
Choosing an AI model is now the easy part. The harder questions sit underneath: where it runs, how it is secured, and what it costs at scale. Without AI-ready infrastructure, even the best model stays stuck in a pilot.
Adoption is rising fast. The government’s AI Adoption Survey 2026 found that 52% of larger UK firms now use at least one AI technology. As use grows, so does pressure on compute, security, and data governance.
This blog is for CEOs, CFOs, CIOs, and CTOs deciding how to fund and scale AI. We look at the Azure updates from Microsoft Ignite 2026 and what they mean for your AI investment.
Today, many AI pilots succeed in a sandbox and then stall. The reasons are rarely about the model itself. Instead, teams hit limits on cost, data residency, security approval, or integration with existing systems.
Moreover, agents add new demands. They run code, call tools, and talk to other agents. Consequently, infrastructure must now isolate agent activity, enforce policy between services, and protect data in use.
In short, AI-ready infrastructure is about control as much as capacity. It decides whether security teams can approve a pilot for production.
Microsoft’s Ignite 2026 announcements, set out in the Book of News and its AI infrastructure blog, form four layers of an AI-ready foundation.
Azure Cobalt 200 virtual machines are now generally available. Microsoft cites up to 50% better CPU performance than Cobalt 100. They are rolling out to 24 Azure regions by year-end. In addition, new Mpsv4 and Lpsv5 types suit large in-memory databases and storage-heavy work.
At the same time, Microsoft pairs its own Maia and Cobalt chips with the latest NVIDIA and AMD hardware. According to Microsoft, it added more datacenter capacity in 18 months than in Azure’s first decade.
Agent Grid, in preview, gives agents, models, tools, and Azure services authenticated, encrypted connections. Notably, it applies identity-driven policies and default-deny network isolation. Similarly, the Agent Sandbox extension for AKS runs agent-written code with virtual-machine-level isolation for each session. For speed, it also offers pre-warmed pools and a serverless option.
In preview, customers can deploy open-weight models and agents from Microsoft Foundry to their own AKS clusters. Therefore, teams can choose GPUs and regions by cost, latency, and data residency. Meanwhile, a new AKS AI Manager places each model where GPU capacity can serve it best. Flex nodes, also in preview, extend AKS clusters to servers outside Azure.
Confidential computing previews now include multi-GPU protection for large models. They also include Azure Confidential Clean Room, which lets partners analyse shared data without exposing raw records. Additionally, Windows Server 2028 is in public preview, so organisations can run AI inference alongside business-critical applications on their own servers.
For example, Rockwell Automation uses AI in its factories to diagnose equipment issues and investigate product failures. Rockwell says that when an approach works in one plant, it can adapt it for the next instead of starting again. In short, consistent AI-ready infrastructure turns one success into many.
| Capability | What it does | Who benefits |
|---|---|---|
| Azure Cobalt 200 VMs (GA) | Delivers more efficient general-purpose compute | Infrastructure teams |
| Agent Grid (preview) | Secures and encrypts connections between agents, models, and tools | Security architects |
| Agent Sandbox for AKS (preview) | Isolates code that agents write and run, session by session | Platform engineers |
| Foundry models on your AKS (preview) | Runs models where cost and residency needs dictate | CIOs and data officers |
| AKS AI Manager | Places each model where GPU capacity serves it best | CTOs and platform teams |
| Flex nodes for AKS (preview) | Extends clusters to on-premises servers | Hybrid IT teams |
| Azure Confidential Clean Room (preview) | Lets partners analyse shared data without exposing raw records | Data and research teams |
| Windows Server 2028 (preview) | Runs AI inference alongside existing business applications | IT operations |
Trusts and integrated care systems want to share data for research and planning without exposing patient records. For instance, a confidential clean room could let partners run joint analytics while each keeps raw data private. We bring NHS DSPT Standards Met status for 2025 to 2026 to that design work.
Data residency and sovereignty shape every public sector AI decision. Therefore, running models on chosen regions and clusters gives departments more control. We help public bodies design hybrid architectures that satisfy security assurance.
Charities need AI without unpredictable cloud bills. In this case, efficient compute and right-sized clusters keep costs under control. We help non-profits plan capacity that grows with demand.
AI-ready infrastructure is the compute, security, and data foundation that lets AI move from pilot to production. It covers where models run, how agents are isolated, and how data stays protected. In other words, it decides whether AI is affordable, secure, and compliant at scale.
Not always. Many organisations use AI through managed services, such as Microsoft Foundry and Microsoft 365 Copilot, without owning GPUs. However, organisations running their own models need GPU capacity in the right regions. We help leaders decide which approach fits their workloads and budget.
In many cases, yes. In preview, Microsoft now lets customers deploy open-weight models and agents from Foundry to their own AKS clusters. Therefore, teams can choose regions by cost, latency, and data residency. We help organisations design architectures that meet their residency and assurance needs.
Confidential computing protects data while systems process it, not just when it is stored or sent. Microsoft’s latest previews extend this to large AI models across multiple GPUs. As a result, organisations can run AI on sensitive data with stronger protection against unauthorised access.
The next phase of AI will reward organisations with strong foundations, not just clever pilots. That said, infrastructure choices are long-lived, so they deserve careful planning. Above all, a clear view of workloads, data, and security needs comes first. From there, AI-ready infrastructure becomes a roadmap rather than a rebuild.
As a Microsoft Solutions Partner, we deliver Azure cloud, security, and managed services. We help organisations design AI-ready infrastructure that fits their sector. In turn, our Microsoft-certified team plans, builds, and runs these platforms over the long term.
At Mazik Global UK, we help organisations unlock the full potential of Microsoft technologies with AI-driven innovation and data-led strategies.
If you are ready to move AI from pilot to production, our team is here to help. Get in touch at mazikglobal.co.uk/contact-us.
Book a discovery call or send us a message.
We’ll get back to you within one business day.
A 30-minute call with our specialists to explore where AI and automation can make a real difference. There’s no sales pitch, just an honest conversation about your goals and challenges.
Tell us about your goals, and we’ll tailor our expertise to fit your needs. Fill out the form below, and we’ll get back to you soon.
Dynamics 365 Activate: How AI Is Changing ERP Migration and Modernisation ERP migrations have a reputation for running long and…
Building AI-Ready Infrastructure: What Businesses Need Beyond AI Models Choosing an AI model is now the easy part. The harder…
AI in Procurement: How Dynamics 365 Is Bringing Source-to-Pay Into the Tools People Already Use Procurement teams spend much of…
Mazik Global UK, a trusted Microsoft Solution Partner and FastTrack-recognised expert, delivers AI-powered, low-code solutions across Dynamics 365, Power Platform, and Azure, driving confident digital transformation.