Pre-training
Train foundation models from scratch on multi-node clusters sized for the job.
serious compute.
close to home.
Reserve dedicated, high-performance GPUs in Africa to train or fine-tune your own models, without sourcing hardware or building a data centre.
Our first AI factory: a planned 376 MW facility at Tinapa Business Resort, Calabar.
Built for the big jobs.
Fine-tuning
Adapt open-weight models to your language, domain and data, then serve them on Inference.
Large batch jobs
Embed archives, label datasets or run simulations across hundreds of GPUs at once.
Research and evaluation
Spin up short reservations to test ideas, benchmark models and compare runs.
Plan your reservation.
Pick a size and a term. Capacity comes online in phases from Tinapa, Calabar, and is allocated in waitlist order. We confirm pricing with you before anything is billed.
GPUs
Term
Comes with every cluster.
01
Ready-to-run clusters
Nodes arrive with drivers, networking and a job scheduler set up, so the first job runs on day one.
02
Fast shared storage
Keep datasets and checkpoints next to the GPUs, in the same region as the rest of your data.
03
Engineers who answer
Work with our infrastructure team on sizing, scaling and getting runs back on track.
Train here. Ship anywhere on Hiperlinks.
train it here.
Tell us what you are building. We size the cluster with you and hold your place in the queue.