In eighteen months, Atlantica Ventures has led or co-led rounds for four African or Africa-linked startups building the underlying layers of AI infrastructure — semiconductor design talent, vector search, DevOps automation and compliance orchestration. It is a marked departure from the fintech- and logistics-heavy portfolio the Lagos firm assembled over its first five years.
The latest, announced on 1 September, is a $1.9 million round for ChipMango, an AI-native semiconductor technology company that trains engineers in chip design and verification and places them on commercial hardware programmes. Atlantica led, with DFS, Kaleo Ventures, Madica, Trilinear Technologies and Malta Ventures participating.
The ChipMango bet
ChipMango’s thesis rests on a labour shortage rather than a technology gap. Deloitte estimates the global semiconductor industry will need one million additional skilled workers by 2030 — roughly 100,000 a year — as revenue grows more than 80% from its 2021 base of just over $550 billion.
The model is to train engineers in browser-based design environments built around industry-standard electronic design automation tools, place them on paid commercial engagements — including verification work for Trilinear Technologies, a supplier of DisplayPort and multimedia IP — and route the resulting expertise into its own edge-AI and hardware products. ChipMango is an Arm Approved Training Partner and works with Ohlone College in California and the University of Pretoria’s Carl and Emily Fuchs Institute for Microelectronics.
Headquartered in Delaware with operations in the US, South Africa, Nigeria, Rwanda and Malta, the company said the capital will fund a European design centre in Malta and workforce development in Kigali.
“The semiconductor industry’s binding constraint is no longer capital, it is people,” said Aniko Szigetvari, a founding partner at Atlantica, arguing that ChipMango converts African engineering talent into chip-design capacity already used by commercial customers.
Semiconductors remain close to untouched territory in African AI. The nearest comparison in recent memory is Egypt’s Infinilink, which raised $10 million in 2025 to develop optical connectivity for AI data centres. Two companies do not make a sector, but they mark the first African activity at a layer the continent has had no presence in at all.
A pattern, not an outlier
ChipMango is Atlantica’s fourth AI lead or co-lead since March 2025, and the four deals span most of the stack rather than repeating a single thesis.
Atlantica led NOSIBLE‘s $1 million pre-seed in March 2025. The South African-founded company, led by Stuart Reid, is commercialising a disk-optimised approximate-nearest-neighbour search index it says cuts the cost of large-scale search relative to existing methods. It has not published independent benchmarks for that comparison.
It led Salus Cloud‘s $3.7 million seed in June 2025, co-led with P1 Ventures, with Lofty Inc. Capital’s Idris Bello, Everywhere Ventures and angel investor Timothy Chen participating. The Cape Town AI-native DevOps platform, founded by Deimos chief executive Andrew Mori, prices its managed enterprise tier from $5,000 a month against what it says is the cost of a single DevOps engineer in Nigeria, Kenya or South Africa.
It co-led Cybervergent‘s $3 million seed in March 2026 alongside Ventures Platform. The governance and security-posture platform says more than 150 organisations across West, East and Southern Africa use it, and that it has cut compliance and audit work by over 70% — a figure it has not had independently verified.
The involvement predates this run. Atlantica co-invested with Mozilla Ventures in the roughly $2.5 million seed backing Lelapa AI, the Johannesburg NLP company founded by Pelonomi Moiloa and Jade Abbott, which builds the Vulavula speech and text engine and the compact InkubaLM model for low-resource hardware. Moiloa told ITWeb this week that Lelapa is now adding voice generation and preparing to take its efficiency methodology beyond Africa.
Small cheques, deliberate shape
Individually none of these is large. Combined, the disclosed amounts across ChipMango, NOSIBLE, Salus Cloud and Cybervergent come to under $10 million.
What distinguishes the run is its shape. Chip-design talent sits at the hardware layer, NOSIBLE at retrieval infrastructure, Salus Cloud in developer tooling, Cybervergent in governance, Lelapa in language and data. Szigetvari has described Atlantica’s approach in similar terms publicly, saying the firm looks across the full stack and sees opportunity in applied AI for cold-chain logistics, agriculture, cybersecurity, DevOps and compliance rather than in large language models alone.
That is a shift for a firm whose better-known bets — Sabi, OnePipe, Sendy, Curacel — sat in fintech, e-commerce infrastructure and logistics. Atlantica was founded in 2019 by Szigetvari and Ik Kanu and closed a $50 million first fund backed by the European Investment Bank and the EU’s Boost Africa programme. Third-party investor data puts its typical participation at the $10 million to $50 million range — well above the sub-$4 million tickets in this AI cluster, which suggests early thesis-testing positions rather than scaled bets.
Against a thin backdrop
African AI startups raised about $1.25 billion cumulatively between January 2019 and March 2025, with roughly 207 AI-labelled startups tracked between 2022 and 2025 — figures researchers caution are incomplete given inconsistent sector-tagging. Germany’s development ministry cited the same $1.25 billion total at the EAC AI Challenge launch, adding that 80% went to four countries.
More broadly, African startups have raised over $1.6 billion so far in 2026, roughly matching the same period in 2025 even as disclosed deal counts have fallen — fewer, larger rounds, concentrated in later-stage and debt financing, with fintech and mobility still taking most of the capital.
Atlantica’s cluster runs against that grain: small, early rounds in a segment with no mega-deals.
It also lands squarely on an argument Roger Jantio has been making in these pages. His June piece held that Africa’s AI problem is investability rather than awareness — too few companies ready for serious institutional capital — and his August series argued that capital must match a company’s stage rather than defaulting to venture equity for everything.
Whether the positioning proves prescient depends on factors outside any investor’s control: global chip demand, the pace at which African enterprises adopt AI-native tooling, and whether larger funds follow into the next rounds. For now, Atlantica is betting that the defensible openings in African AI sit in the infrastructure beneath the applications rather than in the applications themselves.
