For hundreds of startups in the as-a-service world, the scores of product launches at last week's Amazon Web Services Reinvent Conference each were a warning shot across their bows. We are coming. We are coming right after you, with tens of billions of dollars on our balance sheet, hundreds of salespeople, and the broadest suite of software and infrastructure since Oracle. Anything that's open source with traction, we will host. Any business where we see margin is our opportunity. We are coming fast and hard. At least, that's the way I interpreted it.
Amazon isn't the only monocloud. I heard this term from a founder recently. I couldn't tell if it meant a singular cloud that could satisfy all a customer's needs; or monopolistic cloud that will consume all. Perhaps it's a clever portmanteau because both meetings are equally apt.
But getting back to the point, Amazon isn't the only monocloud. Google, Microsoft, and Salesforce each are building broad platforms that serve hundreds of thousands of developers and are generating tens billions of dollars in aggregate revenue and growing at incredible rates. All of these businesses have some pretty powerful advantages they are using to compete.
First, the companies behind them have bolstered these businesses with billions of dollars from their balance sheets.
Second, these dollars finance hundreds if not thousands of engineers are working on developing new products at a torrid rate. The list of new product announcements from Amazon Reinvent underscores that point.
Third, because of the freedom extended by the two most popular open source licenses these infrastructure vendors can host successful open source projects without constraint. And because of their scale, they can do it at a fraction of the cost by the startup would.
Fourth, all of these companies are building substantial sales teams to sell through their existing channels and also develop new customer relationships. Within many large customers, these infrastructure providers already have substantial billing relationships and product penetration.
Fifth, there are very real advantages to using multiple services from a single platform vendor. Integration is easier. This simplifies the task of an account executive upselling or cross-selling.
From their infancy through their adolescents, these monoclouds played primarily in the infrastructure world: databases, hosting, compute, storage. But now they're moving into applications. Microsoft's PowerBI and Amazon’s Quicksight are forays into business intelligence. Microsoft announced the launch of Jupyter notebooks this morning, a push to supply an application to the millions of data scientists. AWS listed business productivity category on their product page which includes collaboration tool to compete with Box and a hosted email product to compete with Gmail and Outlook. Amazon released a mobile analytics product. I could go on.
In addition, all of these monoclouds invest aggressively in commercializing research. This oligopoly on machine learning talent releases advances faster than any start up could. A computer achieves human parity in speech comprehension. Speech recognition model three times faster than typing. Photography apps that recognize locations and objects within their own photographs, or can trace people through their lives as they age. Plus with advances like Google’s TensorFlow Processing Unit, custom silicon for processing machine learning models, the incumbents’ advantage shows no signs of waning.
What is this all mean for startups? Building a web application that’s a view on a database, the characteristic features of a first-generation SaaS company, won't cut it anymore. At some point, these monoclouds will build or buy a competitive version and cross sell their way into the market. Now that everyone has access to similar speech recognition, speech generation, image recognition, anomaly detection, and other deep learning algorithms, the algorithms themselves are not much of a competitive barrier to entry either. Plus, all the technologies are going to continuously improve. For a while, startups might play the on-prem vs in the cloud security cat-and-mouse game. And they might win the battle with that strategy, but they'll lose the war. Cloud migration is an inexorable trend. Last, the “I have the biggest balance sheet strategy” won’t work the way it did when startups were the main competition. The monoclouds coffers will dwarf most others.
In short, defensibility matters more than ever. Startups will have to understand their markets better and leverage their ingenuity. They will likely have to pick a niche that the broad generalized platforms won't pursue. For example, the speech recognition algorithms provided by the monoclouds will be optimized for generic use cases, not medical, not legal, or other vernaculars. Startups will have to find new distribution strategies. They will also have to make the case that a constellation of best-of-breed point solutions trumps the ease-of-use of a suite. perhaps, they will also make use of more restrictive licenses like the GPL family of licenses. And, they must capitalize on the fear that monoclouds are now engendering in their biggest customers: lock-in.
image credit: Don Hammond