The God Who Did it for NVIDIA will do it for You: Thoughts on Building Infrastructure Businesses in Africa

Sometime in 2013, Elon Musk and Larry Page (Co-founder of Google) had a fallout at a birthday party on what direction they felt AGI (Artificial General Intelligence) should go. While Elon advocated for safety and caution in developing these systems, Larry called Elon a speciest and subsequently dismissed his concerns as being overtly superstitious.

Roughly two years after that conversation, Elon Musk, Sam Altman, and Ilya Sutskever banded together to found OpenAI, a non-profit research lab tasked with ensuring that artificial general intelligence benefits all of humanity. Four years later, a fallout with Elon Musk, and extensive compute-related partnerships with Microsoft, ChatGPT was born.

The birth of ChatGPT kickstarted a cascade of events that eventually led to the AI Arms race as we know it today; Google (whose original research paper on Transformer Models was used as input to develop the GPT models) launched Bard in March of 2023, Meta’s open source LLaMa models got leaked online, Anthropic (founded by two former OpenAI employees) raised more than US$1.4Bn, and released its Claude and Claude 2 foundational AI models while adopting a unique approach to RHLF (Reinforced Human Learning Feedback) called Constitutional AI, Amazon partnered with Anthropic to drive its AI strategy, HuggingFace has become the go-to site for developers looking to test out Open Source machine learning models, and every Dick and Harry is building unique AI use cases by slapping fancy frontends on top of fine-tuned LLM models to create new AI products.

While all this ruckus has been ongoing, one technology company has quietly positioned itself as the silent benefactor of all AI development and innovation. That company is NVIDIA.

NVIDIA was founded in 1993 at Denny’s roadside diner in East San Jose restaurant by three seasoned engineering experts – Jensen Huang (present CEO), Chris Malachowsky, and Curtis Priem. NVIDIA originally started off building hardware for video games and was responsible for the chips on the early PS3 and Xbox consoles. However, sometime in 2012, someone realized that NVIDIA’s chipsets had another use case – they could be used to run compute heavy AI training functions that conventional Intel CPUs at the time were not configured to process at scale. That revelation led to a change in management strategy at the business. NVIDIA shifted its focus and began to develop new chipsets optimized to run training models at scale. They didn’t stop there, they doubled down on building an ecosystem of both software and hardware products designed to help AI researchers globally properly train and deploy new models. This led to the launch of new hardware solutions like the much-touted H100s used to train OpenAI’s models.

Today NVIDIA has essentially established itself as the global foundational infrastructure for AI model training and deployment, and how has this been reflected in its bottom line? Since OpenAI launched ChatGPT in November of 2022, NVIDIA’s stock has been up more than 400%. As of the time of writing this, NVIDIA is a member of the coveted trillion dollar club (companies with a market cap over US$1trn), holds an 80% market share of the global GPU market, and continues to innovate and deploy new solutions designed to keep it at the forefront of AI model development for years to come. NVIDIA is the perfect example of what an infrastructure business looks like and why playing at the infrastructure layer is very essential.

Infrastructure Businesses

The biggest misconception I have observed when people try to describe infrastructure businesses (within the technology industry) is assuming that companies providing APIs that extend unique capabilities to third parties are the definition of what an infrastructure business should look like. There are technology startups calling themselves infrastructure businesses primarily because they extend APIs that power some “market-defining use case”.

I personally believe this is a mischaracterization and tends to obfuscate what and how a real infrastructure business looks like and is supposed to function.

To create consensus, let’s attempt to paint a picture; Imagine a man, let’s call him Steve is trying to sell a product to a remote community somewhere in the southern part of Nigeria. Because this community doesn’t have a proper road network (or any at all), Steve is forced to transport his product via boat to get to this community, depending on the value of his goods and coupled with the fact that neither he nor his product can walk on water or supernaturally command storms to be still, this becomes a very risky endeavor. This naturally makes moving goods to this community an unattractive ordeal.

However, what if this community represents a huge market opportunity for not just Steve, but an entire value chain of producers who are itching to supply products to this community but are inhibited by the lack of a viable transportation network, and Steve decides to take it upon himself to build some kind of make-shift bridge to allow trucks to drive into that community and offload goods? That bridge essentially becomes the infrastructure for commercial activity for that community.

Now what if Steve decides to charge a toll for passage on that bridge? Excluding the value he earns from selling his product within that community, what he has virtually built is an infrastructure business that is essentially creating a platform for other businesses to extend and extract value.

The hallmark therefore of every infrastructure business is solving a HARD PROBLEM, a problem that inhibits an entire industry from functioning effectively or taking off by outsourcing that burden and allowing other players to run on the singular platform you create.

Infrastructure businesses are the hallmark of capitalist value capture primarily because they provide the fuel for an entirely new industry to exist (think of what Interswitch did with card switching in Nigeria) and allow a limited number of providers (an oligarchy of some sort) to position themselves at the nexus of a burgeoning industry while the market players go ahead to create new use cases to consume on their platforms (and ultimately generate revenues for them).

So how do you identify Infrastructure businesses? I think true infrastructure plays tend to have three key themes:

Infrastructure Plays are Hard (and usually have some defensibility embedded within them)

Almost anyone can tar the entrance to his/her estate, but not everyone has the financial capacity to build a highway or a flyover bridge, those jobs are infrastructure projects reserved specifically for governments because they not only have the financial resources to do so, but it is partly also their responsibility.

Most of the dominant infrastructure providers today have some inherent capabilities that make them well-positioned to tackle the problem they solve and creates some kind of defensibility around their offerings. Amazon built Amazon Web Services (AWS) because it realized it had an opportunity to outsource some of the excess compute being used on its native eCommerce platform to third parties who needed such compute but didn’t want to go through the stress (or cost) of setting up on-prem servers to run their operations (and managing scaling costs).

While a good number of companies have launched since AWS to provide competing cloud services i.e. Microsoft Azure, Google Cloud, IBM, etc. the kind of firms that have the financial wherewithal, technical expertise, and brand value to directly compete with AWS are limited. Coupled with the head start being first to market affords them, their focus on iterating and improving their core product, and their aggressive push to make sure the startup economy is run on their rails (via their generous credit programs), AWS is continuing to build defensibility within its core cloud services offering.

Bringing this home, companies like Interswitch, NIBSS, Remita, and UPSL have built unique infrastructure products that are exceptionally hard to replicate regardless of how much capital any intending contender has at their disposal.

Anyone (well not anyone) can acquire a switching license from the Central Bank of Nigeria that allows them to process card transactions, but dominating the offline acquiring card business that both Interswitch and UPSL are massive contenders in is another ball game entirely. Building a real-time payments rail like NIP (NIBSS Instant Payments) or RITS (Remita Interbank Transfer Service) that has connectivity across all Deposit Money Banks (DMBs) and allows for seamless direct account debits and credits is more than just having a license and raising money from venture capitalists.

These are really hard businesses to disrupt not necessarily because their products are the best (as at the time of writing this article, NIP is down), but because of the native advantage that comes from being first-movers in a new industry. I’ve written extensively on this in my Understanding the Nigerian Fintech Evolution article.

Infrastructure Plays Serve Large Markets

The next most important thing that makes an infrastructure business relevant is that it empowers a large market. Back to our earlier analogy about Steve; If very few people were interested in selling products to the inhabitants of the community in question, the infrastructure he built (the bridge) wouldn’t necessarily be important or highly impactful. The fact that multiple other businesses were trying to extend products to that community, but the lack of a secure transportation network had inhibited them from doing so is what makes building a bridge (and charging a toll fee for using it) meaningful.

A lot of tech companies build infrastructure for niche markets that tend to be non-existent or too minuscule to be worth the effort in the first place. The nature of infrastructure businesses means that niche opportunities are just not their target market.

Due to how infrastructure businesses are modeled, and value is extracted, real infrastructure businesses need to build solutions that empower a large number of users.

Within the earnings value chain for most industries, the infrastructure provider is usually the smallest benefactor, however, the scale of distribution they command tends to make up for the minuscule margins per product sold.

For instance, NIP (NIBSS Instant Payments) charges N3.75 (US$0.003) per transaction to banks that leverage its service to provide funds transfer capabilities to their customers. The banks mark this up (depending on the transaction band) to extend at N10 (US$0.008), N20 (US$0.016), and N50 (US$0.042) to their customers, thereby positioning NIBSS as the smallest benefactor in that transaction. However, when you factor in the fact that NIP processed 9.6bn transactions in 2023 alone, that N3.75 per transaction quickly turns into N36Bn (US$30m) in annual revenues from just funds transfer operations alone. This is the same reality other infrastructure providers like Remita, Interswitch, and UPSL face. Interswitch and UPSL are massive infrastructure providers for offline acquiring on POS terminals within Nigeria – Merchant Service Charge (MSC) for POS transactions is usually 0.5% capped at N1000, except in cases where transactions are native (issuer and acquirer are within the same scheme) or some kind of card routing is employed, Interswitch and UPSL will keep a small fee on all transactions processed. When you factor in the fact that more than 1.3bn POS transactions occurred in FY 2023 alone according to NIBSS, you realize the massive opportunity sitting with these infrastructure providers.

An infrastructure service designed to serve a niche use case that hasn’t been tested to either have massive existing or latent customer demand is usually an experiment that ends up badly – the margins are too minuscule to justify the operation and the business is eventually forced to shut down (or pivot) due to lack of viability.

Infrastructure businesses must either cater to large existing market demand or potential market demand that can be properly tested and verified.

Infrastructure Plays Can Be Physical.

Infrastructure doesn’t necessarily have to be digital. A lot of infrastructure businesses in Nigeria aren’t core technology companies. MTN and the other telcos for instance are the infrastructure providers for the entire digital economy in Nigeria as we know it.

As more people shift from cable TV to streaming platforms, the telcos benefit, as more Gen Zs and Alpha’s spend their time mindlessly scrolling through TikTok and Instagram, the telcos benefit, as more businesses decide to move online, the telcos benefit and as more content is being delivered over the internet and consumed multiple times a day by different people, the telcos are poised to continue to extract significant revenues for their businesses and stakeholders.

Another unlikely infrastructure provider within Nigeria is Dangote Cement the eponymous jewel of the Dangote Group. Unfortunately, lack of a reliable digital payments platform is not one of the most daunting challenges facing Nigeria today, the US$2.3 Trn infrastructure deficit is, and Dangote, BUA, and Lafarge Cement are among the few companies poised to solve that problem.

While Chinese firms may be responsible for the civil engineering required to construct roads and bridges, Dangote Cement is a key supplier of the raw materials (cement in this case) required for that process to exist in the first place. The Federal Government of Nigeria’s 2024 budget earmarks N7.72trn (US$6.4Bn) for capital expenditure, capital expenditure typically involves the construction of new roads, bridges, buildings, etc. I think it’s safe to say at least 1% of that money will appear in Dangote Cement’s books as topline revenue.

On a side note, if you looked at Dangote Cement and MTN on the Nigerian Stock Exchange (NSE) and redacted their names, you would think that both companies were tech firms. Both Dangote Cement and MTN trade at 29.4 and 25.43 price-to-earnings ratios respectively (the ratio of a company’s share price to its earnings per share – usually used to identify companies investors believe have high growth potential). Those P/E ratios are typically reserved for tech companies (Meta trades at 30.82, Alphabet at 28, and Apple at 30.5), which gives you a glimpse into where investors see growth potential going into the future in some emerging markets. For context purposes, Cemex and Anhui Conch Cement (US-based cement manufacturers) trade at 1.068 and 6.7 P/E ratios respectively, Verizon trades at 8.05, and AT&T trades at a negative P/E of -11.

Note: to avoid intellectual dishonesty, T-Mobile trades at 24.9 P/E which may be more of a testament to T-Mobile’s performance as against an affirmation of the industry’s growth potential (considering the other two major players in its space do not follow the same trajectory).

Banking as a Service in Africa

As technology adoption continues to percolate through the fabrics of society, there will continue to be opportunities for companies in the technical abstraction space (firms that extend technical capabilities of any sort to non-technical people or teams e.g. drag and drop app builders, no code development tools, etc.).

BaaS companies are generally textbook examples of the importance of technical abstraction. They help firms cut down on product development time, reduce compliance burdens, and make it easier to embed financial services. However, the real question is are BaaS providers really infrastructure businesses? If we decide to pass BaaS businesses through our erstwhile identified framework, we may unlock some new insights:

For starters, BaaS businesses aren’t necessarily difficult to build. While not every Dick and Harry can build a BaaS business, every motivated and technically sound team of Dick’s and Harry’s can build one. Building a BaaS provider (in Nigeria) on a high level involves acquiring licenses to provide a plethora of financial services typically; wallet APIs, digital token vending, credit origination, card issuance, etc. The next step involves building the technical stack and yes getting sound engineering talent to do this isn’t necessarily a walk in the park, but it also isn’t impossible either, and it is much easier than replicating what earlier FinTechs like Interswitch have done by developing nationally utilized card switching rails.

The next part of the framework is serving a large market; the honest truth is that on some level, the growth of BaaS seems to be predicated on the growth of digital payments in Africa (which is clearly a growing market), in reality, however, the growth of BaaS is actually predicated on the growth of Fintechs offering digital payment solutions. An established bank or even a tier 1 fintech is unlikely to rely on a BaaS provider unless either temporarily or to quickly embed a solution/feature they do not want to commit engineering resources to build (this is especially true when wholly owning the stack or acquiring a license just to offer that feature doesn’t seem mission critical). In that case, relying on a BaaS provider tends to be the right way to go.

The dearth of high-quality engineering talent due to the Japa syndrome also creates a unique opportunity for BaaS providers to extend their products to firms without the requisite engineering resources to implement or build out new offerings (provided the BaaS provider’s engineers are not Japaing too).

So, startups who want to accelerate their go-to-market and other established firms seeking to add new offerings to their product stacks (that are tangential from their core propositions but reflect emerging market opportunities) are likely to embrace BaaS players.

However, the biggest market opportunity for BaaS providers will slightly draw its thesis from the Fintechnolization concept postulated by Prof Ndubuisi Ekekwe which states that every digital platform will eventually offer a fintech solution at some point. In this case, I think the opportunity to help established locally based “boring companies” embed financial services into their current offerings to extract more value from their users is a huge opportunity that can create massive outcomes for well-positioned BaaS providers.

Embedded payments or card issuance may be a good starting point (even though I personally don’t understand why someone would launch a dedicated card issuance startup), but the meat of the opportunity lies in embedded lending and empowering more firms to profit off their existing distribution. The indicators that this is a huge market opportunity are as clear as crystal; consumer credit within Nigeria in the last 9 months is up 32% or N740bn (US$616Mn) indicating rising credit consumption within the country (largely driven by rising inflation). InDriver recently announced that it was exploring offering financial services to its drivers starting with credit.

While this market opportunity may not have been fully unlocked yet, it does hold the promise of another core revenue driver when it eventually does.

So yes I do not necessarily categorize BaaS companies as infrastructure players because they aren’t necessarily hard to build and they do not serve large markets where they have some kind of inherent defensibility within their models, however, I do believe that unlocking the “boring company” market opportunity may present a unique opportunity for companies playing in the BaaS space, especially from a lending perspective.

The Jumia Miss

Excluding Yakubu Aiyegbeni’s horrendous free post miss against South Korea that saw Nigeria crash out in the group stages of the 2010 World Cup, the most glaring example of an entity missing an opportunity to win a market to me is Jumia’s logistics infrastructure miss.

I’ve written 3 public articles on Jumia so far – with my most recent being in 2021, so this is probably just a slight rehashing of my personal sentiment about the company.

Jumia was originally positioned to be the Amazon of Africa – and that narrative alone was enough to drive more than US$767 million in private capital to the company. Jumia also realized that to build the eCommerce offering it had envisioned, it needed to build the entire eCommerce stack end-to-end, especially the logistics part which was almost non-existent at the time.

However, the one thing Jumia failed to envision was a world where social commerce or even non-Jumia e-commerce would so dominate the market that Jumia itself would only be a fraction of that market. In such a world, the logistics infrastructure would have been a key necessity for that market to take off and anyone who had built a RELIABLE, WELL DISTRIBUTED, AND EFFICIENT logistics layer was well positioned to take that market.

There are two reasons Jumia was well positioned to solve this problem – one was they already had the incentive to do so; they needed that logistics stack themselves to complete their product proposition and continue to improve the user experience for their users and second was they had the money required to do that – funds to hire the most skilled logistics professionals, acquire delivery technology (vans and bikes) and set up warehousing systems to make this work. An offering like this from Jumia would have been a market leader because the economies of scale (serving its own inherent demand and proposed latent market demand) would have allowed it to offer the best service at the cheapest price.

Ideas around offering segmentation for delivery aggregators like Chowdeck, Foodcourt, etc. that prioritizes their services and gives them priority access at marked-up fees would have surfaced, a whole ecosystem of social commerce sellers leveraging Jumia’s infrastructure to distribute their products would have been born and the network effect of more riders coming on their platform to improve distribution density would have positioned Jumia as a frontline player in the local logistics space, and with their African expansion drive, the same proposition could have been extended to other African countries positioning Jumia as the PAPSS (Pan African Payment Settlement System) for African logistics.

In the same way Amazon’s eCommerce platform has become somewhat of a front for its more profitable business line – AWS (Amazon Web Services), Jumia eCommerce would have been a front for Jumia Logistics. I believe that was Jumia’s Moniepoint moment (the moment a company identifies a friction in a market it can solve and leverage on to compound rapidly and morph into a rocket ship).

Today, there is no unified logistics market in Nigeria, multiple players own various segments – from KwikDelivery to GIG Logistics to Red Star Express to the myriads of independent dispatch rider companies scattered across Nigeria. A unified logistics market was the key opportunity Jumia would have leveraged in its Hey Days to build and dominate, thereby positioning itself as the logistics infrastructure for Nigerian and by extension African commerce.

To be fair, hindsight is 20/20, and executing on most of what is here is much more difficult than writing about it, needless to say, Jumia does have a third-party logistics business in Nigeria today as at the time of writing this (which I sincerely doubt operates at the scale hinted in this piece), and is now focusing its effort on reaching reasonable unit economics in its core e-commerce business.

Conclusion

I personally think that due to the maturity level of the African ecosystem, playing as an infrastructure provider positioned to extract value from new opportunity vortexes that may spring up now or in the future is one of the key ways to build massive and impactful technology businesses in Africa.

I hope that after reading this, you go to bed this night praying to the God that did it for NVIDIA to give you the wisdom to identify new and latent infrastructure opportunities, and to properly position your business to capitalize on them.

Inspired By The Holy Spirit