The hard part of running inference shouldn’t be figuring out where to run it. That’s why we’re changing how inference works in AI Studio. With Dedicated Inference, you can now take a model from Hugging Face and deploy it on dedicated GPU resources through Hyperstack. And the new Deploy AI Wizard handles the infrastructure decision with you. Tell it what you want to run and what your workload looks like. It recommends a GPU configuration, gives you an estimated cost and gets you to deployment. So, you don’t need to worry about: -Figuring out which GPU configuration you need -Comparing infrastructure options for your workload -Manually working out what it’ll cost And with support for vLLM-compatible Hugging Face models and a growing model catalogue, there’s more to deploy. Try Dedicated Inference in AI Studio today: ai.hyperstack.cloud/

Olly Brooks takes the stage at AI Infra Summit 2026. 17 September, 10:45 AM, Santa Clara Convention Center. He's breaking down what sovereign AI actually requires: open weights vs. fully open models, the four layers of a sovereign deployment and when it makes sense to own GPU capacity vs. rent it. Then Booth 721. Book a meeting or walk over. bit.ly/46isQU2

See you at AI Infra Summit next week. Hyperstack, Booth 721, Santa Clara, 15-17 September. AI infrastructure and AI inference. Book a meeting or find us on the floor. bit.ly/4xgkOWH See you in Santa Clara! #AIInfra #Inference

Hyperstack and OCF are partnering to bring on-demand NVIDIA GPU cloud to UK HPC and AI teams. The pairing brings together OCF's two decades of HPC and AI engineering expertise with elastic GPU capacity from our Hyperstack platform. For research, life sciences, and engineering organisations, that means scalable GPU compute without the capital cost of expanding on-premise systems: burst demanding AI training, inference, and simulation workloads to the cloud when on-premise resources are stretched, then scale back down once the work is done. As teams move from traditional simulation into large-scale AI, demand for GPU compute keeps rising. This partnership is built to close that gap.

JUST 1 WEEK TO GO. WHO REALLY CONTROLS YOUR AI? AI security and control are too important to leave on the sidelines. Next week, our CPTO, Cory Hawkwelt, and UKAI’s CEO, Tim Flagg, will sit down to talk about what it really takes to build an AI stack you can control, from understanding your dependencies to asking the right questions about security and continuity. 📅 11 September | 2–3 PM BST Register: ukai.co/ai-event-calendar/wh…

Meet Hyperstack at AI Infra Summit 2026. Booth 721, Santa Clara, 15-17 September. AI infrastructure and AI inference with the team on the floor. Book a meeting or walk to Booth 721. hyperstack.cloud/events/infr…

We’re delighted to announce a new partnership between Hyperstack and Daemon. Daemon is an award-winning technology consultancy and AI-Augmented engineering partner. Hyperstack is the GPU cloud built for AI, delivering on-demand access to NVIDIA GPUs for training, fine-tuning and inference at scale. Enterprises have no shortage of AI ambition. What they need is the expertise to build the right solutions, the experience to move them safely into production and the infrastructure to run them at scale. Daemon brings AI-Augmented engineering squads, delivery excellence and technology strategy. Hyperstack brings the AI-native GPU cloud to power those solutions — from experimentation through to production at scale. Together, we will help organisations: Move AI projects from idea to production faster Access high-performance NVIDIA GPU compute on demand Scale AI workloads with confidence, with infrastructure economics designed for production AI This partnership brings together the expertise to build enterprise AI solutions with the infrastructure to run them at scale. Thanks to the teams at Daemon for a partnership we’re genuinely excited about. More to come. #EnterpriseAI #GPUCloud #Hyperstack #Daemon #AIInfrastructure #AIFirst

Watch a GPU go from zero to running in under 60 seconds. Join us in London on 10 September for a meetup hosted with Encode Club. Our Technical Account Manager, Olly Brooks, will demo deploying GPUs and fine-tuning open-source models with zero vendor lock-in. Register now to secure your spot. P.S. Stick around after for drinks, pizza and conversations with fellow AI builders and the Hyperstack team. bit.ly/4yhdkUn

Hyperstack is heading to Santa Clara. AI Infra Summit 2026, 15-17 September, Booth 721. AI inference on dedicated NVIDIA Blackwell and NVIDIA Blackwell Ultra Clusters. Hyperstack On-Demand, including AI Studio, if you want to start now. Book a meeting or find us at Booth 721. hyperstack.cloud/events/infr…

Hyperstack is heading to Santa Clara. AI Infra Summit 2026, 15-17 September, Booth 721. AI inference on dedicated NVIDIA Blackwell and NVIDIA Blackwell Ultra Clusters. Hyperstack On-Demand, including AI Studio, if you want to start now. Book a meeting or find us at Booth 721. hyperstack.cloud/events/infr…

Qwen3.8 Max: 2.4T parameters, 95B active per token and a 262K-token context that stretches past a million. It reasons on every single request. We deployed it across 64 NVIDIA H100 GPUs on Hyperstack and brought up a live endpoint in under 6 minutes, with zero failed requests throughout the full concurrency sweep. Full step-by-step tutorial on our blog: bit.ly/3STag1s

In our upcoming webinar next month, our CPTO Cory Hawkvelt and Tim Flagg (CEO, UKAI) unpack what it takes to build a sovereign tech stack. On the agenda: → What being secure really means beyond geography, not just where your GPUs sit → A layer-by-layer way to assess infrastructure, operations, models and data → The right questions to ask any provider, beyond "are you sovereign" → Concrete next steps you can act on straight after 11 September | 1 PM BST Join us: ukai.co/ai-event-calendar/wh…

05/05

04/05

03/05

02/05

Hyperstack July Release Highlights Catch up on everything we shipped, including the Docs MCP Server, Kimi K3 as a third-party model in AI Studio, multimodal AI Studio features and platform reliability improvements. Read the full update: hyperstack.cloud/technical-r… #Hyperstack #AIstudio 01/05

Our CPTO Cory Hawkvelt took the stage at Raise Summit 2026 to talk sovereignty alongside other leaders in AI. The conversation around AI is moving beyond model performance - it’s now about control, resilience, data sovereignty, and the infrastructure underneath it all. At Hyperstack, we’re seeing firsthand that the teams who get this right are the ones treating compute, storage, and policy as part of the same strategy. Cory’s session was a timely reminder that sovereign AI isn’t a future concept, it’s already becoming a core requirement. Watch the full talk on Raise's Youtube: piped.video/watch?v=-mZEa3Xv…

The Kimi K3 open weights dropped on 27 July. Now you can serve the 2.8T-parameter MoE model on Hyperstack with NVIDIA H100 GPUs. 104B active parameters/token 1M-token context window Native vision Step-by-step multi-node deployment with vLLM Read the tutorial and get started. hyperstack.cloud/technical-r…

We're delighted to announce a strategic partnership between Hyperstack powered by NexGen Cloud and @Computacenter, bringing sovereign, high-performance AI infrastructure to enterprise organisations across the UK. Through the partnership, Computacenter customers gain a straightforward route to Hyperstack's on-demand, NVIDIA-accelerated GPU cloud, hosted in Europe and powered by renewable energy. It combines our specialist AI infrastructure with Computacenter's trusted relationships across the UK's largest public and private sector organisations. We're already off the mark. At the end of June, our teams came together in London for a joint Sales Kick-Off, a half-day of enablement, joint value sessions and account planning that set the direction for how we'll help enterprises move from AI experimentation to production. A huge thank you to the Computacenter team for the energy and collaboration so far.

The model might be yours. But what about everything it runs on? GPU access. Provider dependencies. Closed APIs. Jurisdiction. Capacity. If your AI depends on infrastructure or models you don't fully control, someone else's decisions can become your business continuity problem. On 11 September, Hyperstack is co-hosting a live webinar with UKAI to explore the question: Who controls your AI, really? Join Hyperstack and UKAI for the conversation: ukai.co/ai-event-calendar/wh…

That’s a wrap for Hyperstack at @RaiseSummit Paris 2026. A huge thanks to everyone who stopped by Booth 14A, had conversations and to our team who helped make the event such a success.

06/06

05/06

04/06

03/06

02/06

What shipped on Hyperstack this week: Image Playground on AI Studio, Kubernetes resilience updates and more. Swipe to see the full breakdown. More details on the blog: hyperstack.cloud/technical-r… 01/06

Running a 70B LLM on one GPU? You're leaving performance on the table. 2× NVIDIA H100s on Hyperstack with vLLM tensor parallelism gives not just 2x the throughput, but nearly 4×. Here's why that happens and how to set it up: eu1.hubs.ly/H0wfQMX0

National lab-grade compute shouldn’t be out of reach. Hyperstack Secure Private Cloud is built for large-scale training, HPC, GROMACS, OpenFOAM, genomics and production inference without sacrificing data sovereignty. See you at ISC Hamburg next week. 📍 Booth A39 🔗 eu1.hubs.ly/H0w8y300 #ISC2026 #HPC #AI

We're taking the stage at @RaiseSummit. Our CPTO, Cory Hawkvelt, joins the panel: Sovereign Stacks: Building Trusted AI on National Terms July 8th | 2:00 PM | Ada Lovelace Stage As AI adoption accelerates across Europe, the question of where your infrastructure lives - and who has access to it - has never been more consequential. Cory joins leaders from Lenovo, Together AI, Digital Realty, and DDN to explore how organisations are making infrastructure decisions that determine not just performance, but sovereignty and long-term AI competitiveness. Pre-book a meeting with us: bit.ly/3RQBq8y #Hyperstack #RAISE2026 #RAISESummit #EnterpriseAI

We're heading to #RAISE2026 in Paris. Enterprise AI is shifting from experimentation to execution. @Hyperstack is the full-stack AI cloud helping organisations move from PoC to production on NVIDIA Blackwell & Blackwell Ultra GPUs. Find us at Booth 14A, 8–9 July. Book a meeting 👇 bit.ly/3PTBt2J #Hyperstack #EnterpriseAI #NVIDIA

We're attending The AI Summit London, 10–11 June, Tobacco Dock. If you're working in AI, you already know the gap between ambition and infrastructure is growing. Workloads are scaling faster than planned and access to compute at the right cost is becoming one of the defining questions right now. If you're attending and want to talk about how we can support with compute, come find us. #TheAISummit #Hyperstack

World's first open-source 100B medical LLM just dropped. AntAngelMed: #1 on OpenAI HealthBench. Here's how to run it on 8× NVIDIA H100s in 5 commands. Full tutorial: bit.ly/42T8sHi #MedicalAI #Hyperstack

ISC is where the teams running the world's most demanding workloads show up. And that’s where we fit in. Secure Private Cloud and Managed Clusters powered by NVIDIA GPUs provide the backbone for workloads that can't compromise. Genomics. Drug discovery. Climate modelling. Energy simulation. Large-scale AI training. If that's you, catch us at Booth A39, 22nd–26th June. Book a meeting now: bit.ly/3RrUw4C #Hyperstack #ISC26

Shared infrastructure works. Until InfoSec asks: “Can you prove no one else touches this?” One answer closes the deal. The other delays (or worse, derails) it. See how a single-tenant Secure Private Cloud gets you the right answer: bit.ly/3PYS70N

Our Head of Partnerships, Ashley Williams, is heading to the Qwen Conference in Singapore this Monday. If you're attending and want to talk AI infrastructure, token economics or building out your AI stack, get in touch. #QwenConference2026

Hamburg is about to get a lot more accelerated. Catch Hyperstack at ISC High Performance 2026 — Booth A39 | 22–26 June. Blackwell-powered infrastructure for HPC, genomics, AI training and production inference — from single-tenant SPC to managed Kubernetes & Slurm clusters. Book a meeting: hyperstack.cloud/events/isc-… #Hyperstack #ISC26 #ISCHPC

295 billion parameters. 21B active per token. 600 GB BF16 checkpoint, too large for a single node. We deployed Hy3-preview on Hyperstack using multi-node Kubernetes with 16 NVIDIA H100s across two worker nodes, hybrid Tensor + Expert Parallelism and a 600 GB BF16 checkpoint loaded from local NVMe. In this tutorial: → Multi-node Kubernetes cluster on Hyperstack (two 8x H100-80G PCIe-NVLink) → LeaderWorkerSet API for coordinated 2-node inference → vLLM with native multi-node tensor parallelism and MTP speculative decoding → 256K token context window with three reasoning tiers (no_think / low / high) → Multi-agent code review pipeline with parallel specialist agents and tool calling → Plugging into Claude Code, OpenClaw, and OpenCode as a local backend 80.6 on SWE-Bench Verified. 34.86 on LiveCodeBench v6. Full tutorial on the blog: Deploy Hy3-preview on Hyperstack: A Multi-Node Kubernetes Guide #Hyperstack #Hy3preview

One model. Video, audio, images, and documents - from a single endpoint. We deployed NVIDIA Nemotron 3 Nano Omni on Hyperstack and put its multimodal pipeline to work. In this tutorial: → vLLM serving on a single NVIDIA H100 80GB (62 GB BF16 checkpoint) → 256K token context window with native reasoning mode → PDF extraction - structured JSON from complex financial documents → Hour-long audio transcription with word-level timestamps and action-item extraction → Video summarisation and temporal Q&A from a single prompt → Disabling thinking mode for latency-sensitive tasks 67.04 on OCRBenchV2. 89.39 on VoiceBench. 72.2 on Video-MME. One deployment. Full tutorial on the blog: bit.ly/4duBhjd #Nemotron #MultimodalAI

1.6 trillion parameters. 49B active per token. Too large for a single node. We deployed DeepSeek-V4-Pro on Hyperstack using multi-node Kubernetes - 16 NVIDIA H100s across two worker nodes, hybrid Data + Expert Parallelism, and a 960 GB FP4+FP8 checkpoint loaded from local NVMe. In this tutorial: → Multi-node Kubernetes cluster on Hyperstack (2x 8x NVIDIA H100-80G PCIe-NVLink) → LeaderWorkerSet API for coordinated 2-node inference → vLLM with hybrid DEP topology and MTP speculative decoding → 1M token context window with three reasoning tiers → Long-horizon autonomous code refactoring with self-correction → Plugging into Claude Code, OpenClaw, and OpenCode as a local backend 80.6 on SWE-Bench Verified. 93.5 on LiveCodeBench v6. Full tutorial on the blog: bit.ly/4f1jamb #DeepSeek #AgenticAI

Running Kubernetes or SLURM in-house is a full-time job. Hyperstack Managed Cluster Platform hands you a fully managed cluster environment - delivered at the orchestrator layer, so your team focuses on models, not maintenance. GPU infrastructure. Fully managed. Ready to scale. Enquire now 👉 bit.ly/3QPhHp8 #ManagedKubernetes #SLURM

1 trillion parameters on 8 GPUs. Here's what that looks like. We deployed Kimi K2.6 on Hyperstack - @Kimi_Moonshot's open-weight agentic model. In this video: → vLLM serving on 8x NVIDIA H100-80G PCIe → 595 GB of INT4 weights loaded from ephemeral NVMe in ~6 minutes → Autonomous multi-step refactoring with self-correction → Coding-driven design - single prompt to working website → Local backend for Claude Code, OpenClaw and Kimi Code CLI 32B active parameters per token. 256K context window. 300 sub-agents in a single run. Full tutorial on our blog: bit.ly/4cFJVLF #KimiK2 #MoonshotAI

Full guide: bit.ly/4welSeu (🧵7 of 7)