I Think Most Analysts Underestimate Cloud Computing And Here Is Why

I love to be wrong, but I expect cloud computing revenue will exceed $200B/yr by 2020.

For the last few decades, IT has largely been about enabling employee productivity at lower costs.

Humans have a hard time of seeing new models for what they are and instead they force them into their current frame of reference. Take the canonical example where the first horseless carriages had whip holders. In reality they probably had them in case the engine broke and they were towed by a horse, but I digress and hope you get the point.

With the first wave of cloud industrialized by VMware and the 2nd wave industrialized by AWS, technology has started to take on new meaning. Now with the 3rd wave of serverless computing being pioneered by Google (working for Google gives me some inherent bias) and Azure/AWS is enabling a completely different technological modi operandi.

Instead of just helping move existing workloads to the cloud without modification (often called lift and shift) we are also seeing companies bring us their core business critical application. They do this because this serverless computing is unlocking capabilities that companies never dreamed could exist. This is increasing productivity but also enabling top of line growth for companies.

So instead of a function of shrinking operations, cloud computing spend is growing as a function of top of line growth. This is why cloud computing will exceed $200B/yr by 2020.

How do I come to this number? Simple, 200 by 2020 sounds catchy. But seriously, the economist in me loves back of the envelope calculations. I make a lot of assumptions and can be wildly wrong so I am overly conservative, like assume only 25% of infrastructure spend is on cloud by 2020 or that cloud is 2x cheaper than just hardware. In reality in some cases cloud can be almost as expensive as the raw hardware but cheaper if you include cost to run the hardware. So here goes assuming: the current IT spend is roughly 2.4T $USD according to CompTIA/IDC. Gartner also says $3.3T so I’ll take the lower number, again to be conservative. Let’s say just 20% is hardware, 20% software, 20% services and 40% other (telecom, network connectivity, support employees, etc.). That is means hardware spend is roughly $480B/yr. Let’s say that only 25% will be in the cloud by 2020. That puts us at $120B/yr. Let’s assume cloud is 2x more efficient at just the hardware level. That would put us at $60B/yr potential on just hardware assuming zero growth.

But cloud is much more than replacing hardware. It also removes the need for many other services, software and support like network connectivity.

Wait, network connectivity?! But I thought you need to buy connectivity to the cloud, what gives? Well increasingly with my customers I see cloud encompass more than just virtual machines. Many customers maintain or lease their own connectivity cables to their facilities. In one case a customer had a gorilla problem. Specifically gorillas swinging from their fiber cables caused intermittent outages (bend a fiber just right and some light escapes but doesn’t break.) This added a huge cost in addition to the leasing and/or building and maintaining the connectivity. Instead companies are opting to egress to the internet or just peer directly with one of Google’s 100+ POP or leverage our 800+ CDN nodes. Once on our backbone they don’t rely on the legacy hub and spoke model shedding 10’s of thousands per month, per site.

So let’s say that 20% of the networking is offloaded to the cloud ~ that is another $38B (20% for telecom * 20% in cloud by 2020 * 40% the cost of alternatives today of $2.4T)

Next, Software! Think changes like from Oracle to Cloud SQL or AWS’s Aurora. Or imagine traditional Data Warehouse software to BigQuery. Let’s assume they are 50% of their alternatives. Did I mention how BigQuery is awesome? So lets say again 25% of the software market is replaced by something that is 50% the cost (20% for software * 25% by 2020 * 50% the cost of $2.4T) bringing another $60B.

So without growth we are already at $60B + $38B + $60B = $158B.

Let’s assume there is just a modest 10% CAGR on top of non traditional infrastructure being replaced. That will put us at $210B in 2020 ($174B in 2018, $191B in 2019, $210B in 2020)

Remember I am vastly underestimating? I suspect more that 25% of infrastructure spend in 2020 will be for cloud. I already see much higher CAGR growth in net new workloads in the cloud (sorry, not sharing any details here). Why is this accelerated growth happening? Because these are systems that people are just starting to imagine. Think back to the whip holder on the first cars. Once people come to terms with the reliability and capability provided, then people find amazing ways to use this technology. Hmm, maybe it will look more like $300B by 2020 but I’ll have to save my law of compounding acceleration for another post.

P.S. Hit me up if you can point out a slight flaw in my logic. If you also know how to architect systems in the cloud then I’ll probably offer you a job.

P.P.S. Come hear more about the incredible transformation from companies first hand at Next, March 8–10 2017!