In December of 2000, then CEO Lou Gerstner pledged to the audience at the eBusiness Conference and Expo in New York that IBM would spend one billion dollars on Linux in the year ahead. Even if you adjust that number for inflation, it’s around $1.5B in today’s dollars, which doesn’t sound like much when Amazon, Google and Microsoft have averaged a spend of $1.4B per month in gross PP&E over the last five years. It’s worth remembering that at that time, however, the Linux then was not the Linux of today. It was far from clear that enterprises would accept the product of a Finnish grad student over its biggest competitor, Windows NT. IBM’s pledge helped ensure its path forward.
The interesting thing is that while IBM promised to spend on Linux, what it actually meant was around Linux.
In the year 2000, Microsoft was the dominant supplier of operating systems, which at the time were the most fundamentally strategic piece of technology within the enterprise. The company that controlled the operating system, after all, could exert control over the applications that were written to that operating system.
IBM’s approach to the market was not to try and outcompete Microsoft in operating systems, and it had no intention of spending a billion dollars in trying to be a competitive supplier of them. It explicitly and deliberately left that challenge to partners like Red Hat and SUSE.
Instead, the company’s approach to the market was two pronged.
First, it attempted to limit Microsoft’s operating system market dominance, if indirectly and, in the early stages, tactically. The billion dollars – and perhaps more importantly, the explicit signal of intent that investment was intended to convey – were designed to inject confidence into the ecosystem around the open source operating system and make it easier and safer for enterprises to adopt. Whatever revenue IBM’s Linux partners were able to capture was revenue denied to Microsoft. The lost revenue was a rounding error at first for a company with twin licenses to print money, but that would gradually change over time.
Second, IBM attempted to make the operating system itself less important. The company’s strategy was to diminish the strategic role of the operating system via a higher layer of abstraction generally – the Java application server middleware product WebSphere specifically. Applications written in Java targeting the WebSphere platform, or competitive offerings such as WebLogic, were applications not written directly to the Windows operating system, thereby decreasing the control Microsoft would be able to exert over customers.
Eighteen years after that original promise, IBM invested not one but thirty-four billion dollars to acquire its one-time partner Red Hat. On the surface, the purchase of a vendor best known for its flagship operating system business looked like a repudiation of IBM’s former strategy – one based on the idea that the value to be captured actually was in higher levels of abstraction. In truth, however, the acquisition of Red Hat is perhaps best read as a validation of that original strategy, because the bet is clearly on the idea that the Kubernetes-based OpenShift can replicate the success of WebSphere by providing a middleware-like layer between applications and the operating environments, whether those are on prem or off.
In this context, out of the myriad announcements streaming out of Google’s Next OnAir virtual conference, the most important may be the news that BigQuery will run in competitive clouds.
A year ago this past April, remember, Google announced Anthos. Anthos was interesting not because it was a Kubernetes-based platform for hosting applications – there were many such then and there are more today. What differentiated the platform was that it was a cloud-independent piece of middleware offered by one of the cloud vendors themselves. Historically, a core approach of cloud vendors to the market has been attracting workloads to their platforms via proprietary offerings.
With Anthos, Google inverted that model, taking the proprietary software platform and decoupling it from the underlying cloud. This is notable because it represented a change of strategy, as noted, but also because it’s a heavy lift technically. It’s much easier to get a platform stack operating in one environment than it is in multiple.
Now with BigQuery, Google is doubling down on this approach, extending it beyond its GCP roots into other environments. This may weaken the argument for GCP compute at the margins, but it inarguably expands BigQuery’s addressable market dramatically.
It is also suggestive of a broader shift in strategy from Google. It is possible, of course, that the Anthos-enabled cross platform BigQuery is merely a tactical announcement, one that opportunistically seeks to capitalize on Google’s database prowess but is not more broadly representative.
But it seems equally plausible that Anthos and BigQuery are merely the first manifestation of a fundamentally new approach for Google. For years, would be challengers, armed only with largely similar offerings, dutifully charged up the hill that was an incumbent AWS operating at velocity in its core market. Most were cut down. Google, in particular, never seemed to benefit from this approach.
More recently, however, there have been signs of creativity. It’s arguable, for example, that Microsoft, via its acquisition of GitHub, was attempting to outflank AWS by delivering a tighter, more integrated and developer-friendly experience. There being only one GitHub from a center of gravity perspective, that approach was always going to be a challenge to replicate.
Which left Google with an uncertain path towards its goal – mandate?: competition for the top spot.
BigQuery on other clouds may be something new, however. Once upon a time, IBM took a long and critical look at the enterprise landscape and decided its prospects were best served by moving up the stack from its competition. Rather than attempt to outperform the the unstoppable juggernaut that was Microsoft in operating systems, it would attempt to relegate the operating system to commodity status by making the abstraction that ran above it the really strategic – and thus valuable – piece of technology.
While its product announcements suggest and infrastructure spend confirms the idea that Google is not content to leave the underlying platform to third parties as IBM once did with Red Hat and SUSE (though as an aside this is particularly interesting in that context) Google‘s willingness to abstract itself across competitive cloud platforms looks quite a lot like the classic middleware play. A play the CEO of GCP knows of old from his days running Oracle’s Fusion middleware business.
If it is indeed attempting to replicate IBM’s classic middleware strategy with Anthos and BigQuery, one of Google’s biggest opponents will, ironically, be none other than IBM.
The combination of IBM and Red Hat, of course, along with VMware and its multiple acquisitions, from Heptio to Pivotal, have been following the classic middleware playbook for years. Google’s strategy here – assuming, of course, that that is its strategy – therefore may be unique amongst its hyperscale cloud provider counterparts, the likes of Azure Stack or AWS Outposts notwithstanding, but it is common in the on premises software world. This would be the first time it has been executed by a player with a true hyperscale infrastructure behind it, however, which is an important potential differentiation.
In any event, the next 12 to 18 months will tell us much about Google’s overall strategic approach to the market. One logical question in the meantime is what might come next if BigQuery is indeed the shape of things to come? There are many different possibilities, but watch Cloud SQL closely. With AWS having extended RDS onto its Outpost hardware, a logical next step for GCP would be making its managed relational database infrastructure available cross-platform. If you assume that most customers will be multi-cloud in some capacity, the availability of a platform with core compute as well as analytical and relational database capabilities that would run independent of the underlying cloud would certainly be an attention getter.
One that, importantly, would potentially make the infrastructure it ran on less important much as middleware was designed to do all those years ago.
Disclosure: Amazon, GitHub, Google, IBM, Microsoft, Red Hat and VMware are RedMonk customers. SUSE is not a RedMonk customer.