Blog

Articles from stack8s on sovereign AI, hybrid cloud architecture, and practical GPU economics.

09 May 20269 min read

UK/EU Data Sovereignity Focused | S3-Compatible Object Storage: Best Options Beyond AWS

Most object storage choices look different on the pricing page and strangely similar in production. The reason is simple: the interface most teams care about is still S3-compatible object storage. If your tooling already speaks S3, you usually don't want to rewrite backup jobs, SDK integrations, AI data pipelines, or internal services just because the backend changes. What matters then is not only API compatibility, but also licence terms, egress costs, jurisdiction, and how much operational wo

ArticleRead more

02 May 20268 min read

TTFT Wars: GPU vs TPU vs LPU vs Apple Silicon vs Taalas Chips

Time to First Token, or TTFT, is the pause between sending a prompt and seeing the first word come back. In most AI products, that's the delay users feel first, judge first, and remember first. That makes hardware a latency decision, not only a cost or throughput decision. If you're building chat, voice, or agents, the question isn't only "how many tokens per second can this chip push?" It's "how quickly can it start?" What TTFT really measures, and why users notice it straight away TTFT is

ArticleRead more

19 Apr 202610 min read

U.S. Semiconductor Supply Chain: Why Chips Go to Taiwan for Packaging

0:00/53.2000831× The hard part of AI chip supply is no longer only the fab. A large share of the delay now sits in advanced packaging, the step that turns separate dies and memory into a usable processor module. That matters because even chips built in the US can still be sent to Taiwan for packaging before they go into servers. As AI demand rises, packaging capacity now affects delivery times, system design, and who gets access to the most advanced chips first. Advanced packaging is now as

ArticleRead more

08 Apr 20267 min read

Which GPU for Your LLM Model? A Practical Buying Guide

Picking a GPU for an LLM sounds simple until you hit the real variables. Model size, context length, user count, response speed, and budget all pull in different directions. That's why there isn't one best GPU for every LLM workload. For many teams, VRAM matters more than peak compute, because if the model doesn't fit in memory, nothing else matters. This guide is for technical and budget owners alike. Start with the job you need to run, then work back to the hardware. Start with the workloa

ArticleRead more

06 Apr 202610 min read

Nvidia GPUs vs Google TPUs and AWS Trainium Explained

0:00/76.5606671× AI demand has turned chip choice into a business decision, not only an engineering one. If you run model training, large-scale inference, or edge AI, the hardware mix now shapes cost, speed, power use, and lock-in. That matters because the market is no longer centred on one chip type. Nvidia still leads with GPUs, yet Google, Amazon, Meta, Microsoft and others are building custom silicon for their own AI workloads. The split between training, inference and on-device AI is driv

ArticleRead more

18 Mar 20265 min read

AI Grid Orchestration for Telcos with stack8s

AI Grid with stack8s - podcast0:00/117.21× Telcos no longer run AI in one neat data centre. They run it across towers, central offices, regional sites, and cloud zones. That spread creates a hard problem: how do you manage all of it as one platform without losing control of latency, cost, GPUs, or data rules? That is where AI Grid Orchestration fits. It places workloads where they make the most sense, then keeps policy, scaling, and recovery aligned across the estate. NVIDIA AI Grid Orchestrat

ArticleRead more

12 Mar 20263 min read

Build a System That Lasts..Stop Building AI Agents

I keep seeing founders burn weeks building shiny AI agents, then wonder why nothing sticks. The bottom line is simple: most "agents" don't create durable value, they create moving parts. When the model changes, the tool changes, the prompt breaks, and the whole thing wobbles. I'm not saying automation is bad. I'm saying the lasting part usually isn't an agent at all. It's the plain, boring stuff you can reason about, review, version, and hand over to a team without a long meeting. Why I'm sce

ArticleRead more

11 Mar 20266 min read

GPT-OSS-120B inferencing: which GPUs make sense to host it in 2026?

Running GPT-OSS-120B in production sounds like a pure compute problem. In practice, it's a memory problem first, then everything else. DevOps teams want predictable latency and clean scaling. CTOs want a platform choice that won't stall delivery. CFOs want a cost line they can defend. GPT-OSS-120B is a 117B-parameter Mixture-of-Experts model, yet only about 5.1B parameters are active per token. That lowers compute compared with dense 120B models, but it doesn't magically remove VRAM pressure. W

ArticleRead more

11 Mar 20268 min read

H100 SXM5 vs H100 PCIe vs H100 NVL: real differences and best use cases

If you're pricing an AI cluster in March 2026, the names can feel like a trap. H100 SXM5, H100 PCIe, and H100 NVL all say "H100", so they must behave the same, right? In practice, the module, power limit, memory bandwidth, and GPU-to-GPU links change what you can build, how fast it trains, and how much the rack costs to run. This guide keeps it practical for DevOps, CTOs, CFOs, cloud users, AI analytics teams, and researchers. You'll see what stays the same (Hopper features), what changes (pack

ArticleRead more