Nigerians are building affordable alternatives to AWS and Google Cloud | Hacker News Reader

Nigerians are building affordable alternatives to AWS and Google Cloud

restofworld.org

131 pointsNDAjam53 comments

> While setting up a data center can cost millions of dollars, small Nigerian teams have found a workaround to this with “colocation”

Isn't that what everyone does when starting up a hosting company?

Looking at the websites of those companies, I also wouldn't characterize them as AWS or Google Cloud competitors. They offer Virtual Private Servers, FTP storage, maybe an S3 equivalent and Kubernetes. Maybe we can talk about S3 and Kubernetes being AWS-like, but those that offer only Virtual Servers and FTP space are competitors to "boring" web hosting services, no matter how often they use the word cloud.

It's cool that these services also exist in Nigeria, and I think it's right that Nigerian websites use Nigerian infrastructure. But the article does a poor job talking about it

Yeah...you're not really building an 'AWS' until you're building a hyperscaler and leveraging massive economies of scale. This is more "we're building a service that addressed local reality with local solutions", and that's a pretty awesome thing just in itself.

>you're not really building an 'AWS' until you're building a hyperscaler

AFAICT there's no claim that they're "building an AWS" (or attempting to). The title is "Nigerians are building affordable alternatives to AWS and Google Cloud".

So, these are alternatives and a skim of the article seems to follow.

I think we've been cloud-indoctrinated here (until recently) to believe viability means fullly-scalable cloud infrastructure.

I don't see the real difference between saying this and saying a Raspberry Pi is an alternative to AWS.

You're correct that there's an analogy continuum here that starts at "ridiculous" and ends with "nearly indistinguishable".

I suppose like many things, it's down to a reasonableness test.

You should get your eyes checked...

We're pretty far into the "ditching cloud" hype cycle at the moment. While I think both sides have valid arguments, I think the whole drama tends to lead to unhelpful titles like this one.

Using this same criteria, I'm "building an affordable alternative to AWS" in my house by deploying an HA instance of Proxmox.

I don't think revenue numbers reflect that we are in a ditching the cloud hype cycle, particularly given emerging political risks.. At any given time there is hype for every possibility with controversy or ramifications and the software market is so much larger than what many of us were used to that any group seems large enough to be most of the market.

Oh, I agree entirely with you. When I say hype, I mean there's a lot of positivity around these "We left the cloud" blog posts. I sit in a position within my company where I make these types of decisions (how our infrastructure is provisioned), and every few weeks or so for the past year I've gotten sent some blog post about X company moving off the cloud and reaping Y benefits.

I estimate that many of these cases were self-inflicted with a general misunderstanding of the cloud pricing model. Amazon has a literal pipeline (AKA certificate program) to maximize its revenue. It's so bad that I now include interview questions trying to suss these individuals out (design me X system, and the given solution uses 15 different AWS services).

Now, that's not an entirely fair representation of the problem, as I've seen genuine business cases that shouldn't be in the cloud, period. That's why I say both sides have an argument. But as I said, my opinion is that actual cloud revenue suggests that it's just a vocal minority group spinning up most of the discussions.

It’s also funny because AWS uses (or at least used) colo facilities as well. They just never advertised it afaik.

>Looking at the websites of those companies, I also wouldn't characterize them as AWS or Google Cloud competitors

I'd say they're more "alternatives" at this point, as the article's title states.

But...

>They offer Virtual Private Servers, FTP storage, maybe an S3 equivalent and Kubernetes.

The number of AWS customers that could survive on these services alone is likely higher than we'd suspect, especially for small companies.

Throw in an RDS-equivalent and maybe load-balancing, and the number rises dramatically.

Of course, this also assumes the connective tissue is also in place (e.g. VPC, security groups, etc).

EDIT: I should've said "the number of AWS customers that currently survive on these services alone..."

> The number of AWS customers that could survive on these services alone is likely higher than we'd suspect, especially for small companies

They could, but they will not.If your AWS is small there is no real motive to look for alternatives. Just stick with the brand you know.

Even a lot of developers are very resistant to using something 'non-standard'.I have had "thats a weird thing to do" reactions to suggestions such as running Postgres on EC2 because it would be easier than using RDS for a small DB.

>They could, but they will not.

I intended to say "the number of AWS customers that currently survive on these services alone...". I've updated the comment.

My comment was aimed more at those who assume that providers offering only these services would mean they're technically non-competitive with AWS.

My overall point is that these services likely represent what many customers are actually using. If you're in an enterprise shop, this may be less true. Same if you're in a funded startup that's convinced it needs infinite scalability and architects accordingly.

But, I do agree that it's hard to convince people to resist the dogma. This may be changing as people are starting to question the value of Big Cloud vs its costs, but we'll have to wait for the dogma to follow to see any real change (for those who have access to Big Cloud).

> If your AWS is small there is no real motive to look for alternatives. Just stick with the brand you know.

Lots of reasons. Maybe you want a vastly simpler interface and feature set. Maybe AWS doesn't offer servers in your country, and you've got sensitive data. Maybe the US is threatening your country with economic warfare, and you want to move away from US services.

I agree that those are good reasons. In my experience most people do not care about the simpler interface and feature set. They will put extra time into working with AWS or hiring an AWS expert rather than use something simpler. One reason is that non-technical management think something with a web UI is always simpler than doing things on the command line.

My favourite AWS service is Lightsail because it lets me ignore the rest of it and keep things simple.

People do care about the other two reasons, but only if there is an immediate risk. Most people only care about keeping sensitive data within their country if the law requires it, so that would only be a problem if you are in a country which has such laws, they apply to you with no loopholes, enforces them, and no AWS servers. Smaller companies will not plan for economic warfare unless it is an immediate risk - i.e. sanctions (there are only a few countries that are completely sanctioned).

The article is about Nigeria and AWS has no presence in Nigeria (and very little presence elsewhere in Africa) so there is probably more demand for alternatives there. In most of the world they do.

> She would pay AWS in dollars even as Okra earned in naira because American cloud providers did not accept payments in local currency at the time.

> The payments to AWS gradually skyrocketed as the naira depreciated around 70% against the dollar between 2020 and 2024.

> Combine that with Nigeria’s economic challenges — rising inflation and forex volatility

> Ashiru took things into her own hands, and Okra set up cloud infrastructure with servers in data centers in Nigeria and South Africa in 2024

But the servers will also be priced in a more stable currency than the naira, so I'm not sure what the win is. I also googled "nigeria electric grid" to learn if they'd see how much pricing would be driven by inflation, but before I got there, saw that the national grid had 12 collapses in 2024. Generators and fuel are also priced in more stable currencies.

The data sovereignty argument is interesting. I know there are EU companies that only use EU data centers. We might see (if it's not already a thing) AWS have a nominally independent EU subsidiary that can ignore US regulators.

12 collapses seems incredibly generous, I don't think I ever had fewer than five blackouts in a day for the few months that I stayed there. Maybe it's better for data centers than hotels but I really don't believe that there's a big difference.

Maybe 'the national grid' versus lots of local substations, i.e. all hotels might see 5 a day, but not necessarily at the same times.

The AWS bill comes once a month whereas you buy servers maybe once a year or w/e and then you own them. This alone means that buying servers hedges you against inflation at least to some degree. If inflation is around 20% a year, those savings could be considerable.

The data sovereignty argument is likely more than just interesting, but can be essential to their startups surviving in the future. Who can look at and do what with the data, can be the difference between a country or region keeping a thriving infrastructure or having to outsource an entire related chain to somewhere else that can dictate terms or not be in their best interest.

Just as EU companies have the option of using only EU data centers, African countries can make the same argument for equivalent options, where AU companies (African Union) do the same.

That last part has been announced, but not released yet. Amazon calls is AWS European Sovereign Cloud, see here for details: https://aws.amazon.com/blogs/security/aws-plans-to-invest-e7...

>> She would pay AWS in dollars even as Okra earned in naira because American cloud providers did not accept payments in local currency at the time. The payments to AWS gradually skyrocketed as the naira depreciated around 70% against the dollar between 2020 and 2024.

So this is really a forex hedging problem.

Although there's an interesting twist with the local colo et al. competition who presumably have capital costs still denominated in foreign currencies (e.g. servers, generators, etc).

What should a foreign-currency-denominated cloud pricing do, when a customer's local currency depreciates?

It should be really not that hard to price in local denominations. Netflix charges different subscription costs in different countries and they have their infra built on other cloud providers.

Cost of hardware is constant. And the generator, fuel, local electricity, wages etc. CAN be denominated in local currency if the datacenter is local. Obviously if you’re using a server in the US region while sitting in Nigeria, you have no other discourse than pay USD, but for other regions, you should be able to charge in local currencies.

Ultimately, pricing is what the market will bear.

So if AWS prices in naira, and the naira loses 70% against the dollar, does AWS change their pricing? Or just take a 70% haircut on their Nigerian revenue?

I get local businesses want cost stability, but it seems like inflation will happen both ways: it'll just take longer to cascade through the supply chain with local-owned, local-priced providers.

The operator could hold a large stock of Amazon gift cards.

The premise of the article seems a bit flawed: Nigerians are not replicating hyperscalers, but more like building local Hetzners. And that is a fine idea!

A big problem for people doing this, however, is that peering is pretty much nonexistent in Nigeria (and, for that matter, most of Africa). So, traffic from, say, Airtel (where a lot of consumers are) to Globacom (which hosts a lot of major businesses), will not stay on the continent, but go via London or Marseille instead. And, also worth keeping in mind, from Lagos to those destinations, Joburg or Cape Town are actually double the distance, even though they might 'sound' closer.

So, yeah, I wish everyone involved all the best, but it will be an uphill battle. Convenience and latency make 'big tech' pretty hard to avoid, and 'strong crypto' would be my bet over 'local facilities', but, yeah...

Indeed Africa has some of the highest wholesale internet prices in the world.

Having said that, there's a couple of new subsea fiber optic cables going live in Africa imminently. I would expect wholesale prices to drop substantially.

Yes that's crazy. I didn't know there was serious lack of peering in Nigeria and other countries on the continent.

In South Africa however, peering is excellent and has been for decades. I also love using services hosted in-country. I pinged a service now and got 4ms RTT, try that with a London based server.

> from Lagos to those destinations, Joburg or Cape Town are actually double the distance

As the crow flies, Lagos is slightly closer to Cape Town than London, though Marseille is indeed closer than virtually any part of South Africa (though not by much).

> more like building local Hetzners

So Hetzner is a term now lol. Soon we may also have the verb 'hetznering'...

"Have you people hetznered your datacenter?" "Why, yes we have!"...

I didn't realize we at Hetzner had become a verb. That's a kinda cool thing to find out on a Monday. :D --Katie

How fast are cost per unit compute of state-of-the-art datacenters going down? There is the construction cost and the running cost.

When does it become feasible for poorer countries to start building them?

Is there a reliable modular datacenter construction technique that allows for easy scaling/expansion?

Seems that could have a dramatic impact on bringing these within reach.

Racks.

Yeah, I was thinking something a little more balanced than building an AWS-scale datacenter with full infrastructure—capable of accommodating thousands of racks—then starting with ten 42Us.

Sun had their 'modular data center' concept[1], which was a self contained DC in a standard shipping container. You could chain them together. It wasn't obscenely expensive relatively speaking as I recall. I'm not sure if Sun was 'first', but these days there are all sorts of companies that do the same thing (I was recently in one from Delta). I don't know if they are applicable to the reality on the ground in Nigeria.

[1] https://en.wikipedia.org/wiki/Sun_Modular_Datacenter

Very interesting. Yes, this is along the lines of what I was thinking.

Seems you could calibrate the size of each unit according to anticipated growth and the max a site could sustain, area-wise (i.e. go larger to balance number of containers). Then, prep the site with easily extensible cabling for power/networking between units.

It seems (completely intuiting here) that if the reality on the ground could accommodate full data centers, then this would be possible.

And according to the article, the cost was 1% of traditional data centers.

I don't think there was much choice when it came to the size of the unit. It was a shipping container...the size is standardized. So more "install the first one and start filling it. When you get to some percentage of maxing it out, start budgeting for the next one."

the cost was 1% of traditional data centers

That smells an awful lot like an apples-to-oranges marketspeak. Perhaps "1% the cost for 1% the capacity". Ignoring the things you get when you build out a full DC that these don't have, I very much doubt if you bought, say, 300k sq ft of these it would be 1%.

>I don't think there was much choice when it came to the size of the unit

Yeah, I got the sense the design was intended for smaller on-demand needs versus strategically building out massive datacenters. So, my thinking is that you would borrow from the base idea, but you'd upsize each unit to cut down on the actual number of units (and attendant overhead). Just considering that if you truly are building out a datacenter, you'd probably do better to have fewer, larger units.

On the flip side though, the shipping container format does make it very appealing logistically.

>That smells an awful lot like an apples-to-oranges marketspeak

Yeah, they are no doubt being selective in what they are comparing. That claim has to do with the cost of "connecting [these units] to existing infrastructure".

So, I'm guessing they are assuming infrastructure, then comparing the cost of one of these units to construction of equivalent DC square footage.

1% still seems aggressive, but I could be convinced it's "significantly less". Of course, assuming the infrastructure starts to get murky, since a full DC build is going to be more intertwined with its infra.

I got the sense the design was intended for smaller on-demand needs versus strategically building out massive datacenters

This is absolutely the case. This was not intended to replace small, permanent DCs, much less massive ones. Think "I need compute on space limited offshore oil rig" or "I need some compute here for 6 months".

you'd upsize each unit to cut down on the actual number of units

I've been out of involvement in building out DCs for a long time, so my info is dated, but...

As I said, there are now other companies doing variations on this. The Delta unit I mentioned upthread is built into a (standard?) 18-wheel tractor trailer (significantly larger than a shipping container). But much bigger than that you start getting into "how do I move it to the site" issues, which means costs due to transport are going to go up fast, and you're back to "does it just make sense to build a fixed DC", perhaps with a careful eye towards "I'll want to easily tack on more capacity later". Even if you figure out a way to do off-site pre-fab economically, I don't see how it would make sense to do "DC built out of X many identical prefab units" but instead you'd want to do "X many rack units, and Y many network units, and Z many air handling units, and W power distribution units, and V many CnC units, and...". I know there's a lot of talk about how building houses out of prefab components is going to change the world (one day), but unless you are building a lot of data centers, I don't know how you build out the infrastructure to deliver on that that's going to be cheaper than "haul materials to the site and build it there like we've done forever". My completely unsubstantiated view is the volume isn't there.

Very cool. Huwawai also made a similar thing [1], called the FusionModule1000A All-in-One Container Data Center. But interestingly (as far as I can tell) without any compute, just power, cooling, safety etc. Presumably costs about a million USD [2].

[1] https://web.archive.org/web/20170709195517/https://e.huawei....

[2] https://www.finditprice.com/index/it_product/search.html?pro...

I don't think Sun's base model came with any compute either. It was racks/power/cooling and then you put servers and networking in. Sun would happily sell those to you...just more SKUs on the invoice.

As others mentioned “everyone” has done variations of havac-and-racks-in-teu for the past 25 years. Heres Amazons version circa 2010, https://www.datacenterdynamics.com/en/news/amazon-sheds-ligh... You find them in old satellite photos of places like us-west-2. You still need site prep to support the weight, you still need power, external CRAC/H and/or water supply, and access & repairability sucks.

There is no link in the article and Nebula is a very common cloud hosting name, but must be this one: https://usenebula.io/ https://usenebula.io/pricing

Does not seem cheap though by the first of it.

They sell an alternative to AWS, but their website is hosted at AWS.

Also, pointing the MX record to smtp.google.com will not help with receiving email.

i long for the day where african tech is just tech and not some novelty ooo look at what they are doing with less

happy to hear the prince is wisely investing his money after decades of email donations

> Besides the option to pay in naira, these companies allow Nigerians to store their data within the country — an advantage most of their Western rivals lack. Local servers give businesses the benefits of low latency and data localization at a time when the debate about who has access to a country’s data is heating up

Uhhh... all major cloud providers let you locate data manually? AWS surely has Nigerian datacenters too, that's not an advantage, that's the minimum.

I will grant the argument that random nigerian startups are not subject to foreign government interference, but saying that low latency is a benefit seems incorrect.

> AWS surely has Nigerian datacenters too

AWS's only African presence is in Cape Town.

Good for them if they can keep it up. Keeping your data on foreign servers is idiotic even if the cost is acceptable. Doubly so if it isn't.

Amazon, which was originally planned to be named "Ruthless", will see this as an existential threat and use the full force of their Big Tech allies to make sure they have no additional competition. I support their cloud services and know that Big Tech collusion on cloud infrastructure will be nearly impossible to avoid. There must be more decentralized solutions and I'm proud of Nigeria for innovating when Silicon Valley is more worried about consolidating.

Err. I think you meant Amazon planned to be named "Relentless" (and then Cadabra).

Bezos did register "Relentless.com" and it redirects to amazon.com to this day.

Other early name alternatives were: Awake.com (still works), Browse.com (still works) and Bookmall.com (doesn't work).

Pro Amazon Insider Tip: If you visit each of the four original URLs four times in a row, and then go to Amazon.com, all items are priced at $0.00!

Just kidding.

Reply on news.ycombinator.com