Cloud Software: A Long Journey with Three Layers
SEE LAST PAGE OF THIS REPORT Paul Sagawa / Tejas Raut Dessai
FOR IMPORTANT DISCLOSURES 203.901.1633/.901.1634
psagawa@ / firstname.lastname@example.org
November 18, 2018
Cloud Software: A Long Journey with Three Layers
The exodus to the cloud will take decades. Most enterprises will have to manage a hybrid blend of local and cloud resources for many years, wary of locking in to a single cloud provider. This creates opportunity and competition at three levels of the software stack. Infrastructure – Scale economies are prohibitive, with MSFT, AMZN and GOOGL likely to dominate but facing possible commoditization. We expect them to lever their strength to move up the stack in pursuit of margin. Application – A crowd of best-of-breed SaaS companies (e.g. MSFT, CRM, ADBE, etc.) with full stack solutions are displacing packaged software and custom applications. We expect continued growth and consolidation. Platform – The middle layers comprise SW categories like OS, Databases, and virtualization. Cloud leaders (i.e. MSFT, AMZN and GOOGL) looking to upsell into value added categories with potential to lock in infrastructure business will vie with traditional IT stalwarts (e.g. IBM, VMW, ORCL, etc.) offering discrete solutions built to control performance and cost in multi-application, multi-cloud environments. Together, the three categories address a nearly $4T TAM, with infrastructure the largest but likely least profitable, applications second largest and the fastest growing, but platform the most profitable. We see winners in each category: MSFT, GOOGL, and AMZN in infrastructure; MSFT, CRM, ADBE, WDAY, and others in applications; and IBM, VMW, MSFT and GOOGL in platform software. We are skeptical of many other traditional SW players, such as ORCL, SAP, HPE and others, as cloud players.
- The cloud has three arenas of competition. The advantages of the cloud are profound, but the transition costs are substantial, insuring a long migration that could take decades for many enterprises. In this context, the rising cloud market has developed into three distinct arenas of competition: Infrastructure is a scale game, dominated by AMZN, MSFT and GOOGL. Applications are measured by innovation and performance with many small SaaS competitors vying to by best-of-breed. Platform solutions must tie together messy collections of hardware, software and data, in the cloud and on premises, to give enterprises control over cost and performance. Here competition is just taking form.
- Cloud infrastructure: huge, concentrated, undifferentiated. AMZN dominates IaaS in western markets, due to early entry and leadership with cloud-native businesses. MSFT is half the size of AWS, but growing at twice the pace, levering its strong sales relationships as demand shifts toward traditional enterprises. GOOGL has unique offerings but has struggled to reach enterprise customers. It seeks to address this through its increasingly cozy relationship with CRM. Huge scale economies, high costs of maintaining performance excellence and leverage from internal franchises are substantial advantages for this big three. Going forward, we expect them to concentrate even more share but to face price pressure as new technologies facilitate price arbitrage for basic compute and storage demand. The TAM in this category should be well over $1T, albeit with a strong deflationary effect.
- Cloud applications: rapid growth, narrowly focused, due for consolidation. SaaS application leaders have seen extraordinary growth from their combination of enhanced functionality, superior performance and significant all-in cost advantages vs. packaged SW alternatives. The market is fragmented with a few large competitors with dominant SaaS franchises (i.e. MSFT, CRM, ADBE), a few traditional SW vendors looking to bridge to a cloud future with questionable success, and a flotilla of fast-growing companies vying for best-of-breed status for horizontal and vertical enterprise apps. SaaS solutions not only displace spending on on-premises applications, but also bundled the associated infrastructure and platform functions as well. As such the TAM approaches $1T
- Cloud platform: heterogenous, independent. The platform layers include a lot of functionality – operating systems, run-time, databases, virtualization, containerization, APIs, development rubrics, etc. Within this are all the tools enterprise IT managers use to allocate resources, control costs, monitor performance, develop custom SW, etc. For larger organizations with a heterogenous infrastructure mix, disparate applications to support, and challenges to keep it all under control and within budget, platform software is immensely important. At the same time, most enterprises are wary of solutions that might lock them to a single infrastructure vendor or a proprietary platform architecture. This $500B TAM market is in flux with many technical approaches and no player dominant.
- Infrastructure players look to move up. Facing commoditization, the big three hosts are all looking up the stack. This is easy for MSFT with its position in traditional datacenters. It offers tools for managing applications that run at home and on the web, but its approach is less robust for a multi-cloud approach. GOOGL has very strong tools, particularly for AI development and containerization, and has a strong open source philosophy, but it lacks reach to enterprise accounts. A tight relationship with CRM may help it gain better traction. AMZN has leaned on partners (e.g. CSCO, VMW, etc.) to offer platform level services and has not yet demonstrated a differentiated vision for would-be customers.
- SaaS players need infrastructure strategy, breadth and scale. Scale economies in the infrastructure and platform layers that are integrated into SaaS applications are a future vulnerability for the focused and modestly sized companies competing for best of breed status in what is still a highly fragmented business. We expect consolidation, with larger companies (MSFT, CRM, ADBE, WDAY, etc.) absorbing smaller ones, and smaller companies combining forces. SaaS leaders could also be interesting targets for infrastructure players (e.g. GOOGL, AMZN) who could use their reach into enterprise customers and rely on their application franchises as anchor tenants for their datacenters.
- Platform software must deliver simplicity and control. While smaller shops may opt for integrated platform solutions from infrastructure hosts, most enterprises are likely to opt for a multi-cloud approach and independent platform solutions that can allow them the control to shift loads to the optimal datacenter. We also see strong demand for IT consulting services to help define and implement platform strategies for messy, heterogenous architectures. We believe IBM’s acquisition of RHT and VMW’s deal for hot container startup Heptio help position them as leaders.
- Winners and losers. There are many companies positioned to prosper from the long shift to the cloud. The big three infrastructure players (i.e. MSFT, GOOGL, AMZN), the best SaaS vendors (e.g. MSFT, CRM, ADBE, WDAY, NOW, ZEN, etc.), and top platform suppliers (i.e. IBM, VMW, MSFT, GOOGL) should prosper. We are concerned for traditional SW players playing defense for non-cloud franchises (e.g. ORCL, SAP, etc.).
Send in the Clouds
The largest and most advanced public hyperscale datacenters have enormous cost and performance advantages over private datacenters. Inevitably, the cloud will become the predominant platform for enterprise computing, but the shift will take many years as IT departments must deal with the ramifications for their mission critical systems and for their organizations. The choices enterprises make will depend on the complexity each faces and the impetus to change.
For software companies, we see three distinct arenas of competition corresponding to layers of the “stack” model of computing. In the Infrastructure arena, scale economies, talent scarcity, and first mover advantages see AMZN, GOOGL and MSFT with increasing dominance over other would-be rivals. Still, we believe emerging technologies will reduce the switching costs amongst the three, yielding growing price pressures. The Applications arena – where cloud-hosted solutions sold as a service are growing rapidly at the expense of traditional packaged software – features many dozens of established developers vying for best-of-breed status in focused horizontal and vertical applications. We expect consolidation: from large SaaS operators (i.e. MSFT, CRM, ADBE) buying niche players to broaden their offerings, from smaller players combining to gain scale, and from platform players using M&A to move up the stack in pursuit of customer lock-in.
The Platform arena comprises the software layers that facilitate developing, delivering and supporting applications for an enterprise – e.g. operating systems, run-time software, databases, virtualization, containerization, APIs, development rubrics, and other middleware tools. For many enterprises, the solution must cope with an awkward mix of software systems – some old and others new, some developed in-house and others bought, some running locally and others in the cloud – across a mélange of infrastructure. Above all, platform software is the instrument for IT to manage the performance and cost of the many applications used in the organization. In this context, the infrastructure leaders are looking to move up, offering simplicity in exchange for commitment, while others (e.g. IBM, VMW, etc.) sell flexibility and independence. Both approaches will gain traction, even as SaaS continues to make inroads, as on-premises applications are adapted to run in the cloud, and as the platform challenge gets messier.
In the long run, we believe that enterprises will dramatically reduce their in-house datacenter resources, depending more and more on public cloud infrastructure. Containerization will allow IT managers to easily parse computing across multiple datacenters, squeezing margins for the bottom layers of the stack. Open source-based platform solutions will predominate, as expensive, proprietary products slowly fade, but customization and support will rise in importance. Competition here will be a mix of the big cloud hosts and a few successful independent players. SaaS applications will dominate for common horizontal business functions and for large vertical market solutions. Custom applications are important differentiators for enterprises, but they will also run in the cloud, supported by the platform layers.
This addresses nearly $4T in all-in global spending on private datacenters. Cloud solutions are highly deflationary, but the ultimate opportunity is still huge. Infrastructure will be the biggest piece of it, although pricing pressure may make it the least lucrative. SaaS solutions, which bundle infrastructure and platform layers, will consolidate and move to scale hosting platforms, either by partnership or acquisition. The best will be very profitable. The platform market will be messy for a long time. Customized implementation and reliable support will be critical components for the stickiest, and perhaps, most profitable category.
Isn’t It Rich?
Collectively, enterprises worldwide spend more than $3.0T each year building and maintaining private datacenters (Exhibit 1) not including spending on personnel, electricity, real estate and other operating items, which likely add more than $1T more. Hardware and software bought from IT vendors is less than a third of the spending, the rest going for personnel, consulting, power, real estate and other operating costs required to develop, deploy and maintain systems and to support users. The hyperscale datacenters at the core of cloud-based services yield as much as 90% cost savings vs. private datacenters for basic computing and storage tasks though better efficiencies, scale economies and high utilization.
Exh 1: Global Enterprise Data Center Spending Forecast, 2017 – 2022E
Cost savings are the central draw of the cloud, but there are other advantages (Exhibit 2, 3). Software updates – historically sold to customers at multi-year intervals and carrying significant implementation costs – come frequently, automatically and are included in the subscription fee. Security, disaster backup, and redundancy, all provided with cutting edge performance as a part of the service. Cloud services can also be dialed up at a moment’s notice, adding spare capacity at peak times without paying for it during slack periods. Cloud users have access to specialized hardware, like GPUs and FPGAs for training AI models, software tools, like development rubrics and analytics engines, and pre-programmed APIs, like image recognition or language translation models. Employees can access company systems via a network of datacenter locations, reducing networking costs and latency. Cloud partners can help manage thorny data management issues, like the EU’s GDPR privacy laws. While there are a few industry and application areas
Exh 2: Basic On-Premise vs. Cloud Hosting Cost Comparison
where regulation and strict latency needs demand private, local computing, almost all enterprise applications would run better and more cheaply in the cloud.
The biggest thing stopping an immediate mass exodus are the switching costs. Custom software must often be rewritten to accommodate the new cloud environment. New applications must be tested, and users trained, yielding downtime and inefficiency until the organization gets up to speed. Many IT managers still remember the chaos catalyzed implementing ERP systems two decades ago, and the pain of answering to a CFO when the financial books couldn’t be closed on time. Moving to the cloud is a bit easier than to a new on premises solution, since the new system can easily run in parallel with the old one without a hard cutover. The second major obstacle is the threat to the IT organization itself. Achieving savings from a move to the cloud will require reducing headcount and ceding many responsibilities to 3rdparties. CIOs may be reluctant to recommend downsizing their own department and to rely on outsiders for security and performance.
Still, there are those savings. Ostensibly, that $4T+ addressable market ought to shrink as enterprises take advantage of lower cost cloud solutions. This would still be an enormous opportunity for a cloud software industry that is still less than $175B (Exhibit 4, 5). However, it may be true that the low cost and easy
Exh 3: Other advantages beyond cost savings driving adoption of Cloud
Exh 4: Current Public Cloud Spending Estimate by Category, 2017 – 2021F
Exh 5: SSR Forecast of Total Addressable Market for Cloud Services by Category
availability of cloud-based solutions will find new markets to partly offset the obvious deflationary effect of the cloud.
Layers Upon Layers
The taxonomy of software divides various types of programs into functional layers within a “stack” that rises from the machine level commands that instruct the computing infrastructure at the bottom, and user applications at the top. In between are the layers of code that interpret requests from user programs, parse jobs to processors, keep track of stored data, manage networking, control security, accommodate development, support applications, and other functions needed to keep enterprises computer systems working.
Cloud computing takes responsibility for some or all these layers away from privately run computers (Exhibit 6). At the base level, Infrastructure-as-a-Service (IaaS) hosts provide virtual machines upon which customers can load software, as though they were physical systems in their own datacenter. At the highest level, Software-as-a-Service (SaaS) solutions bundle the whole stack, allowing the users to interface to the applications directly over the internet, bypassing the local enterprise datacenter entirely. In between, Platform-as-a-Service (PaaS) offerings run a customer’s applications providing all or some of the support software – operating systems, runtime systems, networking, development tools, database management, containerization (a system for breaking computing tasks into small executable pieces that can be parsed to different processors and then reassembled), application programming interfaces (API) used to link applications to pre-programmed utilities, and other functions (Exhibit 7, 8).
Exh 6: Public Cloud Services offered at various layers of Software Stack
Hybrid Cloud to Multi-Cloud
The early growth of cloud computing was largely an either/or proposition. SaaS application vendors, like Salesforce or Workday, sold access to hosted applications over the internet that would replace software running on a local computer that provided similar functionality. New web-based businesses, like Snap or Spotify, didn’t bother to build their own datacenters – they relied almost entirely on capacity bought from IaaS partners. Development organizations within larger enterprises would use cloud services as slack capacity, buying computing and storage on an ad hoc basis for discrete projects.
The industry is now beginning to address the major obstacles that have, thus far, kept most enterprises from moving their most important systems to the cloud. Most of this affects the functions comprised by the platform layers. Enterprise systems tend to be intricately intertwined, yet organizations want to transition piece by piece to minimize switching costs and disruption and to optimize sunk investment in existing datacenter capacity. Doing so requires “hybrid” approaches that can manage solutions where some computing is funneled to the cloud, but some is also completed on-site. Ideally, these approaches give IT
Exh 7: High level Outlook of The Modern Technology Stack
Exh 8: Future of Computing Needs Will Evolve in 3 Categories
departments powerful tools to manage costs, performance, security and other important system metrics blended across both computing environments.
The shift to a predominantly cloud-hosted computing solution will take years, if not decades for most traditional enterprises to accomplish. We note that IBM still sells more than $8B per year in hardware and software to customers still running on their mainframes, an architecture whose first iteration was introduced more than 50 years ago (Exhibit 9). Meanwhile, CIOs will carefully assess what vendors to trust to help them deploy and manage this messy, hybrid environment. In this context, enterprises are loath to lock themselves in to a single vendor for an important layer of their computing stack. For most, 25 years of Oracle wringing audit dollars from database contracts without escape or recourse is a painful history. For that alone, enterprise will want to keep their cloud options as open as possible.
Exh 9: IBM Legacy Systems Segment Historical Quarterly Revenue, 1Q15 – 3Q18
This means a multi-cloud approach, with more than one hosting partner, with containerization that facilitates shifting loads and storage from one operator to another as conditions and prices dictate (Exhibit 10). It will not be practical to do this with every application – SaaS apps bundle the platform and infrastructure layers, and other systems may use differentiated hardware, software tools or APIs that necessitate committing to a single host – but the less “lock in” by vendors, the better for customers. This will begin to commoditize the Infrastructure layers. Within the platform layers, enterprises are favoring “open source” solutions that are available in standard form from the infrastructure hosts and from independent software vendors, and that can run on a hosted system or in a private datacenter. The
Exh 10: Snapshot of major computing eras and a Multi-Cloud future
differentiation will come from the ease of implementation, support services, proprietary features layered atop the open source stack, and from the “dashboard” applications offered to give control to IT managers.
Infrastructure: The Big Three (or Four?)
Infrastructure is a scale game. The modern hyperscale datacenter architecture was invented by Google, which needed it to build systems big enough to index the entire web and which donated its innovations to the open source community on the premise that helping internet businesses break through the scaling limitations imposed by the then-standard technologies would be good for business. This proved true and then some. In 2006, Amazon took its Google inspired datacenter infrastructure and began to sell computing capacity to 3rd parties, offering low costs, flexibility and convenience. Fledgling web-based businesses and software developers flocked to the platform. It took Microsoft and Google four years to jump in the business behind AWS, which had locked in marquee customers like Netflix, Airbnb, Yelp, Slack, LinkedIn, Expedia and many others.
There are many cost and performance advantages to scale in the infrastructure game – volume purchasing, leverage on expensive talent, better capacity utilization, demand to justify more esoteric hardware, etc. – and the top players can lever the needs of their own massive consumer franchises as well. Only the company like Google could justify developing an Application Specific Integrated Circuit (ASIC) chip like its TPU just to support its own proposed standard rubric for developing AI programs. Amazon, Microsoft and Google can build more and bigger datacenters than can smaller would-be competitors without those consumer services. For these reasons, most observers believe that the big three, Amazon, Microsoft and Alphabet, will dominate the future of IaaS. But in what order?
Amazon reports its AWS cloud hosting operation as a separate business unit. For the trailing 4 quarters, AWS sales were $23.3B, with an average YoY growth rate of 47.0% (Exhibit 11). These revenues are overwhelmingly IaaS. Microsoft does not break out revenues for its Azure hosting business, combining it with its huge Office 365 SaaS franchise and several other platform and application products for a single “cloud” number. The only pure Azure number reported is its growth rate, which has been more than 90% over the past 4 quarters (Exhibit 12). Alphabet provides no consistent data at all for its Google Cloud Platform (GCP) IaaS product.
IBM reports the revenues generated from software sold “As a Service”, which includes its SoftLayer hosting business along with a range of platform and application software. These businesses generated $10.9B in sales over the last 4 quarters, growing at 22.9% (Exhibit 13).
Industry analysts have published takes on what all of this means for market share. Generally, the consensus is that Amazon is holding serve at about a third of the market; that Microsoft is quickly approaching 15% on a strong growth trajectory; that Google is still less than 10% share but gaining; and that IBM is slightly ahead of Google but slowly leaking market share (Exhibit 14, 15). On a global basis, Alibaba is a comer from 5th place, but its strength is concentrated in China and its sphere of influence. The typical forecast seems to see AWS holding its lead into the future, Azure’s share gains slowing, GCP gaining a little steam, and IBM falling to a distant 4th place.
Exh 11: AWS Historical Quarterly Revenue, 1Q15 – 3Q18
We have some quibbles. Amazon’s lead is substantial, built on web-based businesses and software developers both within and without traditional IT organizations. However, we believe that the next 10 years of cloud growth will be driven more by the migration of traditional IT. Here AWS is less advantaged. It has a modest sales reach and does not have differentiated platform layer services or any applications to sell. Its
Exh 12: MSFT Azure Cloud Historical Quarterly Growth Rates, 1Q17 – 3Q18
Exh 13: IBM Quarterly As-A-Service Estimated Revenue, 1Q16 – 3Q18
Exh 14: Workload Distribution for IaaS Cloud Service Provider, July 2018
relentless focus on providing the services that its customers ask for is less powerful in markets, like the hybrid cloud, where customers do not know what they need or want. Amazon needs a vision for how enterprises should navigate the 10-yr transition to the cloud; it needs unique products that help customers achieve that vision; and it needs a sales organization positioned with enterprises all over the world to sell that vision and those products. Amazon also needs to establish differentiated services up the stack to preserve its margins, lest its IaaS franchise become commoditized through containerization. This may be difficult to achieve organically, leading us to speculate whether Amazon will make an enterprise software acquisition to beef up its platform bona fides and customer reach.
In contrast, Microsoft can leverage sales relationships with nearly every enterprise on the planet (Exhibit 16). As it leads with Office 365, the SaaS version of its ubiquitous productivity software suite, it transitions to Azure. It is one of the biggest vendors of datacenter platform layer software and has well designed PaaS products and hybrid approaches to help traditional enterprises navigate the transition to the cloud. We believe that these organizations are now driving the growth in cloud hosting, and that Microsoft is unusually well suited to serve them. We believe that Azure, which has been growing sales twice as fast as AWS, can ride this to surpass its larger rival within 5-6 years.
Exh 15: Cloud Infrastructure Services Market Share, Q3 2018
Alphabet is a bit of a conundrum. On one hand, it is the smallest of the big three, the last to market, and the least well positioned with enterprise accounts. On the other hand, its engineering prowess is legendary, its innovations at the platform layer groundbreaking, and its leadership in supporting cutting edge tech like AI and containers is extraordinary. It is also able to bundle its IaaS with platform services (e.g. AI APIs, analytics tools, etc.) and with its popular GSuite productivity applications. Still, it is still a laggard, even under the hand of industry veteran Diane Greene, the VMWare founder credited with giving Google’s enterprise operation needed focus. One possible solution to the problem Alphabet’s increasingly close relationship with Salesforce, which could provide the sales impetus needed to move the needle. We note that Google also offers a lot to Salesforce, which will likely need to pull the plug on its own datacenter business.
Exh 16: SSR Scoring for Cloud Infrastructure Services Providers on Key Factors