The Scenario
Consider a machine learning researcher returning to Nigeria after ten years in the United States, where he completed a PhD and spent four years at a research institution in California. He relocates to Lagos with his family and a clear intention: to do the same quality of research in Nigeria that he was doing abroad, and to mentor a generation of Nigerian ML practitioners producing serious work on inadequate infrastructure.
A researcher in that position arrives with a specification already in mind, and it is extreme by any measure. The question is not whether it can be matched on paper — it is whether it can be built in Lagos, on Nigerian power infrastructure, and made reliable.
The Challenge
Research of this kind involves training large neural network models — transformer architectures for NLP tasks specific to Nigerian language data. This is GPU compute work of the most demanding kind. The models require GPUs with large VRAM (to hold the model and its gradients simultaneously), high memory bandwidth (to move data between compute units quickly), and NVLink or equivalent interconnect (so multiple GPUs can share a single model that exceeds one GPU's VRAM).
An American lab has A100 80GB clusters available over the cloud. That is not being recreated in a home office in Lekki. The goal is to get as close as possible with consumer and prosumer hardware available in Nigeria.
The additional challenge is Lagos power. Research runs are sometimes 6–72 hours of sustained GPU compute. A power interruption that crashes a 48-hour training run wastes 48 hours of compute time, potentially corrupts checkpoint files, and breaks the research cadence. The power protection has to be serious.
The Assessment
A build like this needs a technical conversation and a practical one. The technical side covers GPU selection (RTX 4090 vs RTX 6000 Ada vs H100 — the tradeoffs deserve an honest airing), NVLink bridging between consumer RTX 4090s, memory bandwidth limitations, and the thermal envelope of running two RTX 4090s in sustained compute mode in a Lagos home office.
The practical side covers the power situation (somewhere like Lekki Phase 1 — relatively stable but not perfectly reliable), an existing generator (a 7.5kVA unit, say), the room the machine will live in, and how the UPS integrates into the existing power chain so the machine rides through transitions between NEPA and generator without interruption.
Cooling deserves the same honesty. Two RTX 4090s under sustained AI training load produce approximately 600W of heat. In a Lagos home office, without a dedicated cooling system, that heat builds. A portable precision air conditioner for the machine room, controlled by a smart thermostat that activates above 26°C, is what makes the room viable — and it is infrastructure the researcher provides, not hardware in the build.
The Build
AI Research Workstation — indicative ₦18.4 million:
- CPU: AMD Threadripper PRO 7965WX — 24 cores, 128 PCIe 5.0 lanes; necessary to provide full-bandwidth PCIe x16 to both GPUs simultaneously
- RAM: 256GB DDR5 ECC RDIMM — ECC for data integrity in long research runs; 256GB to hold large datasets in system memory
- GPU: 2× NVIDIA RTX 4090 24GB with NVLink bridge — 48GB combined VRAM, NVLink interconnect for model parallelism across both GPUs
- Storage (active): 4TB Samsung 990 Pro NVMe — training data and model checkpoints
- Storage (archive): 2× 16TB Seagate Exos enterprise HDDs in RAID-1 — research datasets backed up nightly
- Motherboard: ASUS Pro WS WRX90E-SAGE SE — workstation-class, supports Threadripper PRO, dual NVLink
- PSU: 2× Seasonic Prime TX-1600 in dual-PSU configuration — 3,200W total capacity with failover
- Case: Fractal Design Define 7 XL — maximum airflow, acoustic damping, 420mm radiator support
- Cooling: Custom hardline water cooling loop — both GPUs and CPU on a single 420mm + 360mm radiator stack
- UPS: 10kVA online double-conversion UPS — seamless transition between NEPA, generator, and UPS battery; no interruption at any transition point
A build at this level is a two-day job on-site: one day building and cabling, one day testing under load, configuring CUDA, PyTorch, and the development environment, and verifying the power chain works correctly through simulated NEPA cuts and generator transitions.
What Changes
Two RTX 4090s bridged over NVLink act as a 48GB unit for model distribution, which is what allows model architectures that exceed a single card's VRAM to train locally at all. For medium-sized transformer training, a setup like this is competitive with a single cloud A100 node — and it has no marginal cost per run, which changes how freely a researcher iterates.
The power chain is what makes it usable rather than merely fast. A 10kVA online double-conversion UPS means NEPA cuts and generator transitions never reach the machine, so training runs measured in days survive an environment that would otherwise kill them. Compute you cannot rely on for 61 hours is not compute you can plan research around.
Key Takeaway
World-class AI research infrastructure is buildable in Nigeria. The components exist. The expertise to configure them exists. What was missing was a vendor who would approach it seriously — designing for Nigerian power realities, thinking about thermal management in tropical conditions, and building with the reliability that sustained research computing demands. The ceiling on what's achievable in Nigeria is being raised by practitioners who refuse to accept that geography equals limitation.
Are you a data scientist, ML researcher, or AI practitioner in Nigeria? Explore the AI Series or talk to our team about a research-grade workstation built for your workload.