The vision of Kenyan ISP Poa! Internet is “to bring internet to every home in Africa.”

Poa! offers unlimited, uncapped, 4 Mb/s fixed wireless connectivity to homes for 1,500 KSh ($11.64) per month plus a one-time installation fee of 3,500 KSh ($27.16). But wait, there’s more. They provide a dual-SSID router, and the home SSID is used by trusted family members, and the open street SSID is for others. Street users get 100 MB of free data each day and are charged 15 KSh ($.16) per GB with no expiry .data if they exceed 100 MB on a given day.

Depending upon your age, 4 Mb/s may sound exceedingly slow, but my first home Internet connection was using an ASR 33 Teletype that printed ten upper-case characters per second—about as fast as I could read. This enabled me to use email, Usenet, Telnet, FTP, etc. and, most importantly, to collaborate with remote colleagues. My first professional home computer had an 8-bit CPU, 64 KB memory, two 8-inch floppy drives, and a 300-baud modem. It enabled me to quit my day job and program, write and consult—I owned my own professional tools.

A low-cost Windows PC and a 4 Mb/s connection would allow all that I did at that time and so much more today—download software, articles, books, movies, etc., stream HD 720p video, make voice-over-IP calls, listen to podcasts, read newspapers, etc. and, importantly, create content and invent applications and services that are relevant to Africa. It would be fascinating to conduct a study of Poa! users to see what they are using the Internet for and how it is impacting their lives.

One thing we know for certain is that people do different things with a computer at home than with a phone on a mobile network or at a Wi-Fi hotspot. For example, I could not write this post on my Android phone.

This point is underscored by the experience of Poa!. Poa! CTO Mike Puchol said that when they deployed about 160 outdoor Wi-Fi hotspots in Kibera, a district of Nairobi, users consumed “very little” data despite the price being 10% of the mobile network operator’s rate. When Poa! deployed residential broadband in Kibera, average consumption hit ~140 GB/month right away and is now over 200 GB/month. (In 2021, average mobile data consumption in sub-Saharan Africa was 2.9 GB per month).

Puchol provisions 1.5 Mb/s per 4 Mb/s account, and he says there is “very little” contention. No doubt that contention grows during busy hours and days, but if they can add capacity as quickly as they add customers, they would be able to maintain that contention ratio without alienating too many customers.

They have fiber loops between their data center and points of presence in each network area they serve and use wireless links to reach individual houses. The service seems to be financially viable in urban areas like Kibera. Puchol reports that they currently have 20,000 home users, and last December, 27,000 people used the public street service. Poa! has received over $42 million in funding

Starlink

SpaceX recently began offering Starlink Internet service in Nigeria and Rwanda and several South American nations and will begin service in Kenya next quarter. Could they substitute Starlink connectivity for fiber and offer their service in rural Kenya?

Today, they could not. Today, there is only one Starlink ground station in Africa, and it is in Nigeria. The only way a Starlink user in Kenya can reach the Internet is via inter-satellite links from a visible satellite to a remote ground station. I just ran Mike Puchol’s Starlink simulation for one hour at a random location in central Kenya, and there was no available coverage six percent of the time.

Since SpaceX has committed to offering connectivity in Kenya during the second quarter of this year, I assume they will have reachable ground stations by then. The same goes for Angola, Eswatini, Mozambique, and Zambia, which will have connectivity during the second quarter, and many other African nations that are slated to connect later in 2003 and in 2004. A rural village with access to a ground station could be used for Starlink backhaul.

If there is excess constellation capacity over a village location, the marginal cost of serving a new terminal will be near zero, and an ISP can count on adding users at a predictable cost. I’ve seen a Nigerian Speedtest result showing a download rate of 238 Mb/s, an upload rate of 45 Mb/s, and a latency of 42 ms. At those speeds, a terminal could theoretically support 188 4 Mb/s customers, but that test was run on an unloaded terminal at a particular time of day.

During a day, the available capacity at a given location varies as the satellites move and the number of users online varies. (Large files should be transferred late at night).

In the long run, capacity increases as more satellites are launched, and technology improves. SpaceX just began launching second-generation Starlink satellites which provide four times the capacity of the first generation and the next generation is expected to more than double that. The number and capacity of ground stations will also continue to increase, and more satellites with inter-satellite laser links will make it possible to reach them from remote sites. We will also see optical links to some ground stations, and terminal cost and performance are also improving.

SpaceX is the only company offering low-Earth orbit broadband connectivity today, but they will have low and medium Earth orbit competitors and several of those future competitors are designing their constellations for organizations and enterprises which may make them better suited than SpaceX for backhaul from a village. For example, OneWeb will offer service-level agreements, which will take some of the uncertainty out of pricing decisions, and SES will offer software-defined beam capacities and shapes.

The one prediction I can make given all this variability is that the prices SpaceX is charging in Rwanda and Nigeria today will change over time as they have in the US and other nations.

In Kibera, Poa! provides both the fiber link to an area and the wireless links to the homes in the community, but decentralized alternatives where the community contracts with Poa! for the satellite link but installs, owns, and operates the local network should be considered. Whether operated by the local government or a private company, local community networks could coordinate to share expertise, training and purchase equipment and tools in quantity. There could be variations on this theme. For example, Poa! could offer discount purchasing as a service to community networks.

The orbit-inclinations and marketing efforts of SpaceX’s constellation were focused on affluent nations at first, but nearly all the African and South American nations will have Starlink service next year and OneWeb and SES will be serving many of them as well. Necessity is the mother of invention, and I can’t wait to see the technology, organizations, and applications invented in the global south and their social impact.

PS: The closest I can come to an English translation of poa is our slang term cool.

Update Mar 24, 2023:

Several people wrote telling me that the Poa! example was unfeasible because the Starlink customer agreement prohibits reselling connectivity and Kenya would require rural community ISPs to have international gateway licenses.

This is true in Kenya today, but SpaceX will change Starlink prices and policy as they have done in the past and will be operating in many if not all African nations. The Kenyan government may modify licensing regulations to allow a provider like Poa! to obtain a single license covering their retail customers, and other nations have different regulations.

As noted above, SpaceX’s global capacity is increasing but they will have competitors with different technologies, strengths, organizational structures, market emphases (OneWeb, Amazon Project Kuiper, and Telesat), and political interests (China SatNet). The title of a future post may be “Connecting every home in rural Africa—LEO satellite backhaul.”

Update Sep 28, 2024:

Africa Mobile Networks (AMN) signed an agreement for mobile backhaul with Starlink in 2023 and the first base station went live in Yebu Nigeria in April 2024. AMN has now rolled out more than 100 terminals to other rural villages in Nigeria.

A node can operate up to 5 simultaneous carriers in either 2G+ 3G or 2G+4G configuration and they will offer 5G later this year. AMN sites are processing more than 25,000 voice minutes daily; on average, traffic has increased by approximately 45% across all sites that migrated to Starlink backhaul.