:format(webp)/cdn.vox-cdn.com/uploads/chorus_image/image/63706843/amazon-web-services-data-center.0.1537225630.0.png)
Remember how a few months ago there was talk of a “price war”in the world of cloud computing services? It’s over and all the signs point to Amazon having won.
As we reported last week, Amazon posted a stunning 81 percent rise in revenue at its Amazon Web Services cloud computing unit, along with a five-fold surge in the unit’s operating income, giving it an operating margin of 21 percent.
During the quarter Amazon said it cut prices on many of its cloud services for the 49th time since the service launched in 2006.
Meanwhile, at least one of Amazon’s rivals — Microsoft’s cloud computing service Azure — raised some of its prices in Europe and Australia. Portions of emails to Azure customers published by the Dublin-based blogger Aidan Finn earlier this month detailed price increases of 11 percent in Europe and up to 26 percent in Australia.
The ostensible reason for the pricing hikes? Currency concerns. With local currencies like the euro and Australian dollar weakening against the U.S. dollar, U.S.-based companies are starting to respond with price increases in affected markets rather than absorb the revenue hit resulting from exchange rates. For its part Google cut its prices on some services by as much as 30 percent in May, sticking to a promise to do so regularly in line with Moore’s Law.
So how does the huge spike in AWS revenue square will falling prices? For one thing there are more companies using AWS than before, though we don’t know how many. On a conference call Thursday, CFO Brian Olsavsky said that “we are seeing continued increases in usage” without being specific. Later he said: “Pricing is certainly a factor [but] we don’t believe it is always the primary factor.”
The last time Amazon gave a ballpark estimate for how many customers it had on AWS was late last year when it said it had more than one million active customers. It hasn’t updated that figure since then. Whatever the number, it suggests that the average revenue per customer is on the rise.
There are reasons for that. One is how the raw scale of Amazon’s computing footprint is growing. In a recent interview with Re/code, Andy Jassy, the Amazon executive who runs AWS, shared a mind-boggling metric. While it won’t surprise you to learn that AWS adds new capacity every single day, it may surprise you to learn how much.
It turns out AWS is adding enough capacity to serve a mini-Amazon every single day. “If you look at how much new server capacity AWS is deploying every single day, it would be enough to have handled all of Amazon as a global company when it was a $7 billion enterprise,” he said. That corresponds roughly with 2004, when Amazon reported revenue of about $6.9 billion. So, think back about to the Amazon of about a decade ago and AWS is adding enough capacity to serve its entire global IT needs — every single day. “And we’re not even close to being done building,” Jassy said.
That appeals to AWS customers — existing ones and new ones — because it gives the impression of nearly limitless capacity. Whatever computing resources you need can be turned on it minutes, not weeks. And the impressive reserve of spare AWS computing power that isn’t rented rarely sits idle. For example, it operates a spot market where customers can bid on renting a few hours or days of compute capacity according to what they’re willing to pay per hour of use. Spot prices are set in real time according to supply and demand, and the bidder understands that the job can be interrupted when the going rate price exceeds what they’re willing to pay. This helps Amazon make money on computing capacity that would otherwise go unused.
Even so it’s reasonable to ponder: If Amazon’s margins are increasing it implies that its operational costs are coming down at the same time that its customer count is going up. Logically speaking, that implies that quality of service might suffer.
There’s a few reasons that it doesn’t, and it has to do with how Amazon’s costs come down as it scales up. As Amazon’s cloud footprint grows, its fixed costs per customer on things like commodity memory chips and electricity decrease. The more chips and power it buys, the more negotiating leverage it has to squeeze a good deal from suppliers. Meanwhile, business and administrative costs associated with keeping the system running — billing and personnel, for example — shrink as more customers sign on and processes become more efficient.
Finally, there’s more that those customers can do with AWS all the time. Amazon has added 350 individual new features to AWS in the first half of the year. Two new services include one called AWS Device Farm which allows mobile app developers to run their software on simulated Android mobile phones. At its current pace it will by sometime this fall have added about 700. And the pace at which those features are being added is increasing too. Last year it added 516 new features. The year before that, it was 280, and the year before that 159. Next year, who knows?
This article originally appeared on Recode.net.
