Add more detail

Author: RoseHJM

articles/dev-box/concept-serverless-gpu.md

@@ -65,7 +65,26 @@ Serverless GPU compute in Dev Box uses Azure Container Apps (ACA) to provide GPU
 
 The following GPU options are currently supported:
 
-- NVIDIA T4 GPUs
+- **NVIDIA T4 GPUs**: Readily available with minimal quota concerns
+- **NVIDIA A100 GPUs**: More powerful but available in limited capacity
+
+### Regional availability
+
+Currently, GPU resources are available in the following Azure regions:
+
+- West US 3
+- Sweden North
+- Australia East
+
+Additional regions may be supported in the future based on demand.
+
+### vNet injection
+
+vNet injection allows customers to integrate their network and security protocols with the serverless GPU environment. While not required for the proof of concept (POC), this feature will be prioritized for public previews and general availability (GA). With vNet injection, customers can achieve tighter control over network and security configurations.
+
+### MOBO architecture model
+
+Serverless GPU compute adopts the MOBO architecture model for ACA integration. In this model, ACA instances are created and managed within the customer’s subscription, providing a more controlled and streamlined management experience. This ensures that the Dev Box service can securely manage ACA sessions without introducing additional complexity.
 
 ### Developer experience
 
@@ -75,6 +94,8 @@ Developers can access serverless GPU compute through:
 - **Visual Studio**: Access GPU compute from within the Visual Studio environment
 - **VS Code with AI Toolkit**: Use seamless GPU integration for AI development tasks
 
+The goal is to provide a seamless, native experience where GPU resources are accessible without requiring any setup from the developer.
+
 ## Administration and management
 
 Administrators control serverless GPU access at the project level through Dev Center. Key management capabilities include:
@@ -83,6 +104,10 @@ Administrators control serverless GPU access at the project level through Dev Ce
 - **Set concurrent GPU limits**: Specify the maximum number of GPUs that can be used simultaneously across a project
 - **Cost controls**: Manage GPU usage within subscription quotas
 
+Access to serverless GPU resources is managed through project-level properties. When the serverless GPU feature is enabled for a project, all Dev Boxes within that project automatically gain access to GPU compute. This simplifies the access model by removing the need for custom roles or pool-based configurations.
+
+Future iterations of the project policy infrastructure will provide even more granular control over GPU access and usage.
+
 ## Related content
 
 - [Get started with serverless GPU in Dev Box (link to be added)]

articles/dev-box/source-serverless-gpu.md

@@ -167,61 +167,3 @@ Instant Access to GPU Compute: Dev Box allows developers to get up and running w
 Centralized Control for Admins: Dev Box integrates seamlessly with Dev Center's project policies, giving administrators granular control over serverless GPU access. Admins can define consumption limits, enable or disable GPU access on a per-project basis, and set permissions for users, all within the familiar Dev Center infrastructure. 
 
 Secure Private Network Integration: Dev Box runs within a private, enterprise-managed network. This ensures that sensitive corporate data used for AI workloads—such as proprietary models, internal datasets, or compliance-bound information—remains isolated and secure at the network layer. This added layer of security is crucial for enterprises handling regulated or confidential data. 
-
-POC Plan 
-
-Stage 1 – ETA 1-2 weeks – Eng: Nick Depinet 
-
-Develop a shell (Windows Terminal extension) that communicates with ACA and can be launched from within Dev Box. 
-
-AI Toolkit Integration 
-
-Checkpoint: Begin collection internal developer feedback on shell functionality and integration. 
-
-Stage 2 – ETA 2-3 weeks – Eng: Sneha 
-
-Implement Agent Management Service (AMS), handle authentication, session management, and related tasks. 
-
-Stage 3 – ETA 3-4 weeks 
-
-Introduce admin controls 
-
-HOBO provisioning 
-
-Begin planning for vNet injection support as a future enhancement. 
-
-Stage 4 – ETA 4-5 weeks 
-
-Finalize portal experience integration, enabling a seamless user interface for Dev Box users to manage GPU compute access. 
-
-Open questions 
-
-What is data persistency story? 
-
-What is the user experience around handling GPU limits per user? 
-
-How do we think about GPU pooling? 
-
-Where does the session pool live in dev center infra 
-
-Rude FAQ 
-
-Experience related 
-
-Why is the GPU accessible only as an external process? Why can't I use the GPU to accelerate my DevBox graphics? 
-
-Why do I have to request for GPU quota separately? Why can’t you auto-grant GPU quota to match the size of my Dev Box pool? 
-
-As an IT Admin for an Enterprise customer, why should I procure Serverless GPU through DevBox instead of directly procuring ACA Serverless GPU? 
-
-Current limitations / Roadmap related 
-
-Why can I only access GPUs via Shell? Why isn't there a GUI?  
-
-Why aren't you giving me the latest generation GPUs? I really need H100s 
-
-I need multiple GPUs attached to a single DevBox, why are you making me create multiple shells which get 1 GPU each instead of giving me N GPUs in a single shell? 
-
-I want to run Windows only software such as GameMaker on Serverless GPUs. Why am I limited to Linux only? 
-
-