As it scaled its way to unicorn status and beyond over the last few years, Pinterest was more focused on adding features to its image-sharing social media site and ensuring its reliability than how much it was spending on cloud services — right up until the moment it realized it had a spending problem.
Pinterest’s resulting five-year $750 million deal with Amazon Web Services is just one example of how spending on cloud services is evolving beyond its roots as an ad-hoc credit-card driven process. In exchange for a more predictable level of spending at a discount to its prior rate, Pinterest happily hitched its wagon to AWS in 2017 for five years, and it’s not alone in adopting such a strategy.
Cloud computing has changed so many things about how the business world uses information technology, and one of the biggest differences is how money flows between buyers of information technology and their suppliers. Over the past year or so, the impact of those changes has materialized in several ways:
- A surge in multiyear cloud-computing deals that lock in discounted rates in exchange for minimum levels of spending.
- Companies that provide services around identifying and tracking cloud spending are enjoying a rush of interest.
- Interest in both extremely granular pricing options and monthly cloud subscription services is changing the product roadmaps at cloud providers.
And more change is around the corner as more and more businesses start to plot their cloud computing strategies after years on the sidelines.
Economic sea change
Building a technology organization was once a capital-intensive undertaking, with procurement departments placing large orders of hardware that took months to complete amid shameless nickel-and-diming on everyone’s part.
In the cloud, it’s very different. Startups don’t have to shell out millions in capital expenditures to support the computing needs of applications that have promise, or scramble to find servers when their app hits the front page of Apple’s App Store. They can try out ideas, see what sticks, and if they hit upon the magic formula they (and their venture capital backers) will be all too willing to shell out for cloud services to handle the incoming demand.
Big established companies can shift IT spending from a capital expenditure to an operating expenditure, the kind of change that warms the cold cockles of the finance department’s heart. The cloud also allows them to take advantage of the world-class infrastructure operations of companies like Amazon Web Services, Microsoft, and Google that most of those companies wouldn’t know how to duplicate, even assuming they could afford to copy that expertise.
Still, this is a generational shift in the way business and technology intersect, and it’s not without growing pains.
Part of that angst comes from the loss of centralized control of IT spending inside many companies, which makes it harder for the bean counters to predict and control technology spending. Part of it comes from frantic decision-making processes inside growing startups, where skyrocketing cloud expenses make sense when viewed against a surge in user demand, until one day they suddenly don’t.
“Businesses care about being able to model and predict accurately,” said Corey Quinn, cloud economist with The Duckbill Group, who helps companies understand the true cost of the cloud. “No one cares how much money they’re actually spending on the bills” when things are going well, they just want to avoid surprises, he said.
The early days of cloud computing pioneered by AWS offered a transactional experience not unlike buying books from Amazon.com: input credit card details, get access to cloud computing resources. This was a very empowering experience for young tech companies trying to find their footing in the wake of the Great Recession, and it also drew the attention of developers with modest budgets inside larger companies as a vehicle for skunkworks projects that would likely face pushback if pitched through the traditional process.
A running joke about this era of cloud computing is that venture capitalists are primarily responsible for turning AWS into the enterprise tech powerhouse it has now become.
And for a time, they were happy to cut checks to service the computing needs of exploding companies that valued user growth above all else. The up-front costs of building a tech business were much lower than what dot-com era investors needed to invest into promising companies, and as a bonus you didn’t have to figure out what to do with the servers when those investments failed.
The rise of open-source source also changed the economics of starting a tech company around the turn of the last decade, freeing startups from having to spend lots of money on expensive packaged software. They could manage those open-source components on cloud hardware and build far more nimble technology stacks than possible just a few years prior, which dovetailed nicely with the explosion of demand for lightweight mobile computing apps sparked by the launch of the iPhone.
But some of those startups turned into Big Businesses a lot faster than anyone could have really anticipated. As Amazon and Microsoft gear up to report earnings this week, we’ll get another data point that suggests just how much cloud spending continues to accelerate.
Pinterest is perhaps the poster child for this evolutionary process, growing like a weed on the back of AWS until one point in 2017 where it realized that its cloud spending was racing out of control. Pinterest renegotiated its deal with AWS, promising to spend $750 million with the company by 2023 in exchange for pricing discounts. That deal dramatically reduced its cost structure and helped the company show better numbers leading up to its IPO last week.
When lock-in is good
Cloud pricing starts off simple: you can rent virtual machines in the cloud by the hour, reserve a certain amount of computing capacity for an extended period of time, or, in a growing number of emerging cases, pay only for what you use.
That simplicity disappears quickly. Compute, the most basic element of the cloud, costs between $0.02 an hour and $13 an hour on AWS, depending on the amount of horsepower you need to throw at your application. Storage is relatively cheap, while moving your data out of a cloud provider’s network tends to be expensive.
Higher-level services, such as managed databases or Kubernetes, generally give cloud providers better margins than basic services like compute and storage. After years of pricing cuts, cloud providers are competing less these days on raw costs and more on differentiated managed services that handle more complex parts of their customers’ technology infrastructure.
The industry is also moving toward more granular units of consumption. Companies can now rent the flagship EC2 compute instances by the second (the minimum charge is a minute), and the AWS Lambda serverless service takes this concept even farther. Serverless pricing is a fraction of regular compute pricing, although serverless applications can also generate a lot more billable activity than you might first think.
But the thinking around cloud buying is starting to change as more and more large companies sign multiyear cloud computing contracts in hopes of pricing discounts and predictable budgeting. You’re still paying for what you use, of course, because the cloud companies bear the costs of acquiring and maintaining the equipment, but multiyear deals offer security and the chance to feel like you’re getting a good deal.
A three-year contract signed with AWS by South African connected car-tech provider MiX Telematics in 2017 (which is somehow available on the internet, unlike most such contracts) provides a template for how AWS strikes these deals, which require a spending commitment of $1 million a year. In exchange for a commitment to spend $1 million through 2020, AWS agreed to give MiX Telematics a 9 percent discount across a host of key AWS services, including EC2 and the DynamoDB database.
Microsoft has a long history as an enterprise tech provider, and also works off standard agreements drawn up in the days when companies would buy thousands of Windows PCs and run their databases on SQL Server. Azure is just one part of those agreements, although Microsoft is known to give existing customers a discount on Azure services if they are moving off Windows Server licenses to its cloud, said Rob Helm, an analyst with Directions on Microsoft.
This year I had AWS contract discussions with multiple Fortune 500s looking at *annual* spend commit of > $500m; many discussions with enterprises negotiating $100m+ annual commits. By contrast, gradual growth in Azure contracts Gartner's seen for more than $10m+/year. #reinvent
— Lydia Leong (@cloudpundit) November 28, 2018
Given all the existing relationships for products like Office 365 or Windows PCs that Microsoft salespeople have with a large swath of the enterprise computing market, it can also use pricing incentives on those products to entice companies to choose Azure for cloud computing, Helm said.
“There’s kind of a classic sales incentive operation going on, where a salesperson has a significant incentive to reel in a lot of Azure business, so that salesperson can discount a lot of Windows or Office software in order to reel in that Azure deal,” he said.
Trust, but verify
But there are still lots of small and medium-size companies that defer decisions on cloud spending to individual departments within the company, rather than routing everything through a centralized purchasing organization as was the case in the old days.
That gives those departments a lot of autonomy to make quick decisions about computing resources, in the hope that internal bureaucracy won’t derail promising ideas. It also tends to result in situations where abandoned projects continue to run quietly in the background on cloud providers racking up small charges that add up over time.
Dozens of vendors have emerged over the past decade that promise to help companies understand exactly how much they are spending on cloud services, including Pacific Northwest companies like Bellevue-based Apptio, acquired for $1.9 billion last year; and Portland’s Cloudability.
“For most companies, cloud is still a small line item in the larger IT budget,” said Cloudability co-founder J.R. Storment. But that is changing quickly, and companies that designed their finance tracking systems around capital expenditures for hardware need help balancing the desire to help individual teams move quickly with the need to manage how much money they’re spending, he said.
They also need help understanding the true cost of cloud computing. The cloud is not cheap, as the surge of revenue booked by AWS can prove, and it requires a different way of thinking about software development when making decisions about how to spend time and money.
For example, a recent Cloudability customer was faced with a decision about an older application that had moved into the cloud. The company could spend six months rebuilding the application in a cloud-native way to operate more efficiently, or spend more money on cloud processing power to make it run reliability.
“They made the decision to throw money at the problem,” Storment said.
Google is behind AWS and Microsoft when it comes to cloud market share, but it made an interesting pricing move this month at its Google Cloud Next conference in San Francisco.
Anthos is a hybrid-cloud services platform introduced at the event that is designed to give cloud holdouts an easier way to balance their existing investments in on-premises hardware with newer applications run on Google’s cloud or even competitive public clouds. It’s basically a set of training wheels for companies new to the cloud, but it also will give companies the option of subscribing to the cloud on the monthly basis.
Google will offer Anthos customers one-year or three-year subscriptions, and it will also allow companies to go month-to-month after a year if they prefer, said Eyal Manor, vice president of engineering and product management at Google Cloud, in an interview at the Google Cloud Next event earlier this month. Google isn’t ready to announce specific pricing details around Anthos, but there will be usage tiers within those subscription plans similar to how G Suite account subscriptions offer different amounts of storage.
Customers interested in Anthos “wanted the predictability,” Manor said. Also, “they don’t want to bundle themselves 14 different cloud services.”
This arrangement might not work for every cloud customer, but it seems to make sense for a certain class of customer that is less concerned with scalable performance and more concerned about stability, which describes a solid chunk of the Fortune 500.
The rising tide
These shifts in pricing strategies reflect a maturing cloud, and it was probably inevitable that Big Business would shape cloud computing in its image to some extent.
AWS and Microsoft have a big edge in this area thanks to their first-mover advantage and deep enterprise sales roots, respectively. New Google Cloud CEO Thomas Kurian has promised to meet big business customers where they are on their terms, which hasn’t always been easy to accomplish inside a company run by the engineers.
Cloud computing is already an enormous business, and given that around the majority of current workloads remain in traditional data centers, an incredible business opportunity awaits the companies that figure out the best way to help customers consume their services. Expect to see a lot more experimentation in cloud pricing as more enterprise customers take the plunge and technologies like serverless computing become mainstream.
“No one’s spending less,” Quinn said. “One year after I do an engagement, I don’t have any customers that are spending less than they were when I started. The difference is now they understand it better and are able to predict it.”