Premise: Amazon has turned the data center into an API. This trend is having profound impacts on enterprise IT customers. In particular, the economics of infrastructure outsourcing (i.e. deployment, provisioning, management and orchestration), which formerly had negative economies of scale at volume, are beginning to track software marginal economics – i.e. incremental costs go to $0. To compete with these cost structures, IT organizations and competitive cloud vendors will have to either have massive scale or become highly vertically integrated.
Full coverage including exclusive video from AWS re:Invent 2014
The Shifting Enterprise IT Landscape
The enterprise IT business is restructuring. BMC, Compuware, Tibco, and Dell have gone private, HP is splitting in two, and IBM is selling off its x86 server business and paying $1.5B to shed its microelectronics division. Meanwhile, activist shareholders are pressuring EMC to give up its most strategic asset, VMware, and pressing companies like NetApp and Juniper to make cuts and/or management changes.
These forces are driven by several factors, of which the most visible are: Enterprise IT spending is flat; large established enterprise players like HP, IBM, EMC, and others aren’t growing; while the street craves growth, private equity has a strong appetite for companies with great cash flows and has plenty of “dry powder” to invest in these firms; Web scale technologies and open source software are bleeding into the enterprise, and rather than keep investing in legacy technology, IT buyers are getting ROI by cutting the denominator in the benefit/cost equation and realizing gain through “reduction of investment.”
The Royal Philips Example
Earlier this year we interviewed Alan Nance, VP of Technology Infrastructure at Royal Philips, a 140-year-old company.
Here’s a summary of what Nance told us:
- For years, Philips has had standalone P&L’s, each with its own IT agenda.
- Philips’ CEO announced that 85% of its IT spend was non-differentiated.
- Philips then decided to completely standardized its IT, end-to-end.
- Its infrastructure is now built on seven cloud-based services:
- Data Analytics,
- IoT and Mobile,
- Security, and
- A Control Plane connects all the elements across the company.
Here’s the key piece of the strategy: All IT purchases at Philips will be consumption-based. No up front fees, no setup fees, no ELAs, no penalties for turning off services. All services will be consumed as “pay-by-the-drink.”
While this is not unique to Philips, it’s a strong example of a leading global company recognizing the reality that there’s a new and better approach to infrastructure technology management.
The Marginal Economics of Enterprise IT
When IBM handed its monopoly to Microsoft and Intel in the early 1980’s with its decision to outsource the PC operating system and microprocessor, the industry learned about the power of marginal economics. In economic terms, marginal cost refers to the delta in total cost that results when an additional unit of goods is produced. It took quite some time during the PC era for some very important people to grasp the implications of this concept.
IT industry historians may remember that when IBM decided to part ways with Microsoft, IBM said, “We’ll take OS/2, you can have Windows…” believing OS/2 was a superior product and not fully understanding the implications of the software model at volume. In another example, Lotus, observing that Microsoft was encroaching on its applications business, initially chose not to support Windows 95 and instead decided to run its apps on the DEC VAX minicomputer operating system, among others. As late as the mid-1990’s Netscape’s Jim Barksdale downplayed the threat of Microsoft bundling the browser into its operating system, insisting that Netscape would thrive as an enterprise software company. The industry became littered with failed attempts of these Microsoft competitors (and others like Novell buying WordPerfect) who didn’t fully comprehend the economic power of software at volume.
Figure 1 below depicts the different business models for the major enterprise IT sectors.
The chart shows marginal cost on the vertical axis and volume on the horizontal plane with the lines showing the conceptual model for each of the major enterprise sectors (hardware, software and services). The following points are noteworthy:
- Software is the most attractive model. The up-front engineering investment is offset at volume as the incremental cost to build a new unit of software eventually goes toward $0. In the case of Microsoft, it became the cost of the floppy disks or CDs on which the software was distributed, virtually nothing.
- Hardware has economies at scale but because of underlying component costs required to build the next unit, marginal costs never go to $0.
- Professional services have dis-economies of scale. As the number of projects goes up and scale goes global, complexity increases and bodies become more expensive and actually less productive at volume.
- Outsourcing has a profile similar to professional services, although some relatively minor efficiencies are achieved. But ultimately outsourcing became “my mess for a bit less” and wasn’t that attractive.
Microsoft’s operating system ubiquity led to a monopoly because its volumes allowed the company to (for example) bundle PowerPoint, Word, and Excel into a suite at a price point similar to standalone products such as Lotus 123, Harvard Presentation Graphics, or WordPerfect—all with large, loyal user bases that ultimately had to give in to Microsoft’s value proposition.
Today, only Apple and Google can match Microsoft’s $400B market cap and $65B net cash hoard (notably, IBM and HP have negative net cash on the balance sheet).
IT Moves to a Utility Model
The above dynamics set the tone for the enterprise IT industry and they haven’t changed much, until recently. While VMware and other hypervisors helped consolidate IT resources, it was really Amazon’s introduction of EC2 in 2006, followed by S3, which began to alter the economics of the enterprise IT business. The downturn in 2008 and 2009 accelerated the move by CFOs to shift CAPEX to OPEX and was a tailwind for Amazon, leading zealots to believe that “the cloud” would kill established players – the same way Microsoft and Intel ascended in the PC era at the expense of established companies. In a famous rant at the Churchill Club, Larry Ellison railed on such thinking leading many to incorrectly believe that he didn’t take the cloud seriously.
Regardless, Amazon’s dramatic and unending rollout of additional cloud services positioned it to compete head-on with established enterprise IT players. The enterprise players were comfortable coining the term and building near-term strategies around “Private Cloud.” However, in 2012, Amazon’s first re:Invent conference underlined its strategy to bring a low margin, high volume model directly into the enterprise. Amazon’s win at the CIA against IBM and a federal judge’s subsequent smackdown of IBM’s appeal was an alarm that sounded within the enterprise IT establishment.
All this has coincided with the realization within IT shops around the world that Philips’ CEO was correct-- there’s no competitive differentiation in running IT infrastructure-- with very few exceptions.
Web Scale Cloud Economics
Our belief is that very few organizations (on the buy side or sell side) will be able to compete with the scale economies of Amazon. Figure 2 depicts our model of the new enterprise IT economics driven by cloud:
The lines look the same as in Figure 1 with the exception of the dashed green line (formerly “Outsourcing”). We’ve called this cloud services (meaning to depict the infrastructure management component of outsourcing) and believe that Amazon has turned infrastructure provisioning and management into a business with a profile that now tracks software marginal economics. Like the lesson’s learned in the Microsoft era of volume software, we believe that scale will overwhelm the ability of the vast majority of IT organizations to provision infrastructure at a competitive price.
Amazon, we believe, understands this well, and it’s why the company is plowing operating cash into CAPEX – building out warehouses (to compete in retail) and AWS infrastructure, to track the green line as fast as possible. Benedict Evans of Andreesen Horowitz described Amazon’s strategy and its use of operating cash—about $5B in the last twelve months—in an excellent analysis recently. The premise we put forth above is that in order to compete, companies (both buy and sell side) either must have:
- Massive scale or,
- Deep vertical integration and a richly competitive stack.
Companies such as Google, Microsoft and Facebook should own their data centers and can compete with Amazon’s cost structure. Microsoft can leverage its massive installed base and continue to bundle cloud services with its enterprise software stack, creating a homogeneous experience for its many customers. Some large banks are known to be contemplating a massive scale strategy, talking to ODMs and component suppliers about building specialized hardware to support massive scale. But beyond this handful of players, who really has the volume to compete cost-wise with AWS?
Google certainly has the volume, and the company has been using aggressive pricing to get attention from reporters, most recently leading up to the AWS re:Invent conference. However, in the enterprise Google is far behind Amazon. Google can compete with AWS for consumer disk space, enterprise personal computing (Google Apps) and some areas of Big Data analytics. However, today Google has nowhere near the set of services, high availability options, track record and sales & support force necessary to compete with AWS. Without a major change in strategy and resource allocation, Google will be able to pick areas of focus that:
- Don't require any call to a real person,
- Don't require low latency, low cost data direct data sharing in mega data-centers, and
- Don't require professional services
The Integrated Stackers-- Who Holds the Best Hand?
Oracle and IBM own much of their own stack and will likely continue to command premium margins. While very similar, of the two, we find Oracle more intriguing for three reasons: Its integration between the database, SOA and Fusion middleware technology looks to be very tight; Its database has a much wider spread popularity, and it has integrated further up the “Red Stack” into the application layer. While IBM has DB2 and numerous SaaS applications in its portfolio, we believe Oracle’s integration is more complete. IBM’s offerings are historically more services-led and its cloud strategy is still evolving.
On paper, Oracle’s cloud offering is compelling, with a clean story around IaaS, PaaS, and SaaS. On its earnings calls, Oracle’s co-CEO Safra Catz is very clear on how much business is derived from each sector with go-forward guidance on the three cloud businesses within Oracle. IBM’s cloud guidance is nowhere nearly as crisp, which we believe is a symptom of its less evolved strategy.
This is by no means to say that Oracle has done all the integration work for cloud customers. Practitioners in the Wikibon community indicate to us that while Oracle’s core technology integration between its database and middleware is solid, Thomas Kurian, Oracle’s EVP of product development has left some work for customers to do when it comes to integrating applications. For example, both Taleo and Fusion HCM, in and of themselves, are well integrated into the Red Stack. However if clients want Taleo and Fusion to work together there’s additional effort, and what integration has been done is not ‘seamless.’
Nonetheless, Oracle is spending on cloud R&D and is willing to be transparent about how much of its business is rental and what percent of that is IaaS, PaaS, and SaaS – a very encouraging sign. Oracle doesn’t need to scale to Amazon levels because: It owns its own hardware, and it commands ridiculously high margins in its software business. While Amazon and Google bash each other in price wars, Oracle’s CEO Ellison can comfortably say that Oracle will match pricing on IaaS while maintaining its obscenely high margins further up the stack. This is a huge advantage for Oracle.
Companies like VMware and HP don’t have this full stack benefit. VMware has vCloud air and is trying to gain first-hand experience to entice its 4,000 cloud ecosystem partners to go with a homogenous VMware stack. In essence, VMware is using vCloud Air as a "petri dish" and a source of leverage to try to line up its ecosystem behind the strategy. Many partners don't like the strategy for obvious reasons.
For its part, HP with Helion is ramping its cloud business, and while it has some “up stack” functions with Vertica it has nowhere near Oracle’s software functionality or Amazon’s volume. It does however have Eucalyptus through a recent acquisition, which gives HP an interesting hybrid cloud/on-prem play. But much more progress needs to be made by HP in cloud. The company is betting on OpenStack and could emerge as a leader in the OpenStack community. As we've said in the past, OpenStack-based clouds are a viable alternative to AWS to the extent that developers can rapidly build out function. But today, many incompatible OpenStack clouds are in the market with lots of competing agendas, and the vision of homogeneity has not yet been realized.
Cisco is vertically integrating with UCS. However it has chosen to be an arms supplier to the cloud, much like Dell and EMC. It remains to be seen how successful such strategies will be long-term, but the market is so large and fragmented that managed service providers and colocation facilities will support ISVs globally by finding ways to differentiate from Amazon – either through proximity, hybrid cloud capabilities, custom efforts for enterprise customers and niche capabilities that Amazon chooses not to chase. Companies like EMC will sell to these organizations for quite some time. EMC in particular will continue to leverage VMware as a strategic asset. However Amazon is a fast-moving target and this large partner ecosystem is very diffuse. Also, it is difficult to see how the cost structure of these solutions will be able to match that of Amazon's.
There is much debate in the cloud community about the cost of renting vs. owning. Amazon claims cost superiority as do the hybrid cloud folks. On a like-to-like basis it would seem logical that owning (at some crossover point) is less expensive than renting because of the rental company's markups. However from an underlying cost structure perspective it's unlikely that owning IT infrastructure, for most companies, will yield the economics of Amazon's internal costs, leaving Amazon holding the best hand in the cloud card game and Oracle dealing a "stacked" deck.
On balance, the market is enormous (in the trillions), but both buyers and sellers must ensure their strategies align with future market directions. What does that mean? Every pundit on the planet is looking for the next disruption in technology. As we've discussed previously, we're observing a change in the way industries are transforming, and that's where we feel the disruptions will be greatest. Digital platforms are enabling phenomena like Stripe and Bitcoin in financial services, Uber and Google in transportation, Netflix in media, and so forth, to challenge the vertically integrated stack within the respective industries. The greatest value seems to be coming from those companies that can leverage digital technology to transform their industry value chains. Cloud is a key enabler but only as an infrastructure platform that allows innovation to ride on top of it. The new technology stack is traversing industries and encompasses cloud, social apps, a security and access control layer, and of course Big Data analytics. The ability to digitize one's business and leverage what author David Moschella calls the "Digital Fabric" is where observers should look for disruption.
From a technology economics standpoint, some wildcards in understanding new IT scenarios have not been addressed in this research. These include:
- Open source software and its impact on established players and proprietary clouds,
- A new economy based on exchanging technology services for advertising (Google), social engagement (Twitter, Facebook and LinkedIn) and opt-in models,
- Society’s willingness to provide information and technology services without compensation (e.g. Wikipedia, open source contributions, user generated content) in exchange for recognition.
Action Item: The cloud generally and Amazon in particular are changing the economics of enterprise IT. To compete, IT organizations must either have massive scale (unlikely for most) or have clear differentiable advantage through vertical integration. CIOs should plan to expend labor for managing IT infrastructure only in cases where it drives direct profit. Otherwise, technology organizations should focus resources on integration and adding differential value through analytics, new services and unlocking innovative digital business technology models.