Incident sheet · /infrastructure

Filed
Read
4 min · 860 words
Severity
Routine

Microsoft Ignite: Agent Grid, ContextCache and AKS-Foundry Bridge Land on Azure

Microsoft used Ignite to launch Agent Grid, ContextCache, Azure Catalyst and a Foundry-to-AKS bridge. Azure claims it loads thousands of agent microVMs in under half a second.

Microsoft’s Azure upgrades add speed and security to agents
FeedMicrosoft’s Azure upgrades add speed and security to agents — Zulfugar Karimov / unsplash

Event log

  1. Azure can load thousands of agent microVMs in less than half a second, Microsoft claims.
  2. OpenAI discounts cached input by up to 95% and Anthropic by 90% or more, per Greyhound Research.
  3. Foundry deployed a Llama 3.3 70B model on AKS; an agent diagnosed a machinery fault in four minutes instead of two hours.
  4. The new Agent Grid, ContextCache and AKS-Foundry features are in preview.
  5. Azure Catalyst modernizes legacy on-premises apps with AI agents.

Microsoft on Tuesday upgraded Azure's infrastructure to speed up and contain AI agents, launching Agent Grid, ContextCache, Azure Catalyst and a bridge from its managed Foundry service to self-managed Kubernetes clusters. The vendor announced the changes at its Ignite conference. The core services isolate agents in sandboxes and make them more efficient.

Mark Russinovich, CTO and deputy chief information security officer at Microsoft Azure, framed the security problem directly. "Agents are insider risks. We need ways to isolate them and control them," Russinovich said during a keynote at Ignite.

The sandbox is wrapped by two services:

  • Agent Grid, a new service controlling which network resources agents can reach.
  • ContextCache, which saves an agent's work so it does not have to start over on a new server.

How fast are the new microVMs?

Microsoft runs agents in mini-computers called microVMs, which lock down agents and keep them from harming outside systems. The microVMs start on demand and shut down when no longer needed. Russinovich claimed Azure can load thousands of them in less than half a second.

Google and AWS offer similar microVM technology to contain agents. "This capability really should be seen as a minimum agentic infrastructure requirement," said Jack Gold, principal analyst at J.Gold Associates.

Agent Grid adds a protection layer around the microVMs. "If it can't access a certain website or you want it to access only a subset of websites, you can enforce those policies right there in the network right next to the agent sandbox," Russinovich said.

Isolation has limits, analysts cautioned. It stops agent-written code from escaping onto the host, but "it does nothing about an agent misusing access it was legitimately given," said Sanchit Vir Gogia, CEO of Greyhound Research.

What does ContextCache save?

ContextCache, first shown earlier this year and detailed at Ignite, saves the context of a long-running agent so it can be reused after a reload on another server. Long-running agents on GPUs reuse context instead of recomputing it. "We need to be able to feed these models so that they ... save money by not having to recompute work," Russinovich said.

Gold called it an important capability, noting shared knowledge between agents and agent runs keeps processing costs down and speeds output. Microsoft cited savings but gave no figure. For comparison, Greyhound Research found OpenAI discounts cached input by up to 95% and Anthropic by 90% or more. "Those discounts apply only to the repeated part of the input, never to the answer the model writes," Gogia said.

Can enterprises mix managed and self-managed AI?

Microsoft also intervened in the build-versus-buy debate over self-managed versus rented AI. Customers can now connect self-managed Azure Kubernetes Service (AKS) clusters to Foundry, Microsoft's managed AI service, and deploy open-weight models and agents from Foundry onto their own clusters, the company said in a statement.

"You can do kind of the middle where you can have Foundry managing things on top of AKS," Russinovich said.

In his demonstration, a platform engineering team set up an AKS cluster with security controls and open-source components. Foundry deployed a Llama 3.3 70B model onto the cluster. An agent built on it diagnosed a machinery issue, using a GPU node with a camera on the factory floor added through AKS flex nodes. The agent diagnosed the fault in four minutes instead of the two hours a human visit would require.

Gogia said the feature "ends a false choice between a fully managed service and a do-it-yourself stack." Teams can start managed and move to their own clusters without changing tools, he said.

What else launched?

Microsoft also introduced Azure Catalyst, which modernizes legacy on-premises applications in the cloud with AI agents. "Azure Catalyst takes a bunch of components that we've already had and brings them together with agents, completing this flow from end to end," Russinovich said.

New VM instances support confidential computing (hardware isolation of data in use) on newer Intel and AMD chips. The instances create a black box around data and applications and require attestation, a verification step, before outside applications gain access.

Russinovich described the combined releases as pushing Azure toward "one connected system, from the silicon all the way up to the platform."

Should CIOs commit now?

Analysts advised caution on lock-in. "No hyperscaler stays ahead of the others for very long," Gold said, adding that the other major clouds have similar capabilities. Azure-specific tools can lock down customers just as AWS and Google tools can, he noted.

CIOs should buy what removes operational work the organization has little reason to own, and build where requirements or business advantage justify maintaining it, said Stephanie Walter, practice leader for the AI stack at HyperFrame Research. "Before committing, establish what it would take to move the agent's state, tools, permissions and evaluations elsewhere. Model choice alone doesn't make an application portable," Walter said.

The new features are in preview. "Platform teams should pilot [the new features] now and hold production commitments until the documentation and service terms arrive," Gogia said.

via techtarget.com (Source)

Filed under

  • microsoft-azure
  • ai-agents
  • microsoft-foundry
  • azure-kubernetes-service
  • ai-agent-security
Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Senior reporter covering money, funding and market shifts around enterprise information technology and the leaders who run at CIO Monitor.

199 articles

The newsroom digest

One email a day with the stories that matter. No noise.

By subscribing you agree to the privacy policy.

Cross-references

‹ Previous articleNext article ›