Four audiences. One vertically integrated stack — adapted to where you operate, what you need to keep sovereign, and how fast you need to ship.
For banks, telcos, energy companies, and regulated enterprises running AI, HPC, and analytics that have to stay in-jurisdiction. Sovereign cloud on Navon-owned hardware, priced for growth markets, with the ergonomics your engineering team already expects.
Bare-metal, VM, and managed inference fleets — physically in-country, not a regional zone of a foreign hyperscaler.
Annual commitments with reserved-capacity discounts. No pay-per-second surprises on a six-figure GPU run.
Kenya DPA 2019, GDPR-aligned audit logs, role-based access, sovereign HSM key management — built in, not retrofitted.
Annual contracts billable in USD or KES. FX exposure managed by Navon, not your CFO.
Multi-carrier fibre into the meet-me room. Cross-connect to your existing footprint without vendor lock-in.
Named account manager, capacity planning, usage reporting tied to your finance close. No ticket queue.
We're best-suited where sovereignty, density, or proximity matters. If your workload is latency-insensitive, low-density, and has no data residency requirement, a hyperscaler is usually cheaper.
Fine-tuned banking and telco LLMs serving regulated customers — Ollama, vLLM, and an OpenAI-compatible endpoint on owned silicon. Weights and traffic stay in-country; no external routing layer in the path.
Climate, financial risk, CFD, structural modelling — dedicated CPU/GPU pools.
Domain-specific model training on customer data that can't leave the country.
End-to-end encrypted comms platforms for cross-border deal teams and executive operations.
30-minute scoping call to map your workload to the right product line and deployment shape. Indicative pricing under NDA.
For ministries, agencies, central banks, and security services that need their digital infrastructure to live where their statute applies. Built on hardware Navon owns and operates — on-soil, in-jurisdiction, auditable in person.
Servers in Kenya, operated by Navon, under Kenyan jurisdiction. Not a regional flag on a US hyperscaler.
Hardware security modules under local jurisdiction. No cross-border key replication unless explicitly authorised.
Post-quantum cryptography embedded at the platform layer — not a 2030 retrofit project.
Local engineering teams. Physical site visits. Sovereign access logs that don't transit a foreign network.
Naivasha Special Economic Zone status — tax, customs, and operating frameworks aligned to long-horizon infrastructure.
Same modular pod architecture supports civilian government workloads and dedicated secure-communications enclaves.
From sovereign cloud baseline through to specialised enclaves for regulated programmes. Procurement-friendly contract structures and multi-year commitments available.
Population-scale identity systems with full data residency and DPO-level oversight.
Government LLMs, language models for local languages, training and inference on classified datasets — model weights and request traffic never transit an external routing layer or foreign jurisdiction.
General-purpose cloud for ministries replacing public-cloud sprawl with a single sovereign tenant.
End-to-end encrypted comms with PQC and optional QKD pathways for sensitive operations.
Dedicated engagement track for ministries, central banks, and security services. Compliance documentation, in-person briefings, and multi-year contract structures.
Direct, dedicated access to Navon's data centre at Hells Gate Deep Tech Park, Naivasha SEZ — alongside a global expert network, structured internships at the site itself, and a pathway from student project to industry career. Built for every faculty where compute is the bottleneck between theory and practice.
Researchers and students live in under a minute. No local installation, no per-user IT overhead, persistent workspaces across sessions.
GPU and CPU/HPC allocations dedicated to the institution per tier. No competition between departments for the same pool.
Annual commitments billable in USD or KES. Predictable budget for grant compliance and finance close.
Quantum sandbox for circuits, hybrid algorithms, and post-quantum cryptography research — alongside classical compute.
Structured placements at the Naivasha facility across infrastructure, ops, security, and research — confirmed within four weeks of nomination.
On-site NAS for datasets, model outputs, library archives, and lifelong research artefacts. Persistent across sessions, accessible to every faculty, never leaves Kenya.
Shared, multi-purpose infrastructure — not a single-use GPU cluster. All compute, storage, and developer environments live on-site at Hells Gate, powered entirely by geothermal energy. Resources are accessible through a single university login.
| Resource | Specification | Purpose |
|---|---|---|
| Data centre | Hells Gate Deep Tech Park, Naivasha SEZ — power, bare metal, network fabric (25G to compute, 100G to storage) Geothermal · 100% green | Navon-owned physical facility. All infrastructure on-site in Kenya; no data leaves Kenyan soil. |
| GPU compute | 2× GPU nodes — 8× NVIDIA RTX 5090 per node, 1 TB DDR5 RAM, 8 TB NVMe per node | AI/ML training, inference serving, image processing, scientific GPU workloads — dedicated allocation per tier. |
| CPU / HPC | 3× dual AMD EPYC 9965 compute nodes — 768 GB DDR5 RAM, 4 TB NVMe per node | Simulation, optimisation, large-memory scientific computing, climate modelling, bioinformatics, engineering analysis. |
| Storage (NAS) | 1× high-performance NAS — AMD EPYC 9965, 512 GB DDR5, 4 TB NVMe boot, 24 × 24 TB SAS HDD (576 TB raw), 2× 100 GbE NIC | Persistent research data, datasets, model outputs, publications — sovereign, secure, accessible across sessions and teams. |
| Developer sandboxes | Isolated per project/group — JupyterHub (university SSO, persistent workspaces), AI environments, quantum sandbox | Full-stack research and development: AI/ML workflows, Jupyter notebooks, quantum circuits, hybrid algorithms, PQC — live in under 60 seconds. |
| Orchestration | Hosted.ai / Kinesis Enterprise — unified platform management, job scheduling, resource governance | IT admins manage users, faculties, quotas and usage reporting from a single dashboard. Fair allocation between departments with priority scheduling. |
| Compliance | Kenya Data Protection Act 2019 — full sovereignty, no international transfer agreements required | All compute and storage on Kenyan soil at all times — fully auditable under Kenyan law. |
One partnership agreement gives every faculty access through the right interface — low-code workflows for non-engineering teams, JupyterHub notebooks for researchers, and full SDKs for engineering groups. The infrastructure scales to match what the institution actually needs — today, and as research programmes grow.
Model training, fine-tuning open weights on sensitive research datasets, NLP for local languages, computer vision, algorithm research, systems programming on dedicated GPU + dev sandboxes. Ollama and an OpenAI-compatible endpoint for the tools students and researchers already use.
Structural simulation, fluid dynamics, finite-element analysis, embedded systems prototyping on CPU/HPC + secure storage.
Quantum computing research, numerical modelling, large-scale optimisation problems on CPU/HPC + quantum sandbox + persistent storage.
Medical imaging analysis, genomics pipelines, protein folding, drug-discovery simulation on GPU inference + CPU/HPC + secure storage.
Weather and climate modelling, satellite imagery analysis, geothermal data processing on CPU/HPC + GPU training + large-memory workloads.
Large-scale survey analysis, demographic modelling, NLP on local-language datasets via JupyterHub + GPU inference + secure storage.
Research repository management, digital archive curation, dataset cataloguing, library systems, and central IT operations — the disciplines that hold the institution's data backbone, on the same infrastructure they curate for everyone else.
Shared datasets, reproducible analysis pipelines, interdisciplinary collaboration on JupyterHub + dev sandboxes + object storage.
Indicative monthly cost for an equivalent mixed university research workload across GPU, CPU, and storage. Annual contracts eliminate pay-per-second unpredictability — student access, sandboxes, and the internship pipeline are included rather than priced separately. Sources: AWS, Google Cloud, Azure pricing (verified April 2026).
| Navon Africa | AWS (p4d.24xl) | Google Cloud (a2-highgpu) | Azure (ND96) | |
|---|---|---|---|---|
| Indicative monthly cost | $10,000–$14,000 | $22,000–$28,000 | $19,000–$25,000 | $21,000–$27,000 |
| Currency | KES / USD | USD only | USD only | USD only |
| Data residency | Kenya · on-soil | EU / US regions | EU / US regions | South Africa |
| Billing model | Annual · fixed | Pay-per-second | Pay-per-second | Pay-per-second |
| Student access | Built-in, included | Not included | Not included | Not included |
| Sandboxes & quantum | Included | Not included | Not included | Not included |
| Internship pipeline | Included | None | None | None |
| Carbon footprint | Near-zero · geothermal | Varies by region | Varies by region | Varies by region |
| Cost vs Navon | — | 60–180% more | 35–150% more | 50–170% more |
Every partnership ships with a named account manager, dedicated onboarding engineer for the first 30 days, and quarterly reviews. The commitments below are written into the agreement.
Monthly availability target across compute, storage, and developer environments.
Under 2 hours from notification — direct line to the on-call engineer.
Under 8 hours from notification, including triage and root-cause communication.
Dedicated onboarding engineer for the first 30 days — until every faculty is live.
Single point of contact for all queries — commercial, technical, escalation.
Usage reporting across GPU, CPU, and storage; capacity planning; roadmap alignment.
Confirmation within 4 weeks of nomination — at the data centre or via the partner network.
Minimum one session per cohort per month from the global expert network.
The internship pipeline is rooted in a real, operating facility. Students don't get placed in generic office roles — they work at or alongside the Hells Gate data centre, gaining hands-on experience with the infrastructure they've been using throughout their studies.
Students work at or alongside the operating facility. Defined responsibilities, mentored project outcomes, credit-bearing where the institution wants it.
Placement opportunities with Navon's partner network across East Africa, Europe and beyond — exposure to live industry challenges in AI, cybersecurity, fintech, agritech and infrastructure.
We recommend starting with a structured 90-day pilot. Research teams across participating faculties run real workloads on the platform before the institution commits to a full annual partnership. Pilot pricing is discounted; no commitment beyond 90 days; full data portability if you choose not to continue.
Accounts provisioned, SSO configured, all resource pools activated. Research leads from each participating faculty receive platform training.
Students and researchers run real workloads across GPU, CPU/HPC, and storage. Internship candidates nominated. First mentorship session delivered.
Usage review by faculty, cost comparison vs current alternatives, outcomes assessed. Go/no-go decision with full data portability if not proceeding.
Working relationships through which Navon brings external expertise directly to partner universities — guest lectures, knowledge transfer, curriculum support, and research collaboration. Delivered on top of the infrastructure your institution is already using.
Guest lectures, curriculum alignment, and research collaboration in AI, computer science, and data systems. Access to UCL's global network of researchers and practitioners.
Knowledge transfer in advanced computing, engineering simulation, robotics, and applied AI — directly relevant to engineering and science faculties at partner institutions.
Curriculum input and guest expertise in telecommunications, AI, systems engineering, and large-scale infrastructure design.
Research collaboration and guest education in data science, quantum computing, and advanced networking. Active contributor to Navon's quantum research programme.
Four research domains where the partner network can deliver guest lectures, curriculum input, and joint research — at any partner university, on top of the infrastructure already in place.
From first call to first faculty live in two weeks. Lightweight MoU, clear obligations, and a structured engagement so the institution can evaluate before committing.
30-minute conversation to understand research workloads across faculties, student cohort size, and institutional requirements.
Navon engineering maps compute needs across GPU, CPU/HPC, and storage to the right tier and configuration.
Lightweight agreement with clear obligations, SLAs, and success criteria — not a 60-page master service contract.
First students and researchers across all participating faculties are live within two weeks of signing.
Annual contract sized to the institution — from teaching-focused departments to multi-faculty research universities. Named account manager, dedicated onboarding, internship pipeline, mentorship cadence — all included. Start with a 90-day pilot.
For AI founders and engineering teams building on the continent who need real compute — without hyperscaler pricing, without data leaving Kenya, and without a six-month procurement cycle. Ship your model from the same infrastructure that enterprise and government already run on.
H100, A100, and RTX clusters available for training, fine-tuning, and inference — no minimum spend, no waitlist.
Your training data and model weights never leave Kenyan soil. Jurisdiction is a fact, not a setting you configure.
Ollama, vLLM, JupyterHub, and an OpenAI-compatible endpoint — your workflow, your stack, our infrastructure.
Monthly pricing with no lock-in. Upgrade to annual for significant discounts when you're ready to scale.
Clean energy and local operations mean we can price below hyperscalers while keeping margins that let us invest in the infrastructure.
Hackathons, developer days, and a growing community of AI builders across East Africa — all rooted in Hells Gate.
From pre-training to production inference — the full model lifecycle, on infrastructure that's physically sovereign and commercially competitive.
Fine-tune foundation models on African-language datasets, sector-specific corpora, or proprietary customer data — weights never leave Kenya.
Ollama and vLLM endpoints for production inference — OpenAI-compatible API, low latency on-continent, scalable on demand.
JupyterHub environments for rapid prototyping — GPU-backed notebooks that persist between sessions.
NVMe-backed storage for training datasets, model checkpoints, and outputs — sovereign, fast, and never subject to cross-border transfer rules.
GPU access from day one — no procurement cycle, no minimum spend. Talk to us about your workload and we'll size a configuration that fits your stage and budget.
Universities co-deploy with ministries. Banks contract through enterprise but pilot through research. Governments incubate startups inside sovereign clouds. We design the engagement around the actual workload, not the org chart.