When I talk to people currently employed by one of the major infrastructure outsourcing units in the major players like Atos, DXC, IBM, or T-Systems, it feels like summer closing. Some of them still embrace what they do, but most see some severe bad weather looming. They explain that the so-called hyperscalers, meaning providers like Microsoft Azure, Amazon Web Services, and Google Cloud are continually growing their businesses, while they are struggling to make ends meet. No compelling portfolio, operations revenue steadily declining, no investments in technology or people, and finally, no new projects are the themes, which are continuously mentioned. I want to dive deeper to see what is behind these feelings, if the financial numbers support these claims, and what my one can conclude for the years to come.
Infrastructure Outsourcing – What?
Before starting, I need to clarify which parts of the companies I mentioned before I want to talk about. As you know the portfolio of the mentioned companies is vast, they sell a variety of services/products like consulting services, software engineering, hardware, software licenses, and of course: infrastructure outsourcing. In all of the companies, infrastructure outsourcing is grouped into a single organizational unit. The table below lists the respective units and the overall share of the business.
|Company||Infrastructure Outsourcing Unit||Responsible Executive 2019||% of Company Revenue|
|Atos||Infrastructure and Data Management (IDM)||Eric Grall||54% based on H1 2019 revenue|
|IBM||Global Technology Services (GTS)||Juan Zufiria||35% based on Q1 2019 revenue|
|DXC||Global Infrastructure Services (GIS)||Eugene O’Callaghan||58% based on FY 2019|
|T-Systems||Managed Infrastructure Services and Private Cloud (MISPC)||Jörn Kellermann||Undisclosed, estimated at 35%|
The table shows that the Infrastructure outsourcing units (which I will call outsourcers from here on) contribute significantly to the overall revenue of their parents. Bear in mind that the outsourcers’ scope of services is not entirely comparable. T-Systems, for instance has a very fragmented organization, while DXC reports figures only separated into two different business segments. However, their success or failure does have an impact on their parent companies.
Infrastructure Outsourcing – A sound business model in the 1990ies
Let’s see why infrastructure outsourcing became a hot thing in the 1990ies and discuss the market conditions which gave way to its rise. For me, it boils down to five major factors, which gave rise to the infrastructure outsourcing market.
- Low cost to acquire technology skills. What I mean by that is a compound factor which describes the cost of acquiring a technology skill of as well as the resulting general availability of skilled employees. In the 1990ies, being, for example, an Oracle DBA was a long term career proposition. Training to acquire and maintain skills like Oracle was expensive and would only be paid by employers, not the employees. It was almost impossible (or at least hard) to install Oracle on a computer in your home to learn outside of certified training centers. The required basic infrastructure to run Oracle was expensive to procure. The strength of this factor certainly differs from technology to technology, but it holds on average.
- The labor part in infrastructure operation was significant. Think of infrastructure operations being the starting and stopping batch jobs, constantly monitoring them, manually sending the output somewhere, making and restoring backups, physically inserting, moving and labeling backup tapes. Of course, the experts made some scripts, ran some cron jobs, and optimized their workload. If someone was smart, he could make himself a comfortable living. However, even if someone did, compared to a cloud-based database service today, more time and effort was required to keep the lights on.
- Automation enabled economies of scale of labor. The magic, which makes the effort of running one additional instance of technology lower than the initial one. The 1990ies we were at the beginning of the automation curve. The initial benefits of automating installation-, maintenance-, or upgrade-jobs were high. If outsourcers invested some time and money in the scripts and low-level tools developed by their employees, the initial payoff was immense. A large amount of labor in each service made investing in automation profitable.
- Balance sheet cosmetics. Customers could change Capex to Opex via outsourcing, which made their balance sheet a bit more lightweight. Instead of buying expensive kit for their exemplary Oracle databases and having the kit sit as assets on the balance sheet, customers rented it for 60 months, including basic operations from the outsourcer.
- Finally, it was attractive to do workforce management (read: reduce staff) via outsourcing. Technology skills were rarer than today; customers had to pay huge to attract and keep talented employees. Over time, the paychecks of employees running services rose, and staff cost became an increasing burden for the CIOs. Outsourcing offered the solution as the outsourcers would more or less happily take over staff. It was often used to make expensive employees redundant almost without legal constraints.
Changes continuously eating into the business model
The difference between the 1990ies and now is that these conditions changed as summarized in the table below. First slightly and during the recent years quite radically. To cost to acquire IT skills are almost zero. Anybody with a laptop and internet access can learn nearly everything for free by using online MOOCs, books, YouTube videos, and communities. Virtually any product can be tried, tested, and learned using a virtual machine using an open-source hypervisor on 500 EUR/GBP/USD hardware. The task of operation is almost not existing anymore. Infrastructure technology has been optimized for stability and ease of upgrades. While in the past you had to wait for two years for the next major release, operators are now force-fed with iterative automatic updates every month. As it is almost effortless to run a piece of infrastructure, economies of scale of labor have become irrelevant. The evolutions of accounting standards prevented that outsourcing is used for balance sheet cosmetics. Recent changes like IFRS 16 make the difference between buying and renting irrelevant from a balance sheet perspective. Moreover, getting rid of expensive employees is not as easy now as it was before because of regulations like TUPE.
|Cost to acquire IT skills||High||Low|
|Labor part in infrastructure operations||High||Low|
|Economies of scale of the labor part in infrastructure operations||High||Medium|
|Financing benefits||Capex to Opex shift||All Opex|
|Workforce management potential||High||Low|
The market conditions, which gave rise to infrastructure outsourcing, turned against outsourcing over time. Also, one more thing happened: Serious competition. In 2006 Jeff Bezos announced Amazon Web Services AWS (https://aws.amazon.com/de/blogs/aws/we_build_muck_s/). Later on, Microsoft Azure and Google Cloud joined the cloud market and thereby changed the game for the existing outsourcers. They focused on the flexible charging, fast provisioning of services, high-level services such as databases or middleware components, and continuously evolving functionality. Since then, they are eating into the business model of outsourcers.
The strategic response: Copy & Paste
I am sure that the leaders of the respective organizations have very well recognized the changes in the market conditions and their strategic consequences which happened in the period between 2006 and 2019. Moreover, understanding did – with some delay – trigger a strategic response.
In 2013 IBM decided to acquired Softlayer which at the time was more or less a US-centered Web Hosting Company. Softlayer was integrated into IBM GTS, led at the time by Erich Clementi, a long-time IBM manager. Reflecting the press release the understanding at that time was that cloud, in essence, meant faster provisioning times and a more flexible billing model. In 2014 IBM followed up by announcing their Platform as a Service offering called Bluemix.
Bluemix was based on CloudFoundry and was marketed to the software engineering community: “ BlueMix provides DevOps in the cloud — an open, integrated development experience that scales […] [and] help developers, independent firms and enterprise teams get started to build enterprise applications more quickly and effectively.”. IBM knew all the relevant marketing buzzwords already back then.
What was striking (and in my humble opinion was already a precondition of the upcoming failure) is that because of marketing towards different markets Bluemix sat below the IBM Software Group (SG), led by Robert Le Blanc and Softlayer sat below GTS. In 2015 Ginni Rometti corrected that and moved the business to a new unit called IBM Cloud, directly reporting to her and led by Robert Le Blanc. In January 2017, the IBM veteran “was retired” by Rometti in a restructuring exercise after 19 quarters of declining revenue. In October 2017 Bluemix and Softlayer were discontinued as product names and everything cloud was consolidated once again into “IBM Cloud” and moved below the SG.
In 2018, despite some the serious multi-billion investments, the market share of IBM cloud sat at only 3,6%. It can be firmly stated is that IBM’s strategic response to declining outsourcing revenue was a profound failure.
Astonishingly Atos moved earlier than IBM. In February 2012 Atos founded Canopy as a separate legal entity with EMC and VMware as partners. Back then, it was rumored that Atos had a fear of making large investments which would sit on their balance sheet. So despite the rumored 50mn. EUR which Atos put into Canopy is looking small in comparison to IBM’s acquisition of Softlayer; it was a significant decision Atos. After founding Jacques Pommeraud led Canopy until it was integrated into a unit called Cloud and Enterprise software during 2014 and reporting to the veteran manager Ursula Morgenstern. Finally, Canopy ended up in IDM. It seems that the limbo Atos did with Canopy was more or less comparable to IBM. At first, Atos believed Cloud was as pure infrastructure play. Afterward, they realized that it was also about Platform as a Service or in Atos case software as a Software as a Service (Yunano which was positioned against SalesForce, and BlueKiwi positioned as enterprise social network).
Finally, German juggernaut T-Systems woke up announced their cloud product called Open Telekom Cloud (OTC) in 2016 based on OpenStack and Huawei. While T-Systems marketing still bangs the drum for OTC, its customer references do not highlight large-scale signings. From an organizational perspective, OTC sat in the division led by Jörn Kellermann when the product was launched. In the recent reorganization went to what T-Systems calls Portfolio Unit Public Cloud.
Summarizing the brief history of of the outsourcers attempts to out invest the hyperscalers it is evident that it did not work for any of them, regardless of the money spent. Interestingly DXC (respectively HPE and CSC) missed this exercise. It seems that because of the turmoil of the merger, they did not invest in their cloud venture, but focused on integrating the two companies instead.
Indeed, all these units did set additional operational measures to protect their installed base. There have been rigorous cost-cutting efforts, near-and offshoring activities for existing accounts, and more or less smart plans for upselling other services to their existing accounts. Coming back to what I hear from employees, it does not work out completely.
While busy trying to copy&paste the services of the hyperscalers the market shifted even more. The revenue of the existing projects has been declining as infrastructure outsourcing projects moved into their 3rd or 4th generation. What I mean by that that with each shift of a customers contract moving from one vendor to another, the degree of standardization rose significantly. The service which at the begin was very customized is in the fourth generation a pure commodity by now. The CFOs of the outsourcers can see their revenue shrinking month by month. The only thing which prevents that, are the outsourcers best friends: the change request and volume increase.
The total size of the market is stagnant – at best. Regardless of customers consolidating or splitting up all companies which either needed, could, or wanted to outsource already did.
Finally, software engineering projects starting from scratch are developed cloud-first or use cloud-ready technology. Nobody (at least I hope so) would nowadays begin to developing his next-generation e-commerce service running using J2EE with an Oracle backend.
Does it show in the numbers?
One might argue that if things look so grim, it should show in the numbers of outsourcers. Unfortunately, the division reporting is sketchy. Revenue can be moved between the units; M&A transactions change the numbers up and down significantly. I do not state that the numbers are fabricated, but they can be nudged a little to support the messages of senior management. Nevertheless, let’s have a quick look at what we have.
IBM reports on quarterly basis separating its units and therefore showing GTS figures separately. Atos, on the other hand, reports half-year figures and due to the M&A transactions likes to restate IDM revenue every year. I hypothesize that this restatement helps to show flattish revenues and keep investors unconcerned. Looking at DXC does not add value as their figures for 2017 and 2018 show only the CSC part of the company. Only the figures for fiscal 2019 show combined values of CSC and HPES businesses. T-Systems does not report publicly at all.
So the diagram below shows the revenue (index 100) development since 2016 only for GTS and IDM. The figures for 2019 are extrapolated linearly, meaning they are H1 figures times two. This is an approximation as usually, the contract negotiation dynamic will heat up just before year-end. However, even this approximation does not change the overall picture of slightly shrinking revenues over the last three years. Despite having tried to do cloud, shift labor to nearshore and offshore locations, and various optimization and sales initiatives, the decline does not stop.
Reading through the DXC annual report for FY2019 I stumbled upon this statement which summarizes this development quite crisp: “The decrease in GIS revenue in fiscal 2019 reflects the ongoing migration out of legacy infrastructure environments, partially offset by growth in our cloud infrastructure and digital workplace offerings.”. In this sentence “GIS” can simply be interchanged with IDM or GTS.
So contrary to the more qualitative aspect of feedback from talking and working with the employees of the outsourcers, who say that the ship is already sinking, the decline in the numbers is very modest. The question now which argument, the qualitative or the quantitative reflect the reality more accurately.
It is not apparent what will be happening with GTS, IDM, MIS, and GIS in the mid-term future until 2025 or 2030. The outsourcing market shrinks while the cloud market is growing; this is quite obvious. The strategic moves completed by IBM, Atos, and T-Systems to protect their market by copying the offerings of the hyperscalers did not work. I also do not think that they will try once more to “out-invest” or “out-market” Microsoft, Google, or Amazon.
So will the divisions be done and dusted by 2030? Well, most certainly not. If I had to predict, then I would argue that they will stay but shrink considerably. New software development will be cloud-first, maybe the SAP S/4 migration will take more customers to the SAP cloud and away from the outsourcers. However, some customers, some businesses, and some technology (Hello Mainframe & z/OS) have a strong tendency of inertia. Somebody needs to run these systems, organize the batch jobs, update and patch the operating systems, and those mentioned companies have proven that they can do that.
I hypothesize that the leaders of these organizations have already recognized this and treat the outsourcers as cash cows, which should generate cash to enable alternative strategic moves. In case of IBM and GTS this might be the RedHat acquisition, in case of Atos and IDM it might be the hardware investment in Bull or the recent collaboration with Google, and finally for DXC and GIS it might be the acquisition of Luxsoft, which brings them closer to consulting and system integration companies like Accenture or Cap Gemini. T-Systems, as stated previously, is on the market and will disappear as a separate entity altogether.
On the operational side, the unit leaders will go into end-of-life management mode for their respective divisions. That means further reducing onshore staff, automating and moving the workload to offshore locations to squeeze the final USD/EUR out of them. For people working in these units, my feeling is that the bad weather (with all the nasty consequences) will be the new normal for everyone working there.