Money’s tight

Hey, Sheriff here 👋 

This week, as always, I have a question for you.

Whenever you see a software startup announce a massive raise, where do you think the money goes? 

For most consumer startups, the answer is simple: cloud bills and ads.

They’re the core growth infrastructure for the consumer internet we have today.

See, building tech products today requires a sprawling web of services: cloud hosting, APIs, communication tools, payment gateways, and now, artificial intelligence. 

Most of these services are built, priced, and billed in dollars by companies headquartered in the US. 

For African startups earning in naira, shillings, or cedis, this creates a quiet but crushing economic pressure. 

We’re calling it “Techflation”.

It’s not one thing, but a host of things that can snuff life out of an African startup before it even crawls out of the cot. 

And with AI changing the way startup economics work, it’s about to get worse.

Let’s look at each line item to see how it all adds up. Starting with…

The cloud bill nobody talks about

The core gospel of tech across the world is scale. 

You build a product, price it reasonably, get a lot of users, and rake in billions of dollars. 

But reaching a massive scale requires massive infrastructure. One of them is the cloud.

Running an online software means you have to serve your code to users via the internet. Historically, people bought servers (racks of computers) to do this. 

The problem is, as the number of people using the software increases, the amount of servers required also increases to the point where a company either can’t buy new servers fast enough, or will spend too much money doing so.

That’s where the cloud becomes important.

Some companies, like Google, Microsoft, and Amazon Web Services, have built millions of servers across the world and made it possible for people to rent them.

And for companies that scale to millions or even hundreds of thousands of users, the rent bill gets huge.

Netflix spends roughly $1 billion a year on Amazon Web Services. 

Snapchat committed $2 billion to Google Cloud and another $1 billion to AWS across multi-year contracts. 

Before Elon Musk's aggressive cost-cutting at X (formerly Twitter), the company was spending an estimated $1.5 billion a year on infrastructure. 

These bills are huge, but the companies earn in the same currency they spend. Their revenue scales alongside their costs.

In Africa, that ratio isn’t the same because revenues come in local currencies, but bills are priced in USD. 

So, in 2023, when Nigeria’s currency lost 50% of its value in a few months, costs went up drastically.

This pressure is real and documented.

In late 2023, Incentro, a Google Cloud partner, sued Twiga Foods, a Kenyan e-commerce startup, over a $2 million unpaid cloud services contract. 

Twiga had been paying around $84,000 a month for cloud hosting alone. 

A TechCabal investigation found that cloud fees and salaries are typically the two largest expenses for African startups. 

When Abolore Salami, a founding partner at Business Lab Africa, ran a LinkedIn poll asking founders about rising cloud costs, more than half of the respondents said the situation was worrying.

And cloud is just one slice of the infrastructure pie, because…

There’s a hidden stack behind every product

To make the point stick, let’s consider a ride-hailing app, a product that millions of people across Africa already use through Bolt and Uber. 

To make one work, you need: 

• The Google Maps API for routing and navigation. 

• A communication API like Twilio for in-app calls and texts. 

• Cloud infrastructure with auto-scaling to handle peak-hour surges. 

• A payment gateway. 

• A real-time support system. Each of these is a separate vendor, a separate contract, and a separate line item on the balance sheet, all priced in dollars.

Uber eventually moved its infrastructure to Google Cloud through a multi-year contract after years of running its own data centres across multiple regions. 

Lyft pays AWS $300 million a year. 

These companies can absorb these costs because their revenue runs into the billions. 

But what would it cost the average African startup to build a product with even half that level of infrastructure? A lot more than most seed rounds can sustain.

And the baseline is about to get even higher.

Because AI changes the math

For decades, software had a beautiful economic property: build it once, run it anywhere. 

After the upfront cost of development, the marginal cost of adding the next user was near zero.

This is what gave SaaS companies their legendary 80-90% gross margins and what made software the most scalable business model in history.

AI breaks that model. 

Every AI-powered interaction, every chatbot response, every smart recommendation, every generated image, consumes compute in real time. 

Unlike vanilla software, this is not a one-time cost. It's a metered utility, like electricity. 

OpenAI reportedly spent $8 billion on compute in 2025 alone. 

Data shows that AI-native companies typically operate at 50-60% gross margins, compared with the 80-90% margins of traditional SaaS. 

One clear example of this is Duolingo.

In 2024, the app had 130 million monthly active users and 46 million daily active users.

That year, Duolingo started powering its Max tier with GPT-4, offering AI roleplay conversations that help users learn a language in context.

By year’s end, its cost had grown to $203.6 million, 43% higher than the previous year.

This year, Duolingo decided to subsidize the cost of this AI for its free users, and that’s projected to drop its margins from 28% to 21%.

That haircut is a luxury most startups don’t have, especially in Africa.

Yet, AI is becoming a baseline requirement.

Users now expect intelligent search, personalised recommendations, and conversational interfaces. 

Products that don't have these features feel dated. 

This matters enormously for African builders because AI is rapidly becoming a baseline feature, not a differentiator. 

When this shift fully takes hold, most African startups building a consumer product will have gained an extra, persistent line of cost: inference. 

Every user prompt, every API call, and every agentic action will all be metered and billed in dollars. 

But revenues will mostly stay in local currencies. 

In this new world, the paradigm of zero marginal cost per user no longer exists 

And techflation sets in.

What is techflation?

It’s just like inflation, but for the cost of producing software instead of everyday goods. 

A perfect analogy for explaining this is a political agreement that happened in the 1940s: The Bretton Woods Agreement.

In 1944, just in the wake of the Second World War, countries across the world agreed to make the US dollar the world's reserve currency. 

It promoted order and enabled global trade, but it also meant that countries producing goods for their own populations still had to transact in a currency they didn't control. 

The resulting dynamic, where American monetary policy could effectively export inflation to other nations, created deep structural dependency.

Simply put, whenever the Fed prints money and weakens the dollar’s strength, it increases the cost of imports for countries that need to ship in dollar-denominated inputs like oil.

This effectively raises the cost of goods for these countries, too.

Techflation is quite similar. As these pieces of digital infrastructure get more expensive, the cost of producing software in Africa goes up. 

And companies need a new alternative.

In the case of the Bretton Woods Agreement, this cycle happened.

Over time, a coalition of countries decided the arrangement was too costly and created BRICS, a bloc of countries trading directly without using the USD.

This reduced their exposure to dollar fluctuations and the hegemony of a single financial system. 

Africa's tech infrastructure is stuck in its own version of this dependency. 

The cloud providers, the API vendors, and the AI model makers are all headquartered abroad and priced in dollars.

Plugging into this stack is, in a sense, like importing inflation.

But is there an alternative?

Let’s look eastward

One answer is already emerging from China. 

The open-source AI movement, led by labs like DeepSeek and Moonshot AI, is producing models that rival frontier proprietary offerings at a fraction of the cost. 

DeepSeek's models offer inference at roughly $0.27 per million input tokens, compared to GPT-5's $1.25 for the same volume. 

Moonshot AI's Kimi K2, released in 2025, matches GPT-5 on several benchmarks while costing up to eight times less per API call. It was reportedly trained for just $4.6 million.

These models are open-source, meaning developers can run them on their own infrastructure, fine-tune them for local use cases, and avoid the recurring per-token toll of proprietary APIs. 

For African startups, this isn't just a cost-saving; it opens up new possibilities. 

Instead of renting compute from OpenAI or Google at premium rates, builders can deploy capable models on their own local infrastructure.

And as for other services, such as cloud and APIs, local alternatives have been emerging.

• Yamify, a Congolese cloud infrastructure company, is building local cloud environments specifically designed to host open-source AI tools on African data centres. 

• Termii built Africa’s version of Twilio that’s ten times cheaper

• And Deimos built a cloud platform priced in local currency

Last month, I met Adeyanju Falade, a founder who runs FikazonSMS, a school management software in Lagos, Nigeria.

The company is ten years old, and serves over 250 schools and 58,000 students with software that handles school operations and student learning.

Students use it to learn, parents use it to track their kids’ performance, and schools use it to manage and pay staff, run their school, onboard new students, and even build their own website.

Yet, Adeyanju runs it all out of a few servers in his office.

When I asked him about it, he said, “I used to be on cloud platforms, but all that has happened is that I’ve watched my bills go up every year. And since my customers pay in naira anyway, there’s no point giving AWS all my money.”

Remember the guy whose company got killed by the Google Maps API cost? 

Someone just built an open-source map that works with AI, costs nothing, and can be hosted on an 8GB computer.

And he built it in one evening using AI.

The trend is clear. The tools needed to make world-class products are being democratized.

And there’s a shift from dependency on Western cloud and AI monopolies to open-source alternatives optimised for local needs and priced in local currencies.

As the economics of tech change in the future, many African startups will have to stay ahead of techflation with smarter alternatives.

What ways have you seen African startups beat techflation?

Let me know!

That’s it for this week. See you on Sunday for a breakdown on African tech.

Cheers,

The Tech Safari Team

PS. refer five readers and you’ll get access to our private community. 👇🏾

Wow, still here?

You must really like the newsletter. Come hang out. 👇🏾