FOR THE past decade few aspects of modern life have made geeks drool more than the cloud, the cumulus of data centres dominated by three American tech giants, Amazon, Microsoft and Google, as well as Alibaba in China. In America some liken their position of impregnability to that of Detroit’s three big carmakers, Ford, General Motors and Chrysler, a century ago. During the covid-19 pandemic they have helped transform people’s lives, supporting online medical appointments, Zoom meetings and Netflix binges. They attract the brightest engineering talent. Amazon Web Services (AWS), the biggest, is now part of business folklore. So it is bordering on heresy to argue, as executives at Andreessen Horowitz, a venture-capital firm, have done recently, that the cloud threatens to become a weight around the necks of big companies.

That possibly explains the defensiveness of Andreessen Horowitz’s Martin Casado, co-author of the blog post titled “The cost of cloud: a trillion-dollar paradox”. On June 24th he described it in a gathering on Clubhouse, a social-media app, as “one of the more misread, misquoted things I’ve ever done”. At the risk of further mischaracterisation, Schumpeter would summarise it as follows. It uses paltry evidence and baffling numbers (where, for instance, does the “trillion dollars” come from?) to propose an excessively all-or-nothing business conundrum: “You’re crazy if you don’t start in the cloud; you’re crazy if you stay on it.” Yet for all its flaws, it is well-timed. It poses a question that businesses will have to think about for years to come. If they entrust all their data—the lifeblood of the digital economy—to an oligopoly of cloud providers, what control do they have over their costs?

It is a problem many companies are already grappling with. On June 29th the Information, an online tech publication, reported that Apple, maker of the iPhone, is poised to spend $300m on Google Cloud this year, a 50% increase from 2020. It is also using AWS and its own data centres to handle overflowing demand for services such as iCloud, a data-storage app. On the same day the chief operating officer of a big software firm told your columnist that the current trajectory of cloud costs is “unsustainable” but that it does not make sense just to leave the cloud. “It is very hard. One can’t be so simplistic as to say it’s all cloud for ever or it’s no cloud.” Jonathan Chaplin of New Street Research likens acquiring flexible data storage on the cloud to flexible office space such as WeWork. Both are similarly expensive, he says. He knows—his boutique firm of analysts is considering renting both.

One reason Andreessen Horowitz has stirred up a storm is because it went a step further. The blog post raises the prospect of “repatriation”, arguing that companies could save considerable sums of money by bringing back their data from the cloud to their own servers. It uses the example of Dropbox, a file-sharing firm that in 2017 said it had saved $75m in the two years before its initial public offering chiefly by clawing back workloads from the cloud. Mr Casado and his colleague, Sarah Wang, estimate that a group of 50 such publicly traded software firms could halve their cloud bills by doing the same, collectively saving $4bn a year. That could, using generous price-earnings multiples, improve their market value by around $100bn. You don’t have to be a super-sleuth to suspect an ulterior motive: if Silicon Valley unicorns take the hint, higher valuations could make venture capitalists like Andreessen Horowitz more money when they go public.

This is an oversimplification, however, in several ways. First, the cloud is not just a cost. It can also boost revenues by providing young companies with the flexibility to scale up rapidly, accelerate new product launches and expand internationally without having to build their own mishmash of racks, servers, wires and plugs. Moreover, cloud providers offer more than storage and spare capacity. Increasingly their most valuable services are data analytics, prediction and machine learning, made possible by the vast troves of data they can crunch. They may also be more difficult to hack. The question is whether a company gets a better return on its investment by paying for cloud services, or by paying to bring data centres, engineers and cyber-security in house.

Second, the supply of engineers is finite. Whereas in the past coders were trained to work with on-premise servers, the latest generation knows more about working with cloud providers. That makes repatriation harder. In a recent podcast about its decision in 2015 to shift entirely from its own servers to Google Cloud, Spotify, a music-streaming app, highlighted the opportunity costs of having engineers tied up managing its own data centres rather than working on new products. (As a geeky relic, it keeps pieces of its last big server in an urn.)

Third, profits are in the eye of the beholder. A company may hope to improve margins by reducing the cost of renting cloud servers. But building its own data centres requires investment. Labour costs will also rise to pay for engineers to manage them.

The silver lining?

There is little to suggest that the stampede into the cloud is slowing. Gartner, a data-gatherer, predicts that worldwide spending on cloud services will increase by almost a quarter this year, to more than $330bn. Repatriation is “an urban myth”, says Sid Nag, Gartner’s research vice-president. “We just don’t see it.”

Continuing to write blank cheques to cloud providers is not sustainable, either. The more firms embrace cloud-computing, the more carefully they must manage its costs. The biggest users, such as Apple, bargain for huge discounts. Smaller ones lack the clout. To keep costs down, they may need to run basic storage in house, diversify into the “multicloud” by spreading computing across several clouds, and make engineers responsible for cloud expenditures. With luck, a low-cost alternative to the biggest clouds will emerge, much as Japanese car companies challenged Detroit’s big three. That took half a century, though. ■

This article appeared in the Business section of the print edition under the headline “Raining on the parade”