🏗 Building backhoes
TL; DR – Backhoes dig trenches, which host cables, which power the internet. So internet speed + reliability can only improve as fast as backhoes can improve. Which is slowly and linearly.
For the cutting edge of tech to make our lives better, we need to give as much attention and energy to the backhoes: those things out of the limelight, but enabling everything else.
So, I was complaining recently to a friend about my wifi here in Rwanda.
It’s not that there’s no wifi. Along with an impressive 95% of the country, I can access an internet connection.
It delivers speeds of around 10 mbps, max. Not terrible, not great: just average. Enough for a small Zoom call (no video). And it’s expensive. Sometimes, it cuts off, like when it rains. Or when there’s a power cut. Or just because.
Nothing seems quite so out of step with the modern world as bad (or average) wifi. We talk, we stream, we browse, we work, we transact, we build over the internet. On the horizon, Web3/blockchain, the Metaverse, Internet of Things, Artificial Intelligence, etc are all predicated on fast, cheap, and widespread internet.
But globally, the story of internet speeds they rely on is the story of steady, linear growth.
Here’s internet speed over roughly a 9 year period (2008-17) in the UK (my home before Rwanda), and the US.
Linear growth isn’t nothing. It’s still a lot; just the line is straight. As you can see, both countries have increased internet speeds by 4-5x in 9 years. Although my hunch is, the average is being skewed by a few 100mbps+ connections at the top. As with many things, inequality skews the average up even as things stay the same for the majority.
Let’s compare this linear growth to everything else that’s happened since 2008. Back then, I was 16. I still texted (WhatsApp was a good 3 years away). Smartphones were scarce, nobody I knew had ever bought anything online, and the bitcoin whitepaper was just about to come out. Those technologies I talked about earlier – AI, blockchain, Metaverse, etc – were more like sci-fi concepts. Since 2008, the world of tech seems to have turbo-charged ahead exponentially, while internet speed has improved linearly.
That’s true in the UK and US. It’s also true in Rwanda. After a massive increase between 2012-15 (from a standing start), things have plateaued. From my experience, we’re not much higher than 10mbps here today. A connectivity hungry future is coming at us, and our internet infrastructure can barely keep up.
Why?
One of the best explanations I’ve seen is in Bill Gurley’s challenge to Moore’s Law. If you haven’t, it’s worth reading the whole, 5-min post.
Gurley is one of Silicon Valley’s all-time great venture capitalists. Looking at the emergent phenomenon of the internet in 1996, he saw it as the convergence of two developments: computing power and telecommunications.
Famously, computing power follows Moore’s Law: it doubles every 18 months.
Conversely, Gurley points out that telecommunications does not follow Moore’s Law. That’s because telecoms are dependent on millions of metres of cabling. Which is dependent on millions of metres of trenches. To dig trenches, you need diggers or “backhoes”. Backhoes can only increase their productivity by increments. Which means the telecoms sector can only improve in increments.
So the internet, which is dependent on both telecoms and computing power, can only improve in increments. In Gurley’s words: “the computer industry is dependent on the telecom industry which is dependent on backhoes which don’t obey Moore’s Law”.
That’s why, while internet speeds are improving, its improvement is still linear. Because backhoes only improve linearly.
More broadly, Gurley challenges us to ask: what are we dependent on? What is that dependent on? And what is that dependent on? When you get to the end, ask: at what rate can that thing improve? Whatever that is, then becomes the rate of improvement for whatever is further up the chain.
A lot of the time, the internet is reliant on things that improve slowly, inconsistently, or are unevenly distributed. Backhoes are a great example of something that improves slowly.
The more you reflect, the more of these dependencies you discover for the internet. Stable, non-extreme weather (also becoming rarer with climate change, and more exclusive to wealthier countries). Non-censoring governments (30 of 70 countries assessed by Freedom House experienced a deterioration in internet freedom in 2020-21). Access to a steady electricity supply (also improving very, very slowly and not at all in many countries).
These things are rooted in infrastructure and geography. I was telling a friend about Gurley’s take on the internet. She’s lived on several small islands, where the internet was dependent on “one cable that runs under the sea”. When that cable had issues, the whole island suffered. Showing starkly the consequences of dependency.
Computing power and our exponential development of new technologies is out-pacing our connectivity. But ultimately, scaling things reliant on the internet will always be constrained by how fast the internet can improve, which will be reliant on backhoes, governments, weather patterns, reliable electricity, and so on. Convergence creates dependence, and can slow everything down. That’s why Facebook and Google are going down the chain, investing billions of US$ in infrastructure and cabling.
Let’s apply this “chain of dependencies” principle to another domain. Between 1950-2006, international trade grew over 4,000% (i.e. exponentially) as more and more countries entered the global market. Then, between 2006-14, it plateaued.
International trade, it turns out, remains reliant on shipping. Which remains reliant on physical ports, ships, and containers. Which (like backhoes) can only get better slowly. So you can only trade so much, so fast. No matter how lean and efficient modern businesses now make their supply chains.
We get most excited by the front part of the chain. The part that’s visible and shiny: the app, the new tool or idea. The internet, or the other slower-moving and slower-improving systems that enable the shiny, get ignored. Until the Zoom call that felt like a communication revolution lags and dies. Until the consequences of the dependency show us the limits of the cutting edge. Then, excitement turns to frustration.
Frustration which compounds when you realise how little energy is going towards the back parts of the chain. Startups are building the next Zoom, not the next backhoe. We’re investing in Web3, with much less attention to meeting the huge energy demands that come with mining crypto. Organisations are hyping artificial intelligence, without investing in the ‘boring’ data storage and processing needed to make it work.
It’s true a lot of this foundational stuff can feel outside our control. It’s systemic, capital intensive, infrastructural. It’s mostly the preserve of governments, and even they’re mostly powerless to improve it anything other than slowly, inconsistently, and unequally.
But I’m optimistic.
Optimistic that – with frameworks like Gurley’s – we can understand the importance of messy systems and un-sexy enablers.
Optimistic that – once we understand them – we can apply humanity’s infinite creative potential where it’s needed.
Optimistic that we can build better backhoes.
🎬 Thanks to Abi Freeman, Lea Simpson, and James Wilkinson for encouraring me to think deeply when I’m frustrated. Thanks to Caryn Tan and Kym Ellis for looking at drafts of this.
🤔 Got thoughts? Don’t keep them to yourself. Email me on asad@asadrahman.io. Let’s figure this out together.
If you enjoyed this, subscribe to get pieces just like it straight to your inbox. One email, towards the middle of each month (and nothing else).
Banner depicts a typical backhoe (trench digger), manufactured by JCB. From Wikimedia Commons, the free media repository.
asadrahman.io 
