diff --git a/app/backend/app.py b/app/backend/app.py
index 9911560227..6948a62cd1 100644
--- a/app/backend/app.py
+++ b/app/backend/app.py
@@ -6,8 +6,9 @@
 import os
 import time
 from pathlib import Path
-from typing import Any, AsyncGenerator, Dict, Union, cast
+from typing import Any, AsyncGenerator, Dict, List, Union, cast
 
+import httpx
 from azure.cognitiveservices.speech import (
     ResultReason,
     SpeechConfig,
@@ -25,6 +26,7 @@
 from azure.storage.filedatalake.aio import FileSystemClient
 from azure.storage.filedatalake.aio import StorageStreamDownloader as DatalakeDownloader
 from openai import AsyncAzureOpenAI, AsyncOpenAI
+from openai_priority_loadbalancer import AsyncLoadBalancer, Backend
 from opentelemetry.instrumentation.aiohttp_client import AioHttpClientInstrumentor
 from opentelemetry.instrumentation.asgi import OpenTelemetryMiddleware
 from opentelemetry.instrumentation.httpx import (
@@ -510,10 +512,22 @@ async def setup_clients():
 
         api_version = os.getenv("AZURE_OPENAI_API_VERSION") or "2024-03-01-preview"
 
+        client_args = {}
+        if AZURE_OPENAI_SERVICE_BACKEND2 := os.environ.get("AZURE_OPENAI_SERVICE_BACKEND2"):
+            backends: List[Backend] = [
+                Backend(f"{AZURE_OPENAI_SERVICE}.openai.azure.com", 1),
+                Backend(f"{AZURE_OPENAI_SERVICE_BACKEND2}.openai.azure.com", 1),
+            ]
+
+            lb = AsyncLoadBalancer(backends)
+            # Inject the load balancer as the transport in a new default httpx client
+            client_args["http_client"] = httpx.AsyncClient(transport=lb)
+
         openai_client = AsyncAzureOpenAI(
             api_version=api_version,
             azure_endpoint=endpoint,
             azure_ad_token_provider=token_provider,
+            **client_args,
         )
     elif OPENAI_HOST == "local":
         openai_client = AsyncOpenAI(
diff --git a/app/backend/requirements.in b/app/backend/requirements.in
index ad552cdae8..9a6f5b90a3 100644
--- a/app/backend/requirements.in
+++ b/app/backend/requirements.in
@@ -2,6 +2,7 @@ azure-identity
 quart
 quart-cors
 openai[datalib]>=1.3.7
+openai_priority_loadbalancer==1.1.0
 tiktoken
 tenacity
 azure-ai-documentintelligence
diff --git a/app/backend/requirements.txt b/app/backend/requirements.txt
index 2cf2c249b8..660063239e 100644
--- a/app/backend/requirements.txt
+++ b/app/backend/requirements.txt
@@ -204,6 +204,7 @@ openai[datalib]==1.30.5
     #   openai-messages-token-helper
 openai-messages-token-helper==0.1.4
     # via -r requirements.in
+openai-priority-loadbalancer==1.1.0
 opentelemetry-api==1.24.0
     # via
     #   azure-core-tracing-opentelemetry
diff --git a/infra/main.bicep b/infra/main.bicep
index 57e5a0b891..5dfba260d2 100644
--- a/infra/main.bicep
+++ b/infra/main.bicep
@@ -43,6 +43,7 @@ param appServiceSkuName string // Set in main.parameters.json
 @allowed([ 'azure', 'openai', 'azure_custom' ])
 param openAiHost string // Set in main.parameters.json
 param isAzureOpenAiHost bool = startsWith(openAiHost, 'azure')
+param deployAzureOpenAiBackendTwo bool = false
 param azureOpenAiCustomUrl string = ''
 param azureOpenAiApiVersion string = ''
 
@@ -310,6 +311,7 @@ module backend 'core/host/appservice.bicep' = {
       AZURE_OPENAI_GPT4V_MODEL: gpt4vModelName
       // Specific to Azure OpenAI
       AZURE_OPENAI_SERVICE: isAzureOpenAiHost ? openAi.outputs.name : ''
+      AZURE_OPENAI_SERVICE_BACKEND2: isAzureOpenAiHost && deployAzureOpenAiBackendTwo ? openAiBackendTwo.outputs.name : ''
       AZURE_OPENAI_CHATGPT_DEPLOYMENT: chatGpt.deploymentName
       AZURE_OPENAI_EMB_DEPLOYMENT: embedding.deploymentName
       AZURE_OPENAI_GPT4V_DEPLOYMENT: useGPT4V ? gpt4vDeploymentName : ''
@@ -401,6 +403,23 @@ module openAi 'core/ai/cognitiveservices.bicep' = if (isAzureOpenAiHost) {
   }
 }
 
+module openAiBackendTwo 'core/ai/cognitiveservices.bicep' = if (isAzureOpenAiHost && deployAzureOpenAiBackendTwo) {
+  name: 'openai-backend-two'
+  scope: openAiResourceGroup
+  params: {
+    name: '${abbrs.cognitiveServicesAccounts}${resourceToken}-b2'
+    location: openAiResourceGroupLocation
+    tags: tags
+    publicNetworkAccess: publicNetworkAccess
+    bypass: bypass
+    sku: {
+      name: openAiSkuName
+    }
+    deployments: openAiDeployments
+    disableLocalAuth: true
+  }
+}
+
 // Formerly known as Form Recognizer
 // Does not support bypass
 module documentIntelligence 'core/ai/cognitiveservices.bicep' = {
@@ -827,6 +846,7 @@ output AZURE_OPENAI_GPT4V_MODEL string = gpt4vModelName
 
 // Specific to Azure OpenAI
 output AZURE_OPENAI_SERVICE string = isAzureOpenAiHost ? openAi.outputs.name : ''
+output AZURE_OPENAI_SERVICE_BACKEND2 string = isAzureOpenAiHost && deployAzureOpenAiBackendTwo ? openAiBackendTwo.outputs.name : ''
 output AZURE_OPENAI_API_VERSION string = isAzureOpenAiHost ? azureOpenAiApiVersion : ''
 output AZURE_OPENAI_RESOURCE_GROUP string = isAzureOpenAiHost ? openAiResourceGroup.name : ''
 output AZURE_OPENAI_CHATGPT_DEPLOYMENT string = isAzureOpenAiHost ? chatGpt.deploymentName : ''
diff --git a/infra/main.parameters.json b/infra/main.parameters.json
index 541bc4cfff..9115cf28c2 100644
--- a/infra/main.parameters.json
+++ b/infra/main.parameters.json
@@ -119,6 +119,9 @@
     "openAiHost": {
       "value": "${OPENAI_HOST=azure}"
     },
+    "deployAzureOpenAiBackendTwo": {
+      "value": "${DEPLOY_AZURE_OPENAI_BACKEND_TWO=false}"
+    },
     "azureOpenAiCustomUrl":{
       "value": "${AZURE_OPENAI_CUSTOM_URL}"
     },