A Dispatch from Cloud City – State of the Union 2016

With a venerable tradition dating back over a year, the annual Platformonomics state of the cloud union strives to combine the exhilaration of the running of the bulls at Pamplona with the hyperbole of Oracle’s annual proclamation that this year they really are serious about cloud. Or at least to land a few jokes along the way.

In summary, we’ve reached the end of the beginning for cloud computing.

There is no longer much question whether public cloud will be the foundation for IT going forward; instead we quibble about timing and implementation details. The largest enterprises as well as the most sophisticated workloads are wafting up into the cloud. The leaders are distancing themselves from the pack while the dreams of cloud wannabes are deflating like footballs around Tom Brady. Legacy vendors’ worlds are imploding. Private cloud proponents are harder and harder to find: except for those few diehards hunkered down in their closet-sized data centers with several years supply of canned goods and tape backup cartridges, previous private cloud proselytizers now talk earnestly about hybrid clouds in hopes of retaining a few on-premises crumbs in the process. And even the very largest corporations are realizing they can’t keep up with the hyperscale public clouds.

I contend there were two critical inflection points for cloud this past year:

Customers tipped, specifically the enterprises who spend vast sums on IT. Most CIOs have shifted from resistance or tire kicking to active embrace, and are doing so increasingly for business reasons as opposed to technical. Sticking your head in the sand is no longer a viable option. The objections have been knocked down one after another. Security turned out to be powerful a reason to go to the cloud, not shun it. The enterprise tipping point is critical because it dramatically expands the size of the cloud opportunity. We can now realistically talk about a trillion dollars of existing IT spend in play, aka the “cloud jackpot”.

Amazon’s transparency, both financial and cultural. The breaking out of AWS financials in April forever banished the platitudes “your margin is my opportunity” and “the race to the bottom”. AWS proved to be a very large, very profitable and very rapidly growing business. Even bulls were surprised to learn not only that the business is profitable, but much more profitable than anyone imagined. The initial operating margins for AWS were almost identical to those of financial engineering savant IBM. Amazon also had some unsolicited transparency inflicted upon it by the New York Times, who took a deep look at the company’s culture.


My thesis for the last two years has materialized: it is a two horse race located here in Cloud City (Seattle) with AWS in the lead and Microsoft the only other vendor who can still see them. Besides being extremely convenient for me, this means your cloud landlord is probably in Seattle. Please don’t be late with the rent check. The geographic version of Gartner’s Infrastructure-as-a-Service Magic Quadrant (™ ® © All rights reserved. p = 0.796513. Trough of disillusionment. Etc.) underscores that the cloud world is not flat.

It is so obvious that even the denizens of Wall Street have noticed, with one brokerage firm hyping it as a ‘206 area code street battle for the cloud’. (Never mind that Microsoft is in a different area code. I’m sure they’re using an area code map from The New Yorker where everything west of the Hudson blurs together, just as all those buy and sell recommendations from east of the Hudson blur together).

Where is Google in this race? In some ways they have the fastest horse and are certainly the third hyperscale player in terms of their global infrastructure footprint. But Google’s horse is sitting some other pasture, contemplating space elevators, indifferent to the idea they need to actually show up for the race to win it.

I have a fundamental question for each of the hyperscale players pertaining to whether and how market shares will shift as this market continues to grow, plus some thoughts on the rest of the rapidly diminishing field.


Amazon remains the cloud trailblazer, maintaining their frenetic pace of innovation while also making necessary investments to become a mainstream enterprise provider. The question for AWS is can they adapt and evolve their culture in order to extend their current leadership into dominant share of that trillion dollar cloud jackpot? (Note that cloud will also bring significant revenue compression, aka customer savings.) This is very much an issue of “what got you here won’t get you to the next level.”

Beyond all the substantive if boring investments required to sell to and support enterprise customers, there are a bunch of cultural issues AWS must navigate. Some stem from their position inside Amazon and some are unique to AWS. The broader Amazon culture issues that the New York Times highlighted also impact AWS’s ability to realize its potential, not least their ability to hire and retain talent. AWS is a very different business from the rest of Amazon and one sitting on the pole position of a trillion dollar opportunity. It requires a different culture than the core Amazon MVP trial balloon autocannon and one that doesn’t resort to zero sum political hackery to assuage its ego.

Public cloud providers are among the most important dependencies any company will take. Successful vendors in this position understand the nature of this relationship with their customers and actively work to build customer trust and mutual co-dependence. Not surprisingly, enterprise vendors are very transparent with their customers. Yet this is at odds with the secretive Amazon culture that seems incapable of putting numbers on the y-axis of charts.

Even more, successful enterprise vendors mitigate customer fears of lock-in. AWS has not figured this out and is struggling with lock-in fears, as evidenced by what can only be seen as disappointing adoption of higher level services like the EC2 Container Service and Lambda, despite their technical appeal. Business as usual will not overcome these fears, and not addressing them means a future where customers only feel comfortable consuming base compute and storage. Being cognizant about your own power is challenging, as big technology companies’ internal mindset invariably lags their growth. They go on thinking they’re the plucky little startup long after they’ve become Godzilla.

I used to think Amazon should spin off AWS so it could maniacally focus on retaining or expanding their current share of the cloud jackpot, and build the distinct culture necessary to fully realize that opportunity (and avoid the distractions from the rest of Amazon). After seeing the financials, I believe AWS should spin the rest of the Amazon e-commerce business.


Microsoft has executed extremely well to emerge as the only credible challenger to AWS, leveraging both their platform heritage plus the fortune of a massive and overly-optimistic infrastructure build-out for search. Further, they’re the only vendor from a previous generation to make the leap to hyperscale. Unlike many of their peers, Microsoft’s survival in the cloud era is not in doubt.

But as the enterprise market for cloud really begins to open, the question for Microsoft is whether they can bring their enterprise capabilities to bear in a way that both reels in AWS and allows them to materially expand their share of the cloud jackpot. It is not clear Microsoft fully appreciates those enterprise capabilities, in relative or absolute terms. It is a long road to become a credible enterprise vendor, and having lived through that process when I was at Microsoft, it brings great cognitive dissonance to realize they are by far the best of the hyperscale bunch (and it is even weirder to see the company getting good marks for “Playing well with others” these days). Microsoft also has an advantage as a full spectrum provider across IaaS, PaaS and SaaS, to which AWS is just starting to react. But more of the same is not going to materially increase Microsoft’s market share position. Further success starts with a strong dose of self-awareness.


The big question for Google is when will they realize cloud is more than just an engineering problem? If they want to build a real business where customers take a enormous dependency on them, they are going to have to do some critically important but mundane things that don’t involve algorithms. Worse, it is likely to involve fickle humans. They must overcome their deep antipathy to both customer-facing operations and enterprises as customers.

Post Alphabet, where any previous inhibitions about pursuing new hobbies have evaporated, it is even harder to imagine the “capital allocators” choosing to invest in thousands of enterprise sales and support people given alternatives involving life extension and/or space elevators. After all, won’t the robotics division eventually solve any problem that today requires humans?


Last year we catalogued the delusions afflicting a long list of public cloud wannabes. This year we simply observe the epidemic of sobriety sweeping the vendor landscape (and the morning-after wreckage). HP managed to exit the public cloud business not just once but twice this year. Helion is Heli-off. Rackspace, still recovering from its OpenStack misadventure, is shifting its center of gravity from the data center to the call center. Both vCloud Air and Virtustream have disappeared into a miasma of highly leveraged financial engineering emanating from Austin. AT&T, CenturyLink and Verizon are all hoping no one remembers they once claimed to be public cloud providers (and probably will get away with it). Cisco, presumably, has filed a missing persons report for their InterCloud.


While the number of hallucinating vendors has plummeted, devotees of delusion should not despair. Despite all the departures, aggregate levels of industry delusion may be hitting new highs between the efforts of IBM and Oracle. These delusional dinosaurs are locked in a battle every bit as fierce as one between the hyperscale competitors, except they are vying for the World Championship of Cloudwashing™. Given cloud poses an existential threat to both companies, it is not surprising they are talking cloud. But their delusion manifests itself in the colossal gap between their rhetoric and their actual capabilities.

I have been arguing for almost three years that IBM is likely to be the cloud’s biggest scalp. Their best outcome is they’re just a much smaller company in the cloud era, not that they’re executing on that path. The stock is down a third since I started beating this drum and is currently exploring new five-year lows. They continue to confuse boutique hosting with hyperscale cloud, and have been reduced to asserting Watson will somehow be their cloud Hail Mary (at what point is it reasonable to expect Watson to progress beyond an endless PR campaign, never mind drive revenue material enough to bolster the ever-shrinking IBM topline?).

A year ago IBM had the cloudwashing title wrapped up but Larry “Lazarus” Ellison is not one to back away from a challenge. Hypercompetitive: yes. Hyperscale: not even remotely. The question for Oracle is do they really believe it when they assert they are the leaders in cloud (or even have a cloud as opposed to some SaaS apps?) or they believe that empty rhetoric is a legitimate substitute for millions of lines of code and billions of dollars of capex? It is embarrassing when your employees feel compelled to point out the discrepancy between announcements and action, and in particular recurring confusion around tenses (also a lesson here for press who happily write the “this time we’re serious AND we are already the clear leader” Oracle cloud story every year without reflecting upon their credibility or past proclamation performance).

But this speaks volumes about Oracle’s cloud:

For instance, when the team was struggling with Oracle’s central IT to get the server resources they needed, the team requisitioned a bunch of desktop computers from Oracle’s Seattle office and turned them into an OpenStack-powered private-cloud-development environment so they could continue their work in peace, right in the middle of the office floor.

IT involved? Check. Private cloud? Check. OpenStack? Inauspicious. Desktop computers under the desk? Are you f*%king kidding me?

To paraphrase William Goldman: “Follow the capex” with IBM and Oracle. We’ll see if they’re still pretending next year.


Dell/EMC/VMware/WTF: the metal-bending M&A muttonheads have likely inflicted irreparable damage to VMware, the best asset in the so-called “federation”. Pivotal also risks being caught up in financial shenanigans perpetrated by those who neither understand nor appreciate software.

DevOps: if you’re buying DevOps tools, you’re doing it wrong.

Digital Ocean: needs to make its play as the dark horse window is closing.

Docker: despite all the political hijinks as competitors tried to box Docker in, Docker has become boring. That is good; the container infrastructure continues to mature. More exciting perhaps are new developer models emerging that are “native” to containers.

GitHub: the Craigslist of cloud?

HubSpot: this is not cloudy, but given the infrequency of my blogging, I will predict their CEO steps down in 2016 with p = .7. The board may follow. The level of transparency has not yet become “uncomfortable”. But it will.

Industry Foundations: after an ugly outbreak of industry foundations last year, we can only hope to be certified Foundation-free in 2016. As we have seen, this affliction is highly contagious. As with cockroaches, when you see one foundation, you will likely see more. So it is important to prevent potential foundation epidemics; the best protection is not letting companies that can’t write code get involved.

PaaS: still a zero billion dollar market though the data is suggesting I might finally have to stop using that line next year. Perhaps more importantly, containers have reinvigorated the endless ontological debate about what exactly constitutes a PaaS. Cloud Foundry is having some success selling to very large enterprises, but they seem to be selling hope more than product. The Fortune 500 is packed with companies grasping for anything that lets them believe they can become software companies.

OpenStack: like a poorly performing European football team, OpenStack has been relegated to a lower division. It is now a solution for telcos. As the saying goes, if at first you don’t succeed, you can still sell it to telcos. OpenStack is a great fit with the NFV misdirection, which gives telcos the infrastructure toys everyone else had a decade ago while leaving the networking crown jewels firmly in vendor hands.

(Free) Stock Tips: if wave one of the cloud disruption hit enterprise hardware, wave two is hitting enterprise software. VMware preemptively tubed its stock by letting itself be the funny business in the Dell-EMC deal, so it not clear how much more downside there is in VMW. Oracle’s stock has already started to roll over. But there is still time to short Red Hat who, despite being irrelevant to cloud, sports a multiple of over 75 yet will see a much smaller fraction of every dollar that shifts to the cloud. If you have a cloud infrastructure software company to sell, Red Hat is your first call.

Open Season: A Short Industrial Drama

Cloud Foundry had a pretty good week with endorsements from Baidu and IBM. Both relationships were developed after I left VMware so what follows is purely speculation on my part. But some companies have a tough time getting over their history and playbooks, so it is easy to imagine how things went down.

Warning: this post contains serious “inside baseball” about the past and present of software standardization and open source mechanics. If you don’t know what ECMA oxymoronically used to mean or haven’t debated the merits of different open source licenses, you may want to stop reading right now (go see Pacific Rim or read up on Bitcoin instead). I may be the only person who gets some of these jokes. Apologies to David Mamet.


Scene: a hipster office in SOMA populated by dogs, twenty-something Siamese programmers and two older gentlemen trying with limited success to project a casual air.

Jim – a Pivotal executive
Dan – an IBM executive
Angel – an IBM standards executive

Dan: We’re from the IBM company and today is your lucky day. We have decided Cloud Foundry is going to be the platform-as-a-service for the cloud.

Jim: Oh…

Angel: It’s still early days for the cloud.

Dan: The way this will work is IBM will make Cloud Foundry open and therefore viable for the enterprise. We know how to do this and will tell you what you have to do.

Jim: I’m not sure I follow as our strategy has been to make Cloud Foundry as open as possible from day one. Am I missing something?

Dan: Cloud Foundry cannot be used in the enterprise until IBM gives it our blessing. It is critical that enterprises only use open technologies.

Jim: Open like the mainframe?

Dan: Watch your tone son. We’re from IBM and we make sure that enterprises are not locked into proprietary technologies.

Jim: I’m definitely not following you. What do you mean by “open”?

Dan: Openness depends on having a comprehensive governance strategy. We will work with you to create a Cloud Foundry Foundation to manage the governance of Cloud Foundry.

Jim: What exactly would such a Foundation do? And isn’t “Cloud Foundry Foundation” kind of awkward phrasing? Did you consider just Foundration? That domain might still be available.

Dan: The Cloud Foundry Foundation will handle the governance of Cloud Foundry. With a formal governance process as defined in bylaws, Cloud Foundry will then be open so enterprise customers can embrace it.

Jim: Hmm.. I assume you’d describe GE as an enterprise customer. They’ve embraced Cloud Foundry to the tune of investing $105 million. And they’ve never mentioned the word governance as far as I can recall. They have been known to throw around terms like productivity and time-to-market.

Dan: Let me help you understand how this will work. Do you remember Java and Linux? IBM made those technologies successful in the enterprise by ensuring they were open. We will do the same thing for Cloud Foundry. But you will need to follow our direction.

Jim: You’ll have to excuse me as I was in junior high school when you were running that playbook for Java and Linux. But I’m still not certain what this has to do with Cloud Foundry.

Dan: Enterprise customers expect new technologies have formal governance processes so they can trust them to be open. For example, it is critical there be explicit rules to specify the voting rights for different classes of membership and how to deal with conflicts of interest on the board of directors.

Jim: I am afraid I still don’t understand what this has to do with making it easier for customers to build applications for the cloud. I defer to your knowledge of the previous century as well as conflicts of interest, but it seems customers today are more focused on functioning code that solves their business problems than governance processes. But you should explain to me what governance you think is necessary.

Dan: The Cloud Foundry Foundation will be a legal entity with a steering committee which will define all the subcommittees necessary for different aspects of Cloud Foundry. Obviously, we will use Robert’s Rules of Order.

Jim: Is this how you created the Java programming model?

Dan: Exactly.

Jim: You do know the majority of enterprise Java development is done today with the Spring Framework which was developed to shelter developers from the horrors of committee-developed technologies like EJB?

Dan: Son, enterprises can’t build enterprise solutions without enterprise technologies like EJB. When you start doing transactions, it is no longer child’s play. You may have your simple solutions for simple problems, but IBM solves enterprise problems.

Jim: I won’t ask how it is that the biggest Internet companies on the planet somehow manage to do transactions at vastly greater scale than any enterprise that uses IBM technology. I guess I should also be surprised the cloud has gotten as far as it has without any committees and Robert’s Rules of Order.

Angel: It’s still early days for the cloud.

Jim: Your faith in committees is touching, but pretty much for every broadly successful technology in the world today, there was a committee-driven alternative that failed. Take IP vs. OSI, or HTML succeeding only by throwing away the vast standardized bulk of SGML. I think the world has learned from these experiences. We have seen over and over that committees are prone to making bad political tradeoffs, delivering least common denominator solutions and losing sight of the real problem at hand. Premature standardization is a killer; you need to allow for experimentation, evolution and finding the proverbial product-market fit.

Dan: Son, you need to understand how things work in the enterprise.

Jim: Is there more to IBM’s cloud strategy than a vague appeal for standards? Do you really think a bunch of random committees are going to keep up with Amazon Web Services? I guess it could be a good strategy if they’re laughing so hard they can’t get any work done.

Angel: It’s still early days for the cloud.

Dan: Let me give you a recent example. Have you heard of OpenStack? IBM is making OpenStack part of the open enterprise cloud.

Jim: I am familiar with OpenStack as it turns out. In fact, Cloud Foundry runs on OpenStack. I do seem to recall you guys jumped on the OpenStack bandwagon a couple years after it got started. Are you saying IBM is somehow responsible for OpenStack’s momentum?

Dan: IBM is making OpenStack open and acceptable for enterprises.

Jim: So what contributions have you made to OpenStack?

Angel: We have dozens of our best standards people working on OpenStack.

Jim: I was thinking more in terms of code. NASA and Rackspace have contributed major pieces of technology – what has IBM brought to OpenStack?

Angel: It’s still early days for the cloud.

Jim: Well, even if software development is not your focus, you do operate a lot of outsourced IT infrastructure. With IBM’s enterprise presence you must have a lot of customers running OpenStack today. How many megawatts of OpenStack capacity are you operating?

Angel: It’s still early days for the cloud.

Jim: I’m not sure where you guys have been for the last decade, but the world has changed. We now achieve openness at the engineering level, not with lawyers writing bylaws and Robert’s Rules of Order. The days of heavyweight governance via committees staffed by people whose primary skill is sleeping while sitting up have probably come and gone. Cloud Foundry is extremely open today by any practical measure. The code is all on GitHub under the very permissive Apache license. Is there something we’re missing?

Angel: What is this Geet Hub? Can you spell that for me?

Jim: GitHub is a public code repository. Anyone can submit a pull request and contribute code to the project. If you don’t like the vision or want to do something different that is more tailored to your specific needs, you can always fork the project and take it in whatever direction you want.

Dan: (visibly flinching and frothing) Are you mad? Anyone can just contribute code? To a product that will be used by enterprises?

Dan: You encourage people to fragment the project by modifying it and making derivative works? (pause)  Do you not know any history boy? We spent years trying to minimize Java fragmentation. Microsoft would taunt us that even Ivory soap was only 99 and 44/100th pure. Despite Herculean efforts, we never quite achieved it, but we tell ourselves, much like with the current economic recovery, it could have been so much worse. You would let anyone do whatever they want with the software? (aghast)

Dan: How will enterprises ensure they’re getting the official version of Cloud Foundry? Do you not see how critical it is to have a Foundation that controls Cloud Foundry?

Jim: The market decides what the best version of Cloud Foundry is, not some committee. If you don’t like the direction, you could always fork and go in whatever direction you think is most appropriate. One would think with 400,000 or so employees, IBM would have some people who could write code as opposed to committee minutes.

Dan: Surely you jest. We can’t rely on the market to make decisions for the enterprise. That is IBM’s role and has been since the dawn of information technology. If you don’t fully appreciate the criticality of governance, we can go elsewhere. We have options. We could bless OpenShit, sorry I mean OpenShift instead. I bet Red Hat would play ball. We’re old friends with their standards guys.

Jim: Good luck with that.

Dan: Or we could bring the full might of IBM’s research labs to bear and build our own platform-as-a-service. Don’t underestimate the technological prowess of the IBM company. We get more patents every year than any other company. We can write the letters IBM at the atomic level. We are going to positively own the burgeoning robotic game show contestant market. We can make WebSphere the application platform for the cloud. WebSphere is the biggest middleware on the planet, though I’m not sure why the development team was laughing when they said that. If you don’t hand over Cloud Foundry to the Cloud Foundry Foundation, we’ll just compete with you.

Jim: You’d think with all those great patents, you’d have more innovation to show in your product line and wouldn’t be here trying to figure out how to co-opt the fruits of someone else’s R&D. I get that what’s yours is yours, like the mainframe, but you’d also like what other people have developed to be under your control. You’re welcome to participate in the Cloud Foundry ecosystem on the same level playing field as everyone else, but we’re not going to distract ourselves from building a great platform with some giant bureaucratic foundation. If you want to compete, by all means compete, but at some point you’re going to have to write some code people actually want to use. Maybe you can create an IDE that lets people write code at the atomic level. And with all due respect, WebSphere at this point is just a middleware museum. It is about as relevant to the cloud as the mainframe.

Dan (quietly to Angel): They’re onto us. Our strategy of blessing different piece parts defined by multiple slow-moving and conflicted committees that don’t work together well and need busloads of consultants to make them limp along may not fly in the cloud. This may be a problem for our earnings roadmap. Our CFO told Wall Street we’d have $7 billion in cloud revenues by 2015 and SmartCloud unfortunately isn’t looking that smart.

Angel: It’s still early days for the cloud.

Jim: I’ll tell you what. I’d hate for you to have to go back to Armonk and get yelled at by your CEO again for not working hard enough and not bothering to return customer calls. We’re doing a Cloud Foundry developer conference this fall and how about IBM sponsor breakfast there or something? You can buy some healthy fare and we can explain how in the past you would have brought donuts, but you’ve gotten religion about reducing middleware girth. You can even come to the advisory board meeting. And of course you can submit all the code you want to the project, but I realize that may not be your thing. But I do have one request if we’re going to work together: please don’t ever use that the word governance again in my presence.

Angel: It’s still early days for the cloud.

Jim: Yes, it’s still early days for the cloud…at IBM.


Note: the voices in my head for this are the default Xtranormal voices. In the sequel, the Bernank will make an appearance.