This article could just as easily have been headlined "How various empires will strike back." But what are the empires and who will they strike back against?
For the last few decades, the biggest players in most major industries have been sitting-out the computer and Internet revolutions as they've watched technology behemoths and startups sprouting-up around them, and then servicing them with goods and services that chomped away at their bottom lines. If the Microsoft of the 1980's taught us anything, it was the degree to which software can be cheaply manufactured and massively reproduced to drive the kind of margins that most industries dream of. As if software companies weren't already making enough profits, along came the Internet to drive down both distribution costs (think downloads vs. physical channels like retail) and marketing costs. Then, the Web and the cloud took those bits (and the profits) to another level.
Why distribute software when you can just distribute the U/X and programmatic access across networks like the Internet and the Web? Imagine how the bottom line of a software vendor starts to improve when all it takes is the click of a browser's refresh button to upgrade the entire customer base at once? (Boy, how today's "kids" stand on the shoulders of those of us who lived through tortuous installations and upgrades with 5 1/4" floppy disks). To other industries watching from the outside -- the ones like mining, manufacturing, and aerospace --- it was almost unfair.
But if the Internet, the Web, and the cloud were forces that conspired to make IT vendors richer at the expense of other industries, there's a new set of forces in play that could start to shift the tide back to "industry." If they play their cards right, the old empires that had to sit out prior revolutions could be in position to strike back.
If you're a CxO in some industry, it's best that you pay attention. Just like with tech (think "iOS and Android"), within each of the major industry verticals, there may not be room for more than one or two first movers. If you're not a CxO, make sure this article gets to a CxO's desk.
The thesis for this idea of Industry-as-a- Platform (IaaP) started rather innocuously. A press release that arrived at ProgrammableWeb's offices hailed the general availability of GE's Predix; "the company’s purpose-built cloud for digital industry" according to the email accompanying the announcement. Having read such announcements many times during my 25 years as a tech journalist, I rolled my eyes. A "purpose-built cloud for digital industry?" If ever something sounded like marketspeak BS, this did. But this is GE, right? I've seen frivolous gratuitously worded announcements from big players before. But if GE was putting its stake in the ground as a new cloud provider, my gut told me that it could barely afford an early faux pas. Maybe there was something to this.
Deep in the announcement was mention of Cloud Foundry as the underlying muscle behind the initiative. So, instead of calling GE just to be served a helping of its Kool-Aid, I called Cloud Foundry CEO Sam Ramji to find out if there was any merit to this idea of an industry-specific cloud.
Ramji identified some of the forces at play that will enable GE and other like-minded companies (each in their own industries) to get their slices of the pie; Cloud (where the last set of conspiratorial forces left off), big data, and machine learning (ostensibly couched under the broader rubric of Artificial Intelligence). Google and Amazon may have no idea of the extent to which they have informed this next wave of disruption.
Here at ProgrammableWeb, over the last year or so, we've been fascinated by the rise of what we call Ph.D. APIs. These are APIs that pack the brains of a team of Ph.D.s into a single API call. Sure, you could have hired your own team of Ph.D.s for a couple billion dollars. Or, you can just outsource your Ph.D. requirements for facial recognition, semantic analysis, or entity extraction to an API for pennies.
Take Google's recently announced Cloud Vision facial recognition API for example. Compared to your typical API-driven database query, facial recognition is the stuff of rocket science. But only a handful of companies like Google and Facebook are in a position to get it right. Why? More than half the planet's faces along with other identifiers are stored on their servers. Remember when Google announced unlimited free storage for Google Photos? Yes, it's unlimited. But only when you agree to let Google deliver compressed versions of your images upon request. Meanwhile, you uploaded uncompressed versions. More pixels means more data. More data means better accuracy. But better accuracy of what?
Enter Machine Learning. If you think about it, machine learning can't be all that useful unless the machine has gazigabytes of data to sift through. The more data, the more the machine learns and the more Google's rocket scientists can control the machine learning knobs and levers (for example dialing image compression up or down) in an effort to optimize accuracy. Bottom line: Google couldn't do this without access to the data (and its own team of Ph.D.s, of course), the elasticity of a compute cloud, and machine learning.
So, what if anything does this have to do with industries as platforms or GE Digital's Predix announcement?
According to Cloud Foundry's Ramji, in the end, it's actually the guys with the data who will win. Compute power and storage are interesting. But thanks to open source and Amazon, the margins in that business have been all but wiped out. When GE announced its so-called cloud for industry, it had no interest in competing over bare metal. "GE is going to sell aircraft engine uptime to air carriers like American Airlines." said Ramji. "Given the cost of downtime, what airline wouldn't buy that?"
Wait. Aircraft engine uptime? That's not a cloud, is it? Or is it?
Consider the thousands of miles that all of GE's aircraft engines rack up in aggregate every day. Then imagine GE having access to all the data that those engines are generating through their sensors and other diagnostics as they fly each mile. This is beginning to sound a lot like having access to the high-resolution versions of millions of faces. Now throw some Ph.D.s and some machine learning at that dataset and pretty soon, GE is able to predict with uncanny accuracy what's going to go wrong with an engine, and when. So, instead of waiting for an engine to break in a way that results in a delayed flight, bad ratings, and angry customers, American Airlines has a better idea of when to schedule an aircraft for maintenance instead of a flight. Like Ramji says, "[What airline] wouldn't buy that?"
Then, there's the long game. Regardless of who it buys its aircraft from (Boeing, Airbus, etc.), American Airlines could start to request GE engines on its new aircraft because of the long term business benefits of engine up-time.
So far though, this still sounds more like an application rather than a cloud for industry. For example, couldn't anybody go out and build a highly targeted vertical app like this on top of Amazon or Rackspace? Maybe yes. For example, Rolls-Royce, another company that makes jet engines, could create the same benefit for its customers.
But, according to Ramji, without a certain critical mass, it may be difficult to recoup the exceedingly non-trivial investments that must be made to get the combination of cloud, Big Data, and machine-learning right. "Unless you're able to leverage that investment across multiple applications," said Ramji, "for example, across a whole bunch of industries for which you're collecting tons of data on a daily basis [the way GE is], that investment is less likely to work out."
And this is where the empire strikes back. GE has it hands in just about every major industry on the planet; Aerospace, Energy, Oil & Gas, Healthcare, Mining, Telecommunications. You name the industry. More than likely, GE's footprint covers it. Like an industrial Internet of Things, with each of those industries streaming petabytes of device data on a daily basis (through APIs no doubt), the ROI from a cloud involving a central investment in machine learning that can be leveraged across industries -- an "industry cloud" -- starts to sound a little less like Kool-Aid.
Lastly, what if anything does Cloud Foundry have to do with the idea of an industry cloud? Cloud Foundry isn't just a way to auto-erect an enterprise cloud with all the requisite things you'd want a great cloud to have (application hosting and lifecycle management, Authentication, message bus, etc.). It also does this like middleware across IaaS platforms like Amazon Web Services, OpenStack, vSphere and Azure. In other words, whatever you build to run on Cloud Foundry is portable across a variety IaaS stacks. "That kind of leverage," says Ramji, "means your IaaS provider can't hold you hostage." Furthermore, for reliability's sake, that kind of portability means that you can also distribute your cloud across multiple IaaS providers "the way Apple's iCloud is distributed across AWS and Azure" said Ramji. "This means your cloud can survive events like when Amazon performs a region-wide restart because the Azure part of your cloud is still running."