Issue #22: The Cloud

Five Computers

There is no direct evidence that Thomas Watson, Sr. predicted in 1943 that there would be “a world market for five computers”. First, he probably didn’t know about computers back then: in the 1940s computers were classified research projects at the UK Government Code and Cipher School and the US Army, while Watson’s IBM were famously involved with the other side during the war. Second, it’s just not something there’s evidence of him talking about. When IBM presented plans for their early commercial computers to various businesses, they got about ten immediate sign-ups and as many expressions of interest. That’s already a market for twenty computers. Various people in government on both sides of the pond, including the UK National Physical Laboratory’s head Sir Charles Darwin (grandson of the one you’ve heard of), may have said that their respective government’s laboratory needs could be satisfied by a small number of computers, but these seem to be estimates of immediate resource usage rather than predictions of future market growth.

Like the 640k RAM thing attributed to Bill Gates, the five computers quote seems to be a fabrication from a later age, invented to teach us a lesson about the futility of grand predictions. It’s a shame, though, because the five computers prediction looks like it may come true very soon.

Through a combination of adoption of virtualisation technologies (theoretically improving the portability of software deployments) and consolidation onto a small number of providers (practically ensuring that no such portability occurs), three of the five computers can easily be identified. Amazon, Alphabet’s Google and Microsoft each run one of the world’s three largest commercial computer-time rental businesses. Their three CFOs each centrally command an economy worth in the region of $1T (particular quotes taken from Yahoo! Finance on 14th June valued Amazon at $1.27T, Alphabet at $0.96T, and Microsoft at $1.42T), though as we shall soon see it’s probably best to see them as department heads of a single multi-trillion global computer economy encompassing the three businesses and more. To put that into comparison with other planned economies, the Soviet Union reached a GDP of $1T in 1979 and $2T before its dissolution a little over a decade later.

So which computers are fourth and fifth in the global market? Certainly none of the other cloud providers, who are all tiny in comparison. The closest runner is Alibaba ($0.58T), but forget the other western runners, Oracle ($0.16T) and IBM ($0.11T). They can both be seen as attempts to extract rental income from their existing hardware platforms – SPARC and RS6000/Power, respectively.

And talking of RISC CPU designs leads us to disregard another wannabe competitor who’d definitely like to be in the big five, but seems destined to come crashing back down to the ground. As excitement mounts over whether certain high-profile computer vendors may switch to using the ARM CPU design in their systems, ARM sales and revenue are down and parent company Softbank is suffering losses, too (both directly, and its $100B Vision Development Fund). ARM’s heritage is in desktop and mobile computing (the chip was designed to run the Acorn Archimedes line of personal computers, and when ARM was spun out of Acorn they received funding from Apple to develop chips for the Newton line of handheld computers). This is clearly a market that current CEO, Simon Segars, sees as legacy business, the cash cow that he can squeeze to fund his real interest in the Internet of Things. Thus ARM’s new goal, to enable “a trillion connected devices”.

You could see Softbank’s pattern of investments as an attempt to create that fourth computer in the market of five, the distributed computer of things. Whether you’re in the office (Slack, We), at home (Grofers, FirstCry, OpenDoor), or inbetween (Cambridge Mobile Telematics, Mapbox, GetAround), Softbank are hoping that you’re connected to Masa’s computer. Their particular choices, and scattergun approach, mean that this is unlikely to coalesce within the funds and time available to Softbank. It’s unlikely that they’d take ARM down with them before someone stepped in to buy it out. Which brings us on to the real contender for the world’s fourth computer:

Apple. The Apple cloud is a computer that contains a lot of music, TV shows, movies, games, books, news reports, and storage space that users of Apple’s consumer electronics can rent access on. As a cross-selling deal, a small amount of access is offered with the sale of each of the consumer devices. Unlike the other three computers we’ve seen so far (Amazon’s, Alphabet’s, and Microsoft’s), you can’t rent time directly on the Apple computer to run your own code. You can upload your own software, but you rely on people discovering and choosing to run it themselves. The execution is actually done on the end devices, not on the Apple computer, though the documents people create may be synced via the Apple computer. By the way, Apple’s planned economy currently amounts to $1.47T.

You may notice that described this way, the Apple computer sounds a lot like the Ethereum computer, and expect me to suggest that it’s the fifth computer, or maybe the bitcoin computer, but no. It’s the Facebook computer. Like the four previously mentioned, the Facebook computer is a huge distributed computer with billions of users, running lots of different applications. Like the Apple computer, you can’t just run your own code on the Facebook computer (well you can run games), but that doesn’t matter, because most people use computers for the same reasons. The big feature of the Facebook computer is the contacts app, which has about a third of the world’s population on.

Those are the five computers of the world market, together directly controlling $5.8T of the world’s assets, roughly 2/3 the value of the total amount of gold in the world. Indirectly, their influence through app stores, subscription fees, software providers who are dependent on their services, resellers and so on is significantly larger.

And it’s important to consider them together, because the veneer of competing in an open market only goes so deep. Look at any relevant trade association and you’ll see the trillion-dollar computer vendors working together toward their common interests, for example the Cloud Native Computing Foundation (Alibaba, Amazon, Apple, Softbank ARM, Google, IBM, Microsoft, Oracle, others), or the Linux Foundation (Fujitsu, Google, IBM, Microsoft, Oracle, Alibaba, Facebook, Amazon, Softbank ARM, others). There is really one central computing economy, controlling well over six trillion dollars of technology investment, with five different public faces.

How did the market for computers get consolidated to this extent? It certainly wasn’t inevitable. Early examples of outsourced computing were advantageous because the hardware was so expensive, needed specialist and frequent handling, and there just wasn’t much of it about. Those early IBM systems that Watson found more than five customers for could be rented for around $12k-18k per month, so you had to be sure that your computing job was going to save a few people’s salaries before you even ran it in the cloud.

Where companies did invest in computing hardware, offering the platform as a service to others help to offset the purchase and development costs. The first commercial computer in the UK was the Lyons Electronic Office I, developed from the EDSAC machine by food producer Lyons. LEO/I also ran payroll for Ford UK, and weather simulations for the Met Office (until they bought their own Ferranti computer). It was in use between the 1950s and the 1980s, demonstrating great longevity in the days before planned obsolence.

During those decades, the smaller size of components, greater reliability, lower costs, improved usability and application combined with the business vision of key people who set out to provide “a computer on every desktop”. Rather than a small number of computers becoming more capable, capabilities became more widespread, more distributed, more affordable, and more accessible. Computers shrank in size and cost until every office building, every cubicle, every home, eventually every pocket could contain one.

None of the technical, physical or practical trends described above have reversed to lead to the cloud computing consolidation we see today, and the world market for five computers. The only change that can account for this behaviour is the business vision, and the direction in which the cadre of executives take their companies.

Since the dot com crash of 1999-2001, investors have looked to seed companies with the smallest amount of money possible to lead to either a rapid collapse or rapid growth supported by multiple follow-on investment rounds until an eventual exit. Hockey-stick charts, “disruption”, “MVPs”, “growth hacking”, “fail fast” are not neutral business advice, but a particular way of working designed to minimise risk and maximise payoff for a small clique of rich investors who expect the money they lend out to work its way back into the system – perhaps through cloud subscription fees paid to companies they hold positions in. Don’t go buying expensive capital investments like offices, desks or computers when you can rent them…from us.

This has led to the situation where computing applications are architected with the goal of rapidly meeting the goal of tomorrow’s investor presentation, regardless of the effect this has on the later costs or usability of the product. One workload I now support would cost an estimated £6000 per month to run on one of the large cloud providers, which would allow me to purchase the actual hardware in use four times over every year. But “spinning up an instance” today using free credits means you can worry about that cost later, whereas ordering a box from Dell means waiting two to three business days by which time the hackathon is over. Besides, you’ll hit your Series A round before you get that far, and can hand the money straight back to the cloud providers taking a few percent for foosball tables and sodas for your team of devops engineers.

Make no mistake, the VC money in software is a small raindrop let loose from the cloud, with the intention of finding the next valuable idea to absorb. A tiny morsel of this multi-trillion-dollar planned economy is sent out to see where it sticks, and what new ideas the computing Gosplan department should factor into their forecasts.

It’s actually pretty easy to stay out of this system, by designing software applications that don’t depend on it. Treat your users as your customers, and give them choices: do they want to use the facebook computer, or their own computer? Do they want their files in the Google computer, or the Apple computer, or maybe on a Nextcloud or in Dropbox? Does your processing have to be done on the Amazon computer, or could you buy and house your own machine, or colocate it, or use a virtual private server? Do the interactions of your customers with your software need to become part of the global technology planned economy, or can they be part of an online commons like Wikipedia, OpenStreetMap, or the Internet Archive?

Photo by Derek Thomson on Unsplash.

Graham is a senior Research Software Engineer at Oxford University. He got hooked on making quality software in front of a NeXT TurboStation Color, and still has a lot to learn.