Stelia

6 weeks until @Cannes_Lions. We'll be there, will you? Cannes is where the boldest ideas in media and advertising come together. We'll be there showcasing how forward-thinking organisations are turning ambitious thinking into outcomes with production-grade AI. Not prototypes, but reliable, governable, and cost-effective AI that can be operationalised at scale. Our recently launched Stelia AI OS is providing a significant opportunity for organisations: faster time to market, new revenue streams, and a competitive edge that compounds with every campaign cycle. All built on proprietary data that never leaves your environment. If you're a business leader at Cannes thinking about how AI translates into revenue, efficiency, and advantage, we'd love to meet with you. Set up a meeting here with our VP of Media and Entertainment Ula Nairne and VP Applied AI Paul Heathcote: platform.stelia.ai/canneslions Read more about the type of outcomes being delivered on the Stelia OS Platform here: newsroom.stelia.ai/introducing-st… Looking forward to joining discussions with fellow innovators, including companies such as @GoogleDeepMind, @OpenAI, @madebymonks_ alongside brilliant speakers like Adam Mosseri from @Meta, David Sandstrom from @Klarna and Grace Kao from @Snap Inc. Watch this space for some exciting project news with a major media company coming soon. #CannesLions2026 #FutureOfMedia #MediaAI

Most enterprise AI stacks are being built around a flawed assumption: that a single agent, given enough context, can reliably orchestrate and execute complex workflows at scale. But in reality, that context has limits, errors can quickly multiply, and the consequences of each mistake grow with the complexity of the task. Stelia AI OS was built on a different principle: that the most effective AI systems aren't built around broad agents – they're the ones that use AI purposefully, at the task level, within well-governed workflows. The result is enterprise-grade AI that delivers genuine business value in production settings – governed, observable, and significantly more cost-efficient than the single agent alternative. Our VP of Software Engineering, James Dobson makes the case for why it's time to rethink what we mean by agentic AI in his latest piece: newsroom.stelia.ai/its-time-to-re… Learn more about how Stelia AI OS puts these principles into practice: newsroom.stelia.ai/introducing-st… #AIagents #AgenticAI #EnterpriseAI

Today we are proud to launch Stelia AI OS: the full-stack operating system for enterprise AI. Our team has experienced first-hand the challenges of getting AI into production at scale for some of the world's most complex organisations. Fragmented tooling, costly integration, and governance that's nearly impossible to maintain end-to-end. That's why Stelia AI OS was built. It's not another orchestration layer or model gateway. It replaces the need to assemble and maintain an entire AI stack in the first place. It enables you to move at commercial speed, scale to enterprise demand, and deploy the latest innovations safely, compliantly, and cost-effectively while staying under your control. The result: ▶️ 5x faster speed to market. ▶️ 50% reduction in LLM costs. ▶️ 3x engineer productivity. ▶️ An architecture that’s built to evolve with the pace of AI. ▶️ High quality business outcomes, new revenue streams, diversified products. One platform, multiple capabilities at your disposal, such as: ▶️ One interface to access all models: switch providers, upgrade models, select regions without touching code. ▶️ Create complex agent workflows, run them on any model instantly. ▶️ Set the guardrails you need. Model agnostic, out-of-the-box. ▶️ Track token usage, latency, performance, live in one dashboard. ▶️ Create a native knowledge graph, build institutional memory. ▶️ Fine tune LLMs, image and video generation models on your proprietary data – no deep data science expertise required. And much more. With no vendor lock-in Most importantly, we’re running Stelia AI OS internally to power real systems that need enterprise-grade reliability, quality, and trust. More on how Stelia AI OS can accelerate your speed to market here: newsroom.stelia.ai/introducing-st… Speak to the team or request a demo: aios.stelia.ai/form

Stelia’s Chief Legal Officer, Imogen Armstrong, will be speaking at @GetJuro's Scaleup GC: The Operator Era next week at County Hall, London. Imogen will be joining Tom Draper from Coalition Inc., Kit Lewin from Capsule Insurance, and Laura Brodahl from @wilsonsonsini on a panel discussing, Cyber Attack: What Investors and Boards Demand from the General Counsel. When a breach hits, the GC becomes the focal point - fielding pressure from all sides. With cyber claims against European scaleups on the rise, this session will bring together in-house counsel, external counsel, and insurance experts to share the "war room" strategies you need to protect your company's reputation and capital. This event will also feature brilliant speakers such as @maryshenocarro1 , former president of CLOC, Elena Leichty, Ethan Forrest, Ronan Davy and Broc Cocoman all from @AnthropicAI, Emily Turner, CEO of @HSBCInnovation Banking UK and Antti Innanen, author of ‘Prompted’. Thank you Richard Mabey and the Juro team for organising such a great event. For more information, see here: juro.com/scaleup-26 #ScaleupGC #CyberSecurity #LegalLeadership

Introducing AI, Stacked: a new monthly brief connecting what’s working today with where the future is being built. From practical insights for engineering teams, to strategic perspectives for business leaders, we’ll cover what’s moving at the frontier, why it matters, and what most people are missing when it comes to building production-grade AI for durable business advantage. Written by Stelia’s full-stack AI team, this is grounded in real-world delivery: from infrastructure to agents, through to measurable business outcomes, across some of the most demanding production environments. We’ll be focusing on areas such as: ▶️ How to design AI systems that scale reliably in production, from the ground up ▶️ Case studies of how agent workflows are reshaping enterprise operations across industries — from advertising and media, to retail, space, and more ▶️ The governance standards that will define who survives the coming wave of AI accountability ▶️ The latest developments in the ecosystem - and what they mean in practice, beyond the hype Sign up here to get this delivered straight to your inbox: mailchi.mp/stelia/newsroo…

We are pleased to be joining our partners @nvidia and @Lenovo next week for a special event bringing our AI ecosystem together. This forms part of a number of exciting projects we currently have underway. The event will be taking place at the De Vere Wokefield Estate, with the Stelia team stationed at the 12th hole for our PAR 3 challenge, where we have a few surprises up our sleeves. Keep an eye out for the Stelia buggy and be sure to grab some of our custom golf balls! We look forward to welcoming teams from @nvidia | @Lenovo | @ReadingFC | @RFCCommunity | @Computacenter | @BellIntegration | @Econocom_UK | @VAST_Data | @Scenegraph_UK | @QAOps | @devereofficial | @NHSuk | @nscale | @SkyUK | XSAIA | @IWFhotline | @wwt_inc And all for a great cause with the Berkshire Animal Connection Centre. We look forward to seeing you there.

Not long until Stelia CTO David Hughes speaks at @Leaseweb's Berlin Tech Meetup, tackling a question many CTOs find themselves asking: Do we really need all these GPUs? The short answer: it depends. Certain use cases absolutely demand GPU-heavy setups. But too often, organisations jump to maximum capacity without understanding what their application actually requires, resulting in significant overprovisioning and avoidable cost. Dave will dive into the approaches organisations can take to effectively run their AI workloads across distributed compute, and when high-cost GPU clusters aren’t necessarily needed. We're excited to be joining a fantastic lineup, including Ömer KARABACAK from Youzu and Felix Meyer from @cloudeteer, with @ArthurPoot from AI Builders moderating, bringing together valuable perspectives on what it takes to build resilient AI infrastructure at scale. A big thank you to the Leaseweb team for organising and continuing to support the Berlin AI community. Secure your spot here before it’s fully booked: luma.com/klg1zbe2 Ahead of his session, check out Dave’s recent piece on how different workloads demand entirely different infrastructure strategies and how to avoid over investment by understanding behaviour first: lnkd.in/e3rYQ8PA #Leaseweb #AIInfrastructure #AICommunity

Most cloud providers over-provision hardware. Your 20 vCPUs are actually oversubscribed with a dozen other tenants. At Stelia, we decided that wasn't good enough for AI infrastructure. When you're running a large model inference workload, the difference between a CPU that's truly yours and one that's being time-shared can be a 15–30% performance gap — sometimes more. That's the difference between a product that feels responsive and one that doesn't. We built our compute instances to avoid this entirely. It's a lot of complex plumbing, but the outcome is simple: when you get 20 vCPUs from us, they're genuinely yours. As a company, we never opt for the easiest and fastest route. We invest in complex engineering to ensure our technology is future-proofed, flexible, and highly performant for our customers in what is an incredibly fast-moving landscape. In our latest technical piece, Senior Platform Engineer Peter Bangert talks through the hands-on implementation — from configuring @kubernetesio Topology Manager to building a custom cgroup controller that pins vCPU threads and enforces NUMA-aware scheduling. Read it here: dev.to/steliadevs/no-… #NUMA #CPUs #AIinfrastructure #cloudcompute

Following our recent collaboration with @nokia, Stelia’s Lukas Stockner has been invited to speak at Nokia Wavelengths in Madrid next week. Our Principal Platform Engineer will join Alberto Carrasco Colomer from Telefonica and Aleksandar Ristovski from @exainfra for the “Advancing Connectivity” roundtable, moderated by @DoironTim from Nokia, bringing together perspectives from across telecoms and infrastructure. The session will be an open discussion, exploring both how networks need to evolve to support AI-driven workloads, and how AI itself can be applied to optimise and operate those networks. With AI placing new demands on infrastructure, we look forward to being part of the conversation that will focus on what it takes to build resilient, scalable networks ready for the next generation of compute. Thank you to Anthony I'Anson, Hélène Zalkind, Gillian Hanna, @jonasvermeulen and @RolandT for organising such an interesting event. Read more about our collaboration with Nokia here: newsroom.stelia.ai/stelia-and-nok… See the full agenda for Nokia Wavelengths EMEA-LAT here: nokia.com/events/wavelen… #Nokia #AIInfrastructure #Connectivity

At Stelia, we pride ourselves in having an engineering toolkit that spans some of the most advanced languages, frameworks and capabilities. Of these, systems programming language Rust has earned a specific place in our system design for its reliability, precision, and stability at scale. Gary Tierney, a key member of our engineering team, has unpacked what Rust is; why we use it to enable robust, production-scale system design; and equally where a different approach is simply the better fit. Read the full blog here: newsroom.stelia.ai/what-rust-is-a… #AIdevelopment #AIinfrastructure #AIarchitecture #Rust #ProgrammingLanguage

Last month, Stelia’s engineering team joined the @cloudfest community in Europa-Park, Germany. It was great to connect with friends and peers across the ecosystem such as @nokia, @Leaseweb and Neo Networking Consultants. Having recently announced our collaboration with @nokia, it was brilliant to deep-dive into the latest developments on their Event Driven Automation (EDA) platform - a cloud-native network management solution built on @kubernetesio that treats all networking hardware as nodes. This has great synergies with Stelia's own architecture. Peter Bangert and Lukas Stockner also had the pleasure of taking part in @Leaseweb's VIP dinner, strengthening long-standing relationships ahead of David Hughes speaking at their Tech Meetup at the end of this month. Across the event, it was clear that while technologies continue to evolve, it's the exchange of ideas and openness that’s driving progress in the ecosystem. Read more about our collaboration with Nokia here: newsroom.stelia.ai/stelia-and-nok… Thank you to @cloudfest for bringing the community together, and the brilliant speakers including Soeren von Varchmin, Christian Jaeger and @rogerrohatgi. #CloudFest26 #cloudcomputing

"GPU cloud" is now common language in AI, but even among technical leaders, what actually underpins these systems and how a GPU cloud is actually designed remains surprisingly unclear. Today we are launching a new series to build this foundational understanding. From first principles through to deep dives of each layer of the modern AI stack, we're breaking down what it actually takes to build robust, reliable cloud infrastructure for AI. And we're starting at the beginning, as Stelia's David Hughes defines what constitutes a true cloud, why GPU environments fundamentally change the design constraints, and the systems-level thinking required to build these environments at scale. Read the first piece here: newsroom.stelia.ai/understanding-… More to follow over the coming weeks as we unpack the components that drive both technical performance and business outcomes at scale. #AIinfrastructure #Cloudinfrastructure #GPUcloud #AIsystems #AIengineering

How does @Google's newly released TurboQuant compression technique change how much memory, hardware, and cost it takes to run modern AI systems? Memory has become one of the biggest constraints in scaling AI. LLMs are extraordinarily memory-hungry, and that has a direct impact on what it costs to deploy them at scale. Stelia's Tristan Beecroft has unpacked what TurboQuant actually is, what it means for the economics of running AI in production, and what real-world adoption of capabilities like this could look like for AI deployment in practice. Read the full breakdown here: newsroom.stelia.ai/llms-are-memor… #AIinfrastructure #AImemory #Quantisation #LLM

Stelia has a new home on London's South Bank, just in time for the spring sunshine. Our UK team are settling in well and it's great to be right at the heart of such a vibrant AI ecosystem, surrounded by energy that inspires what we build every day. We have some amazing facilities and we're excited to start welcoming customers, partners, and friends of the business through our doors. If you're nearby, don't hesitate to drop us a line, we’d love to host you. #AIecosystem #AIcommunity

We're collaborating with @nokia to advance the deployment of high-performance, high-trust AI at enterprise scale. Stelia is embedded in some of the world’s most data-intensive enterprise environments, serving as the operational backbone for governed, production-quality AI. By integrating Nokia's leading open-standards-based networking technology into its AI ecosystem, Stelia is enabling reliable performance across distributed systems - built for the most demanding business use cases. “Our collaboration helps organisations support demanding AI workloads while maintaining the trust, governance and connectivity required for real-world operations”, said Paul Alexander, Vice President and Country Manager at Nokia. Alongside Stelia’s wider ecosystem of partners, the companies are redefining what trusted, enterprise-wide capability looks like, underpinning critical services across industries including space, media and entertainment, retail, and finance. For businesses, this is accelerating decision-making, reducing operational risk, and unlocking measurable returns from production-grade AI deployments Read more about the collaboration here: newsroom.stelia.ai/stelia-and-nok…

Stelia CTO David Hughes’s next stop is Berlin where he'll be speaking about building robust infrastructure to scale AI workloads at @Leaseweb's Tech Meetup. The great line-up of speakers also includes Ömer Karabacak from Youzu, and Felix D. Meyer from @cloudeteer, each bringing their own perspective on building and operating resilient AI infrastructure. In his talk, “Do we really need all those GPUs?”, Dave will explore how organisations can run AI workloads efficiently across distributed compute - without defaulting to high-cost GPU clusters. Looking forward to having @ArthurPoot as moderator, organiser of the excellent AI Builders community events we spoke at earlier this year. Thank you to the Leaseweb team for bringing the Berlin AI community together. We'd love for you to join us, register for free here: luma.com/klg1zbe2 #Leaseweb #AIInfrastructure #AICommunity

AI adoption at scale requires a fundamental shift in approach. Tool-first thinking creates friction at every level: teams waste time on integration overhead, leadership makes strategic decisions based on inaccurate perceptions, and costs compound invisibly across organisations, eroding competitive advantage. The organisations extracting genuine value are those that have flipped the model entirely, designing solutions within an operational context rather than in isolation. As AI capabilities increasingly move into production deployment, this shift is what will separate sustainable acceleration from compounding inefficiency. We’ve explored what a disciplined approach to AI adoption actually demands: newsroom.stelia.ai/the-hidden-cos… #EnterpriseAI #AIAdoption #AIStrategy #TechLeadership

The @nebiusai GTC Watch Party in Berlin kicked off with a betting pool on what would be announced within Jensen Huang's @NVIDIAGTC keynote in San Jose. Predictions ranged from the number of times AI would be mentioned to whether robotics, new architectures or hyperscaler partnerships would dominate the @NVIDIAAI keynote. The bets weren't far off with Huang reinforcing the rise of AI factories, agentic systems, and physical AI; providing a deeper look at the Vera Rubin platform; teasing what's next with Feynman; and reiterating just how quickly global demand for compute is accelerating. What nobody bet on: that @Disney's Olaf the snowman would end up centre stage and steal the headlines. A more in-depth analysis of NVIDIA GTC 2026 is coming next week from our team on the ground in San Jose. #NVIDIAGTC #AIInfrastructure #AIecosystem

In the race to productionise AI, storage decisions are often optimised for immediate results and speed-to-market alone. But designing production-grade AI systems demands building for long-term dependability with resilient, battle-tested foundations. In our Principal Architect's recent article, he examines why @ceph – a community-governed distributed storage platform built to scale linearly with demand and recover autonomously under failure – reflects exactly that kind of thinking. Read the full piece here: newsroom.stelia.ai/why-we-chose-c… #AIstorage #AIinfrastructure #Ceph #opensource #AIengineering

New Stelia merch has landed. And we've levelled up. It's been a pleasure rolling this out to the team via our Stelia Swag store and seeing it in the wild. Looking forward to sharing more sightings soon. In the meantime, we've had a surprising number of questions about public rollout. All we can say is watch this space…