After investing in tech startups throughout Africa for the previous seven years, Enza Capital has launched a devoted fund just for synthetic intelligence (AI) startups and AI-enabled startups in Africa, on the view that African enterprise capital is maturing and that specialist funds discover and assist the very best corporations extra successfully.
“AI is the most consequential technology shift in a generation, and it needs to be harnessed to create value for a broad set of people,” Mike Mompi, Enza’s managing associate, stated on a name with TechCabal. “That is why investing in AI in Africa now is an incredible opportunity and even a responsibility.”
The agency’s identify comes from a Zulu phrase that means ‘to build’, ‘to make’ and ‘to create’; it has raised $58 million throughout its first two funds and says it has made 75 investments throughout 47 corporations in 9 African nations.
Mike Mompi started Enza in 2019 with two co-founders. Sir John Lazar ran Metaswitch, a communications software program firm backed by Sequoia Capital and Francisco Partners, which Microsoft purchased in 2020. David Cohen labored at Goldman Sachs in London earlier than spending about 20 years as a associate at Farallon Capital, the hedge fund.
By Enza’s rely, AI took about 60% of world enterprise funding in 2025 and just over 5% in Africa. While Enza doesn’t anticipate African startups to compete with frontier fashions as a result of the capital required is too massive, it is betting on corporations that adapt AI to African circumstances.
These African variations might embrace voice instruments that perceive Pidgin and code-switching for clients who can’t learn, smaller fashions educated on native information, and computing that works the place connectivity is patchy. The agency calls this the constraint benefit, the concept merchandise constructed for Africa’s limits also can work far past it.
That thesis shapes what Enza seems to be for. As constructing a prototype will get cheaper, the agency says, founders want deep data of their market and a moat that can’t erode: information nobody else has, present routes to clients, or a product constructed into the each day work of the individuals who use it.
Enza additionally screens arduous on value. AI startups pay for each question their product runs, so a enterprise that assumes low cost inference will be worn out if that value doubles. The agency backs corporations with excessive margins or a clear path to them and expects open fashions and cheaper native computing to carry prices down.
In this conversation, Mompi explains why Enza is elevating a fund devoted to AI, why he believes African AI corporations will win by native information, distribution and match into clients’ each day work, what he seems to be for in founders, how he thinks about the price of operating AI fashions, and the way Enza plans to return cash to its traders.
This interview has been edited for size and readability.
Why a devoted AI fund for Enza?
African enterprise capital is following the sample of different rising markets. Funds begin as generalists and specialise because the market matures. A specialised fund is higher at discovering the very best alternatives and serving to corporations attain good outcomes, as a result of focus brings networks, relationships and perception into what is taking place globally and domestically.
We have invested in African expertise for almost a decade, and with our staff, our data and our mixture of native and world networks, we imagine we’re one of many companies greatest positioned to run an AI fund for Africa.
Sir John Lazar spent three a long time as an engineer, raised cash from Sequoia Capital and Francisco Partners, and bought that enterprise to Microsoft. He is president of the Royal Academy of Engineering, which runs the Africa Prize for Engineering Innovation; that is the place we hung out collectively earlier than launching Enza. He additionally chairs the Raspberry Pi Foundation, which owns the buying and selling enterprise Raspberry Pi, and guided that enterprise by its itemizing on the London Stock Exchange.
Sir John runs our workplace of the CTO, alongside two CTOs in residence, which is uncommon for a enterprise staff on the continent. One was head of engineering at Yassir, the Algerian tremendous app. The different led engineering at Founders Factory Africa. That depth lets us assist our corporations on expertise. We additionally run a group the place CTOs throughout our portfolio be taught from one another and workshops with outdoors specialists on placing AI to work for corporations which might be AI-first and people that aren’t.
We introduced AI Tinkerers to Africa as an working associate and sponsor. It is a community of about 1,300 builders working on AI merchandise in Nairobi, Lagos, Cairo and Johannesburg, and now we have run classes with companions together with OpenAI and Microsoft. The purpose is to present engineers and early founders a place to indicate their work and meet individuals they won’t in any other case meet.
Since 2025, now we have revealed our thesis on what AI means for Africa by our Substack, African Vector. It is the place we share our considering and invite problem.
AI is probably the most consequential expertise shift in a era, and its worth wants to succeed in a broad set of individuals. Private AI funding has grown fivefold during the last 5 years, however Africa’s share of that development has been marginal. AI took round 60% of world enterprise funding in 2025 and just over 5% in Africa. That hole is a gap, and we imagine AI will probably be one of many greatest long-term sources of worth on the continent.
Beyond the scale of the chance and the funding hole, what attracts Enza to AI investing?
AI is collapsing the price of constructing issues, and it lets African founders promote globally from day one. More particularly, three issues.
The first is native context. There is a chance in AI methods educated or tailored for African languages, cultures and working circumstances. Building the identical factor as Anthropic or OpenAI would take extra capital than Africa can increase. The query is which fashions swimsuit which locations and makes use of, from small language fashions with hundreds of thousands or billions of parameters to massive ones with a whole lot of billions or trillions.
Global fashions usually miss native context. We noticed it early on with facial recognition constructed within the US that would not recognise non-white faces, and we see it now with voice AI that struggles with Pidgin and with audio system switching between languages mid-sentence. A financial institution buyer or cellular subscriber who can’t learn and might solely use voice is shut out. Add background noise and completely different dialects, and the issue grows. Africans at the moment are constructing fashions that deal with this, educated on African information and deployed in African corporations. We have invested in a single, Aethex AI, and others embrace Lelapa AI, Intron and ToumAI.
The second is proprietary information. Companies want defensibility, so they don’t seem to be constructing a characteristic that sits on a frontier mannequin’s roadmap. They can construct distinctive datasets and suggestions loops that make their AI extra correct over time. Hakimu AI, one in all our corporations, has entry to a whole lot of hundreds of beforehand unpublished Supreme Court case information, which it makes use of to coach a small language mannequin that helps East African courts ship quicker, extra constant justice.
The third is infrastructure and compute. How do you make AI work the place connectivity is restricted or patchy? That means operating fashions on gadgets on the edge, mixing sorts of {hardware}, and constructing methods that can’t assume abundance. Solutions constructed for these constraints also can work globally.
Five years in the past, we invested in Guidewheel, based out of Stanford, whose first deployments had been in Kenyan factories. It collects machine information and processes it within the cloud to assist factories run longer and minimize vitality prices. Its clients report a median of 41% extra output from present gear, and now we have seen vitality prices fall by about 15% a month. It has since scaled throughout North America, elevating from Greycroft and from Decarbonisation Partners, a three way partnership between BlackRock and Temasek.
Many traders I converse to see AI as a software for fixing native issues quicker, with the sting coming from native context. Do you agree?
Yes. We see AI as a means to an finish. We are constructing an AI fund as a result of we imagine AI will let a era of African founders, companies and probably governments remedy issues that would not be solved earlier than.
AI adjustments the economics. Smaller groups can construct extra formidable corporations with much less capital, which issues much more in skinny capital markets like ours.
We see two sorts of corporations. AI-native corporations have AI at their core, and their income comes from it. AI-enabled corporations, which is virtually each expertise firm elevating cash in the present day, spend a part of their finances on AI instruments to work extra effectively.
You say Enza has completed probably the most AI investments in Africa not too long ago. How many offers, over what interval?
I imagine we’re probably the most lively enterprise fund making AI-native investments in Africa, however I’d be comfortable to be taught that others have completed extra so we will co-invest. The alternative is far bigger than anyone agency. We have made 11 AI investments and are working on our twelfth.
Quro Medical, a South African well being tech firm, is one now we have exited.
Hakimu AI is the authorized tech enterprise I discussed. When individuals in New York or San Francisco ask how one can construct AI in Africa, it is one of many examples they perceive.
Lua was the primary African AI enterprise to undergo Y Combinator in its Fall 2025 batch. We had backed one in all its co-founders in an earlier, non-AI enterprise. Lua builds AI brokers for companies, largely African corporations, although it has clients on 5 continents.
Aethex AI builds native language fashions, beginning in Francophone and Anglophone West Africa, and is already promoting them. It additionally constructed Kora, one of many main benchmarks for native language fashions.
WideBot, began in Egypt, constructed the primary Arabic language mannequin that handles a number of dialects. It now has 14 governments and greater than 400 companies as clients.
At the infrastructure layer, there are Guidewheel and Satlyt, which introduced its seed spherical in October. Orca Fraud and Buguard are AI-first cybersecurity corporations.
What does the perfect startup for this fund appear to be?
I’d describe what we glance for, because it doesn’t rely on a particular enterprise or sector.
People nonetheless matter most. Founders want deep experience available in the market they’re constructing for. Prototypes are getting simpler and simpler to construct, so a founder who doesn’t know the market finally ends up alongside everybody else who constructed one thing related. We look for founders who’re formidable and reasonable, with ambition that comes from deep data of the market.
Then it relies upon on the place a firm sits. We consider the stack in three layers: infrastructure, platform and software. Building will get cheaper as you progress up towards functions, so there is extra competitors, and also you want a stronger moat. At the appliance layer, that moat will be proprietary information, distribution by present buyer relationships, or workflow integration, the place the product turns into a part of the each day work of the individuals who purchase it.
Moving down the stack, capital prices rise, and competitors falls. Look at Cassava and the GPUs it is bringing into Africa. We are constructing one thing for the ecosystem with Cassava that we are going to announce shortly. Few corporations can function on the scale Strive Masiyiwa can, along with his experience and entry to world capital. Infrastructure wants economies of scale. Edge compute is one alternative, and that is the place Satlyt sits. As an early-stage software program startup, it plugs into present scale, optimising communication between satellites and GPU capability, with out constructing that scale itself.
The platform layer, between the 2, is the trickiest, as a result of competitors comes from each facet. There you want a actual expertise moat that a massive language mannequin firm won’t roll over.
The very last thing is the enterprise mannequin. Many AI corporations are transferring from charging for utilization to charging for outcomes. If a financial institution receives 10,000 calls a day and might deal with 1,000, it misses 9,000. A voice agent that handles all of them and resolves clients’ issues may very well be price a whole lot of hundreds, or hundreds of thousands, of {dollars} to that financial institution and may very well be priced as a share of the revenue it creates.
How do you concentrate on exits for this fund, and might you inform me concerning the exit you talked about?
For an funding ecosystem to work, traders want their capital again plus a return for the danger. Exits at scale have been lacking in Africa, however now we have sufficient indicators to imagine they’re coming.
We are seeing extra mergers between corporations. Industry gamers are constructing groups to amass expertise corporations. Global corporations are trying for development outdoors their residence markets. Several IPOs are in progress, which is elevating consciousness and participation.
We are additionally working on it ourselves. With co-investors, we helped launch Clearflow Markets to make it simpler to purchase and promote shares in personal corporations. We can’t wait for liquidity to seem. We must construct routes to it for our corporations, our LPs and ourselves.
Quro Medical constructed wearable expertise, utilizing machine studying, in order that a affected person who would have spent ten days in hospital might spend two there and eight at residence. It labored so nicely that hospitals started asking why sufferers ought to solely put on it at residence. We co-led a spherical early in Enza’s life, and in 2025 we bought our full stake, alongside different traders, to the biggest publicly listed hospital group in South Africa.
Application-layer AI startups pay for mannequin and inference prices, which squeezes margins. What gross margins does Enza wish to see early on?
If you construct a enterprise assuming a sure value for your core product and that value doubles, you’ll be able to exit of enterprise in the event you can’t adapt.
One motive an Africa-specific AI fund is smart is that our corporations do not need to rely on world fashions to resolve native issues. Open-source and open-weight fashions, and initiatives we’re a part of, are chopping the price of coaching, operating fashions and accessing compute. Some work wants GPUs, and a few wants CPUs, so figuring out the place a mannequin ought to run issues. There are additionally initiatives such because the Enzi Prize, a $10 million prize for extremely succesful, open and inexpensive AI in Africa.
Margins matter. Many companies acquired into hassle between 2020 and 2022 by chasing headline numbers reminiscent of complete transaction worth and ignoring margins. Net revenue margin and free money circulation matter much more than gross margin. We look for companies with excessive margins or a clear path to them. Companies are additionally innovating with smaller fashions, native compute and on-device AI, so we anticipate prices to fall.
There is additionally big and doubtless unreasonable overinvestment in AI infrastructure globally, which dangers overcapacity. Not all AI is fundable, and capital put to work with out a deep understanding of the expertise dangers being misplaced. That is why it issues to grasp the place worth is being constructed and the place it’ll keep, for every enterprise, every nation and the continent.
