Cassava Technologies is adding NVIDIA’s open Nemotron models to its African AI platform, so enterprises can call them as a service without building their own model-serving infrastructure.
The Pan-African technology group announced on 21 October 2026 that it is expanding its AI Factory with the Nemotron family. The models will run as a locally hosted, managed inference service through the Cassava AI Multi-Model Exchange (CAIMEx), where they are offered to customers as APIs.
One gateway, several models
CAIMEx is Cassava’s exchange for AI models. Its product page describes a single API that lets customers, particularly telcos, use several large language models and switch between them without re-engineering their integrations, running on regional infrastructure for lower latency and in-region data handling where required.
Nemotron adds reasoning and tool-use capabilities, which Cassava says will help customers build AI agents that work through complex tasks and interact with business tools. Because the models’ weights, training data and recipes are open, developers can also fine-tune them for specific applications.
“Expanding our AI Factory across our markets is about creating the capacity and economics needed for businesses to build and deploy AI solutions at scale,” said Ahmed El Beheiry, Group Chief Technology and AI Officer at Cassava Technologies. He said the company was giving enterprises and service providers “more choice in how they access and deliver these capabilities to their customers.”
Security and cost controls for AI agents
Cassava said that as AI agents take on longer-running tasks and gain access to business systems, the infrastructure must include security and guardrails. It is using NVIDIA OpenShell to bring model compute, governed agent execution and independent security enforcement together for customers’ agentic workloads on the AI Factory.
It is also introducing a CAIMEx Token Optimiser that identifies information the system has already processed and, where appropriate, reuses it instead of reprocessing it. Cassava said this reduces computing requirements and helps customers use their AI tokens more efficiently, and that it applies to both Cassava-hosted and third-party models.
Part of a wider NVIDIA push
NVIDIA’s separate announcement on telecom operators building AI infrastructure lists Cassava in Africa among the operators expanding NVIDIA-powered infrastructure, alongside Deutsche Telekom in Germany and Fastweb and Vodafone in Italy.
The move extends the Cassava group’s run of technology partnerships this year. Its security arm, Liquid C2, recently agreed to resell KnowBe4’s security awareness training to African enterprises. Cassava also supplies the AI layer for the AI data centre that Vodafone Business and Elsewedy Electric plan to build in Egypt.
Cassava’s businesses also include Liquid Intelligent Technologies, Africa Data Centres, Distributed Power Africa and the fintech Sasai, and the group says it serves customers in 94 countries.
Cassava did not say in its announcement when customers can access the Nemotron models, what they will cost, or which markets the service will launch in first.
Leave a Reply