UK’s trusted IT infrastructure partner since 2003
Servnet
FinanceToolsConfiguratorGet in Touch
AI Infrastructure — Insights — networkAI Infrastructure — Insights — reach
Insights · AI Infrastructure

AI Infrastructure

29 articles in this category

AI Infrastructure

Fine-Tuning vs RAG: The UK Infrastructure Decision for 2026

Most fine-tuning vs RAG comparisons argue about accuracy. Few mention that a full fine-tune of a 70B model needs roughly 700GB of VRAM, while QLoRA ca

· 8 min read
AI Infrastructure

AMD Instinct vs Nvidia 2026: A Real Alternative Yet?

AMD's Instinct MI355X posted 100,282 tokens per second on the Llama2-70b-99.9 Server test in MLPerf Inference v6.0 — a 3.1x jump over the MI325X resul

· 8 min read
AI Infrastructure

Mac Studio LLM Performance vs GPU Server: 2026 Reality

Apple's March 2025 launch material said the Mac Studio could be configured with 512GB of unified memory and run LLMs over 600 billion parameters entir

· 8 min read
AI Infrastructure

RAG Infrastructure Requirements: A UK Buyer's BOM for 2026

Most RAG explainers stop at a three-box diagram: documents in, vectors stored, answer out. What they skip is the bill of materials. Supabase's own ben

· 8 min read
AI Infrastructure

GPU Utilisation Monitoring: Fix It Before Buying More GPUs

UK IT leaders are about to sign off another six-figure GPU order to fix a problem monitoring could solve for free. A June 2026 survey of 573 technical

· 7 min read
AI Infrastructure

Small Language Models: The Smarter UK Business AI Buy in 2026

Mid-2026 frontier API pricing tells its own story: GPT-5.6 Pro costs up to $60 per million output tokens, while a Phi-4-mini model needing roughly 3GB

· 9 min read
AI Infrastructure

Best AI GPU Server 2026: XE9680 vs HPE vs Supermicro

UK IT leaders comparing the Dell PowerEdge XE9680, HPE's Apollo/XD685 platforms and Supermicro's SYS-421GE for 8-GPU AI training in 2026 are chasing a

· 7 min read
AI Infrastructure

Used H100 GPU 2026: Bargain or Trap? UK Buyer's Verdict

The clearest signal in the 2026 GPU market isn't the Blackwell launch — it's what's happening to the Hopper cards it's displacing. H100 cards that sol

· 7 min read
AI Infrastructure

AI Server Power Requirements 2026: Amps Beat GPUs

Everyone still talks about GPU allocation. But a single NVIDIA GB200 NVL72 rack draws 120-132 kW — eight to twenty-five times a standard enterprise ra

· 7 min read
AI Infrastructure

Sovereign AI UK 2026: Build or Rent British AI Capacity?

Sovereign AI has stopped being a slogan and become a live UK procurement category, with government now running a dedicated Sovereign AI proof-of-conce

· 9 min read
AI Infrastructure

Inference Server Hardware 2026: Buy For Agents, Not Training

UK IT leaders are still buying training-grade GPU clusters for a workload that has already moved on. The UK government's own compute annex puts total

· 8 min read
AI Infrastructure

When to Switch to Liquid Cooling: The 2026 UK Threshold

Air and liquid don't compete on efficiency — they compete on physics. Liquids move heat between 50 and 1,000 times more efficiently than air, and CIBS

· 6 min read
AI Infrastructure

Do You Really Need a 130kW AI Rack? (2026 Guide)

A single NVIDIA GB200 NVL72 rack draws 120-140 kilowatts and costs roughly $2-3 million in hardware alone — built to deliver 1.4 exaflops for trillion

· 8 min read
AI Infrastructure

Blackwell vs Rubin: The UK 2026 AI Server Buy Decision

NVIDIA's Rubin R100 entered full production at CES 2026 and is rated at 50 PFLOPS of FP4 inference per GPU against the Blackwell B200's 18 PFLOPS — a

· 7 min read
AI Infrastructure

AI Workload Repatriation 2026: A UK Buyer's Cost Case

Everyone claims to be 'repatriating AI from the cloud' in 2026, but the honest answer is: it depends on utilisation and data sensitivity, not ideology

· 6 min read
AI Infrastructure

AI Server Prices 2026: The HBM Crunch Hitting UK Buyers

Every AI server quote landing on a UK desk in mid-2026 seems to have grown since the last one. The reason sits inside the GPU itself: HBM memory now m

· 8 min read
AI Infrastructure · GPU Sizing

How much VRAM does a large language model need? (2026)

The first question when self-hosting a large language model is deceptively simple: will it fit? The answer is a sum of three parts - the model weights

· 8 min read
AI Infrastructure · Cost & ROI

Self-hosting LLMs vs cloud GPUs: the break-even (UK, 2026)

Renting GPUs by the hour is the obvious way to start with AI, and often the right one. But once a workload runs steadily, the meter never stops - and

· 8 min read
AI Infrastructure · Facilities

Hosting a Blackwell AI rack: power, cooling & colo vs on-prem

Buying the GPUs is the easy part. The harder question — the one that stalls more AI projects than budget does — is where you physically put the rack.

· 8 min read
AI Infrastructure · Procurement

Refurbished H100 & H200: is the 2026 secondary market worth it?

As Blackwell ships, the previous flagship GPUs are flowing onto the secondary market — and for a lot of real workloads, a refurbished H100 or H200 is

· 7 min read
End-User Computing · AI

Copilot+ PCs in 2026: buy for the workload, not the badge

Every business laptop now ships with an 'AI PC' or 'Copilot+' sticker and a neural processing unit (NPU). The pitch is that on-device AI is the reason

· 6 min read
Technical · AI Infrastructure · How-To

Building your first UK on-prem AI cluster: step-by-step

Cloud AI training is convenient but at sustained scale (multi-week training runs, hundreds of GPUs), on-prem becomes meaningfully cheaper. UK enterpri

· 11 min read
Technical · AI Infrastructure · Cooling

AI server cooling 2026: air vs liquid vs immersion

NVIDIA H100 SXM was the last GPU generation comfortably air-cooled at rack density. H200, B100, B200, and GB200 NVL72 push thermal density beyond what

· 9 min read
Server Infrastructure · AI Hardware

Dell PowerEdge XE9680, XE8640 and XE7745: choosing a Dell AI / GPU server in 2026 (UK)

Dell's PowerEdge XE line is its purpose-built AI and GPU range, and choosing within it is really a question about your workload, not your budget. An e

· 12 min read
Compliance · AI Governance · UK

The EU AI Act: UK impact analysis for IT teams

The EU AI Act entered force August 2024 with phased application through 2027. UK organisations deploying AI in EU markets, AI vendors serving EU custo

· 9 min read
Networking · AI Fabric · Buyer's Guide

Arista vs Cisco vs Juniper for 800G AI fabric (2026): UK data centre networking buyer's guide

AI training clusters and large-language-model inference workloads have pushed data centre networking into 800G as the new default. Arista 7060X6, Cisc

· 11 min read
AI Infrastructure · Trends

Ultra Ethernet vs UALink vs InfiniBand: the AI interconnect standards to watch

When you build an AI cluster, the network between the GPUs matters almost as much as the GPUs themselves - a fast accelerator starved of interconnect

· 13 min read
AI Infrastructure · Explainer

Liquid vs immersion cooling and RoCE: decoding AI data-centre terms (UK)

Two clusters of jargon dominate any AI data-centre conversation: how the heat gets out, and how the network stays lossless. Air, direct liquid and imm

· 10 min read
AI Infrastructure · GPU · Buyer's Guide

NVIDIA H200 vs H100 vs B100 vs B200: UK AI GPU buyer's guide 2026

NVIDIA H100, H200, B100, B200, and GB200 NVL72 are the GPUs UK enterprise AI buyers shortlist in 2026. H100 + H200 (Hopper architecture) are mainstrea

· 10 min read

Talk to a UK specialist

Get expert advice or a no-obligation quote — servers, storage, networking, maintenance, finance and cloud. We reply the same working day.

or call 0800 987 4111