
-|------|------------|----|-------------|
-| `faas.mem_usage` | Histogram | bytes | Distribution of max memory usage per invocation |
-| `faas.cpu_usage` | Histogram | milliseconds | Distribution of cpu usage per invocation |
-| `faas.net_io` | Histogram | bytes | Distribution of net I/O usage per invocation |
+| Name | Instrument | Unit | Unit ([UCUM](README.md#instrument-units)) | Description |
+|------|------------|------|-------------------------------------------|-------------|
+| `faas.mem_usage` | Histogram | Bytes | `By` | Distribution of max memory usage per invocation |
+| `faas.cpu_usage` | Histogram | milliseconds | `ms` | Distribution of cpu usage per invocation |
+| `faas.net_io` | Histogram | Bytes | `By` | Distribution of net I/O usage per invocation |
 
 ## Attributes
 
diff --git a/specification/metrics/semantic_conventions/http-metrics.md b/specification/metrics/semantic_conventions/http-metrics.md
index a24e2bbb5cd..f664e97b95e 100644
--- a/specification/metrics/semantic_conventions/http-metrics.md
+++ b/specification/metrics/semantic_conventions/http-metrics.md
@@ -17,18 +17,18 @@ type and units.
 
 Below is a table of HTTP server metric instruments.
 
-| Name                          | Instrument                 | Units        | Description |
-|-------------------------------|----------------------------|--------------|-------------|
-| `http.server.duration`        | Histogram                  | milliseconds | measures the duration of the inbound HTTP request |
-| `http.server.active_requests` | Asynchronous UpDownCounter | requests     | measures the number of concurrent HTTP requests that are currently in-flight |
+| Name                          | Instrument                 | Unit         | Unit ([UCUM](README.md#instrument-units)) | Description |
+|-------------------------------|----------------------------|--------------|-------------------------------------------|-------------|
+| `http.server.duration`        | Histogram                  | milliseconds | `ms`                                      | measures the duration of the inbound HTTP request |
+| `http.server.active_requests` | Asynchronous UpDownCounter | requests     | `{requests}`                              | measures the number of concurrent HTTP requests that are currently in-flight |
 
 ### HTTP Client
 
 Below is a table of HTTP client metric instruments.
 
-| Name                   | Instrument | Units        | Description |
-|------------------------|------------|--------------|-------------|
-| `http.client.duration` | Histogram  | milliseconds | measure the duration of the outbound HTTP request |
+| Name                   | Instrument | Unit         | Unit ([UCUM](README.md#instrument-units)) | Description |
+|------------------------|------------|--------------|-------------------------------------------|-------------|
+| `http.client.duration` | Histogram  | milliseconds | `ms`                                      | measure the duration of the outbound HTTP request |
 
 ## Attributes
 
diff --git a/specification/metrics/semantic_conventions/rpc.md b/specification/metrics/semantic_conventions/rpc.md
index 97f47586023..12c6d57b939 100644
--- a/specification/metrics/semantic_conventions/rpc.md
+++ b/specification/metrics/semantic_conventions/rpc.md
@@ -31,26 +31,26 @@ MUST be of the specified type and units.
 
 Below is a table of RPC server metric instruments.
 
-| Name | Instrument | Units | Description | Status | Streaming |
-|------|------------|-------|-------------|--------|-----------|
-| `rpc.server.duration` | Histogram  | milliseconds | measures duration of inbound RPC | Recommended | N/A.  While streaming RPCs may record this metric as start-of-batch to end-of-batch, it's hard to interpret in practice. |
-| `rpc.server.request.size` | Histogram  | bytes | measures size of RPC request messages (uncompressed) | Optional | Recorded per message in a streaming batch |
-| `rpc.server.response.size` | Histogram  | bytes | measures size of RPC response messages (uncompressed) | Optional | Recorded per response in a streaming batch |
-| `rpc.server.requests_per_rpc` | Histogram  | count | measures the number of messages received per RPC.  Should be 1 for all non-streaming RPCs | Optional | Required |
-| `rpc.server.responses_per_rpc` | Histogram  | count | measures the number of messages sent per RPC.  Should be 1 for all non-streaming RPCs | Optional | Required |
+| Name | Instrument | Unit | Unit ([UCUM](README.md#instrument-units)) | Description | Status | Streaming |
+|------|------------|------|-------------------------------------------|-------------|--------|-----------|
+| `rpc.server.duration` | Histogram  | milliseconds | `ms` | measures duration of inbound RPC | Recommended | N/A.  While streaming RPCs may record this metric as start-of-batch to end-of-batch, it's hard to interpret in practice. |
+| `rpc.server.request.size` | Histogram  | Bytes | `By` | measures size of RPC request messages (uncompressed) | Optional | Recorded per message in a streaming batch |
+| `rpc.server.response.size` | Histogram  | Bytes | `By` | measures size of RPC response messages (uncompressed) | Optional | Recorded per response in a streaming batch |
+| `rpc.server.requests_per_rpc` | Histogram  | count | `{count}` | measures the number of messages received per RPC.  Should be 1 for all non-streaming RPCs | Optional | Required |
+| `rpc.server.responses_per_rpc` | Histogram  | count | `{count}` | measures the number of messages sent per RPC.  Should be 1 for all non-streaming RPCs | Optional | Required |
 
 ### RPC Client
 
 Below is a table of RPC client metric instruments.  These apply to traditional
 RPC usage, not streaming RPCs.
 
-| Name | Instrument | Units | Description | Status | Streaming |
-|------|------------|-------|-------------|--------|-----------|
-| `rpc.client.duration` | Histogram | milliseconds | measures duration of outbound RPC | Recommended | N/A.  While streaming RPCs may record this metric as start-of-batch to end-of-batch, it's hard to interpret in practice. |
-| `rpc.client.request.size` | Histogram | bytes | measures size of RPC request messages (uncompressed) | Optional | Recorded per message in a streaming batch |
-| `rpc.client.response.size` | Histogram | bytes | measures size of RPC response messages (uncompressed) | Optional | Recorded per message in a streaming batch |
-| `rpc.client.requests_per_rpc` | Histogram | count | measures the number of messages received per RPC.  Should be 1 for all non-streaming RPCs | Optional | Required |
-| `rpc.client.responses_per_rpc` | Histogram | count | measures the number of messages sent per RPC.  Should be 1 for all non-streaming RPCs | Optional | Required |
+| Name | Instrument | Unit | Unit ([UCUM](README.md#instrument-units)) | Description | Status | Streaming |
+|------|------------|------|-------------------------------------------|-------------|--------|-----------|
+| `rpc.client.duration` | Histogram | milliseconds | `ms` | measures duration of outbound RPC | Recommended | N/A.  While streaming RPCs may record this metric as start-of-batch to end-of-batch, it's hard to interpret in practice. |
+| `rpc.client.request.size` | Histogram | Bytes | `By` | measures size of RPC request messages (uncompressed) | Optional | Recorded per message in a streaming batch |
+| `rpc.client.response.size` | Histogram | Bytes | `By` | measures size of RPC response messages (uncompressed) | Optional | Recorded per message in a streaming batch |
+| `rpc.client.requests_per_rpc` | Histogram | count | `{count}` | measures the number of messages received per RPC.  Should be 1 for all non-streaming RPCs | Optional | Required |
+| `rpc.client.responses_per_rpc` | Histogram | count | `{count}` | measures the number of messages sent per RPC.  Should be 1 for all non-streaming RPCs | Optional | Required |
 
 ## Attributes