Hammer AI Works: getting more out of every GPU

Hammer AI Works: getting more out of every GPU. Nordic mountain range at night under the Milky Way, with the white headline and the logos of hammer, AMD, Aixia and Liqid

How do you get more out of your GPUs when AI workloads keep changing? On 8 December, Aixia invites you to Hammer AI Works, a lunch seminar at our Gothenburg office, where we take a closer look at how Liqid’s infrastructure with AMD Instinct GPUs lets you adapt compute resources to different workloads and make better use of the capacity you already have.

AI workloads rarely stay the same shape for long. A model that ran comfortably last quarter is suddenly constrained by memory bandwidth, a new service pushes concurrency up, or a training job needs a burst of GPUs that sits idle a month later. Building for the peak means paying for capacity you use a fraction of the time. Building too lean means teams wait.

Adapt the infrastructure, not the workload

Liqid’s approach is to treat compute as a pool rather than a fixed allocation. Resources are composed to match what the workload actually needs at that moment, so the same physical hardware serves different jobs over the course of a day instead of being locked to one team.

Chris Bone, AI Sales Specialist, will walk through how a more flexible infrastructure shortens the path from planning to deployment and makes it easier to scale up as demand grows. The session is practical rather than theoretical: what changes in practice when the fabric is composable, where the gains show up, and what to plan for before you commit to a design.

What we will cover

  • How composable infrastructure with AMD Instinct GPUs adapts to shifting AI workloads
  • Where increased utilisation translates into measurable gains, and where it does not
  • How to shorten the time from planning to production deployment
  • What scaling up looks like when demand outgrows the initial design

After the presentation we continue the conversation over lunch, with time for questions and for comparing notes with others working on the same problems.

Who it is for

The seminar is aimed at people who own or influence AI infrastructure decisions: platform and infrastructure leads, architects, data science leaders, and anyone weighing the next GPU investment against what the existing estate can still deliver. No prior knowledge of Liqid or AMD is required.

When Tuesday 8 December 2026, 11.00-13.30

Where Aixia, Hälsingegatan 10, Gothenburg

Agenda 11.00 Welcome reception · 11.15 Aixia · 11.30 Presentation from Chris Bone · 12.30 Lunch and networking

Contact Sonny Clark, sonny.clark@aixia.se, +46 763 16 59 63

Places are limited and registration is binding. Sign-up takes under a minute.

Register for Hammer AI Works

Latest News

Hammer AI Works: getting more out of every GPU. Nordic mountain range at night under the Milky Way, with the white headline and the logos of hammer, AMD, Aixia and Liqid

Hammer AI Works: getting more out of every GPU

On 8 December, Aixia hosts Hammer AI Works in Gothenburg. Liqid and AMD Instinct GPUs, composable infrastructure, and how to…
Läs mer

Så får ni mer nytta av varje GPU-timme

GPU-utnyttjandet ligger ofta under 30 procent. AiQu samlar kapaciteten i en gemensam pool, och tre mått visar vad infrastrukturen faktiskt…
Läs mer
Produktionslinje för plastfilm med inline-kamerainspektion, en tekniker granskar kvalitetsdata på en skärm

AI i kvalitetskontrollen: så minskade Trioworld materialsvinn i produktionen

Hur kan AI förbättra kvalitetskontrollen i produktion? Läs om Aixias samarbete med Trioworld och vad tillverkare bör tänka på innan…
Läs mer

Palo Alto Prisma AIRS passerar 100 MUSD ARR: AI-säkerhet har blivit affärskritiskt

Prisma AIRS passerade 100 MUSD i ARR på fyra kvartal, och Palo Alto bygger nu ut skyddet till utvecklarnas datorer….
Läs mer