Looking forward to Cloud and AI Infrastructure 2026

With the Cloud and AI Infrastructure Event in London right around the corner, we’re looking forward to the topics on the agenda. The event is massive, with multiple tracks spanning cloud and AI infrastructure, DevOps, cybersecurity, big data, and data center–specific themes, so it’s important to focus on what will be most impactful for our clients. Here’s what we’re expecting to hear.

Scaling AI Infrastructure

Unsurprisingly, there are a number of sessions focused on AI infrastructure, including The Inference Imperative: A New Playbook for AI-First Infrastructure with Nirmal Ranganathan, CTO, Managed Public Cloud at Rackspace, AI Infrastructure at Enterprise Scale: Cloud, Cost, & Control, and Scalable AI Infrastructure – Structuring for Innovation with Charles Ewen, CIO at Met Office.

The first is intended to explore the “data gravity” principle: that smaller, more efficient language models running at the edge (next to your data rather than forcing it to travel) demand a fundamental infrastructure rethink as cost-per-inference becomes a competitive advantage. The second is a panel discussion among leaders in heavily regulated sectors on how they’re building AI-ready infrastructure that balances flexibility, cost, and compliance. The third is a use case in which Charles Ewen examines how the Met Office is using AI and how it’s changing the way they think about people, processes, and technology.

We’re especially interested in what speakers and attendees see as the biggest challenges created by the pace and scale of infrastructure growth, and we’re particularly keen to hear how the panelists plan to tackle them.

For example, deploying smaller nodes closer to the data, as Rackspace plans to discuss, resurrects a problem that hasn’t been as critical in recent years: the cost of delivering that payload. To be profitable, organizations need to ensure they aren’t deploying GPUs that sit underutilized. Even smaller sites can draw thousands of watts continuously, and when you add the cost of cooling, expenses can balloon quickly.

How are organizations planning to scale the ability to turn these nodes up or down, with all the GPUs and traffic controls involved, without losing control of cost and reliability? What does provisioning look like in practice? Identifying where individual APIs are being driven and how, whether DNS-based, latency-based, or influenced by a long list of other factors, is already a struggle for many companies. How are they planning to manage this complexity in a way that doesn’t restrict growth?

One way we’ve observed in the past is through dynamic pricing models such as “follow the sun” cost-averaging, where cloud providers stratify GPU cluster costs based on utilization at-a-time. There can be a standard, lower cost option that comes with a bit of a delay, with users switching to premium locations for time-sensitive workloads, and finally shift lower priority workloads to cheaper GPU cycles. This allows more accessibility and can be a factor in not restricting growth – but we’ll see what they have in mind.

Managing Hybrid Cloud Environments

This is a significant topic for us because it touches two of the top operational concerns our clients raise about network infrastructure. Redefining Hybrid Cloud with Chris Cowdry, Cloud Solutions Architect at Trust Systems, will cover how organizations can regain control of their cloud infrastructure, reduce lock-in, and build resilience through data sovereignty. Fleet at Planet Scale: Challenges in Managing Millions of Bare-Metal Servers by Igor Kanyuka, Staff Software Engineer at Meta, centers on end-to-end, automation-first lifecycle management for global environments.

As we said in our State of Network Infrastructure piece earlier this month, network infrastructure is simultaneously more capable and more fragile than many organizations would like to admit. The number one priority in addressing both issues is visibility: knowing where your infrastructure is, what it does, and where traffic is coming from and going is the foundational step to proper management. From there, you can prioritize nimble processes that automate manual work and improve system reliability. It’s also important to take these learnings and deployments and reinvest in institutional knowledge, but we’ll have more to say on that in a larger piece soon.

Looking Forward

As we head into the event, these sessions feel like the best window into what’s next, what’s actually working right now, and where the gaps still are. More than anything, we’re hoping the discussions move beyond polished takes and into the practical, with real lessons learned, honest trade-offs, and clear signals on what teams plan to do differently in the months ahead. We’ll be listening for the themes that connect the room, and we’ll share the most useful takeaways once the conversations unfold.