Warren Buffett and IBM



I have been bearish toward IBM for years. However, throughout 2016, Berkshire Hathaway kept remaining as the largest shareholder of IBM. I have a deep respect for Warren Buffet (as I touched on high conviction investing in this article) and IBM is not my favorite. It made me ask questions myself and naturally forced me to look into what presumably made Warren Buffett convinced about IBM.

In the interview last year, he said, “And then we went around to all of our companies to see how their IT departments functioned and why they made the decisions they made. And I just came away with a different view of the position that IBM holds within IT departments and why they hold it and the stickiness and a whole bunch of things.”

It looks the stickiness of IBM’s business was the critical reason WB got convinced. I guess it was IBM’s mainframe business because it has a reasonable backdrop, as an individual product, inside of IBM.

Other products IBM has do not look sticky due to product life cycle or competitive landscape. The WebSphere middleware of on-premises, for which IBM is pretty strong, is getting challenged by the migration toward the cloud, where on-premises middleware such WebSphere is not used. GBS and GTS are the No2 business in the industry, not competitive and growing fast enough. It does not have the application software that will remain in the client system of cloud-based IT. I do not think those products & services can pass the analysis of WB and be viewed as a sticky & sustainable business.

Then, how has IBM’s mainframe business been? Why was WB confident that it would be sticky and profitable?

Although the mainframe is neither an innovative product nor a product that delivers new value-added, it is still widely adopted, particularly among the financial service companies and the government. There are specific types of applications that require mainframe, and the typical ones are mission-critical applications. They are highly reliable, zero-downtime, and secure apps for high-volume & centralized transactions.

Can the mainframe get replaced with Apache servers on-premises or private cloud? Can even a public cloud service be a feasible alternative?

The short answer is “not impossible but extremely difficult.”

The biggest reason is not only the reliability & security of the system but also the difficulty of rewriting codes for Linux servers while keeping backward compatibility., i.e., migrating applications off the mainframe.

Why is it challenging to rewriting apps? It is COBOL that is 60 years old software still running on the mainframe, and it requires both COBOL expertise as well as Java or C++ expertise for the migration team, and this it is quite costly to rewrite it.

The application COBOL codes on the mainframe are usually very complicated after the decades of patches and updates. It is tough to correctly analyze huge lines of COBOL codes and then translate them with another language on Apache server.

Additionally, there are minimal numbers of programmers with COBOL expertise, which is has been and is in massive shortage. It is the least popular programming language because it does not feel intuitive but looks ancient for programmers. It is not a choice for young and talented programmers, and Java, C++, and Python are the most popular choice. Maybe some wonder why shortage made more people learn COBOL. Yes, demand-supply would make it so on theory, but it is a decades-old programming language platform. It is almost like becoming a specialist in IT programming language archaeology. Even those who used to handle it have not used it for so long. Young IT professionals would not pursue a career to become a curator of museum pieces of dinosaur-like software.

Legacy mainframe computer systems also depend on Assembler and VSAM files, which are a part of the outdated IT technology with similar issues. Thus, it is very challenging to build a team of COBOL software engineers for migration and execute the plan.

Say, for instance, what about giving up the entire legacy IT asset of the mainframe system and starting from scratch and writing a completely new program in Java?

It is a feasible option, but this decision requires substantial new IT investment. In fact, it is known that such a plan has not been successful in the past, often ending up with hundreds of millions of dollars without achieving the goal.

Aside from the technical issues, another challenge is the return on IT investment on the migration off mainframe project.

It is costly due to technical and skill resource shortage reasons, and it does not generate much return, such as cost savings, agility, scalability, or mobility. Applications on the mission-critical system often prioritize very high reliability of undisrupted operation and backward compatibility. They do not need something new, and to guarantee “the same as before” is often the goal of development. However, if you end up with just the same function from the migration with updated IT infrastructure, there are no incremental benefits. It isn’t easy to justify the significant IT spending for the migration reasonably.

I guess WB’s GEICO is a loyal user of the mainframe system, and it could be a source of ideas that built the thesis of “stickiness,” as touched here.

So, is the investment case for IBM good and sound? It seems the mainframe will remain sticky and stable, although it is getting a small niche. The entry barrier is not low. If not, what are the potential flaws?


Perceive IBM as a Whole

What is the essence of IBM as a whole? How can its business model be described?

One anecdote you can sense from IBM’s annual report is IBM’s Global Financing segment. At first, you might think that it is a small and complemental business segment, but it is not. It recorded a substantial pre-tax profit of over $2bn in 2015, close to Global Business Service. In the recent 10-K, IBM described Global Financing as follows.

Global Financing facilitates IBM clients’ acquisition of information technology systems, software and services by providing financing solutions in the areas where the company has the expertise. The financing arrangements are predominantly for products or services that are critical to the end users’ business operations. These financing contracts are entered into after a comprehensive credit evaluation and are secured by legal contracts. As a captive financier, Global Financing has the benefit of both deep knowledge of its client base and a clear insight into the products and services financed. These factors allow the business to effectively manage two of the major risks, credit and residual value, associated with financing while generating strong returns on equity. Global Financing also maintains a long-term partnership with the companies’ clients through various stages of IT asset life cycle—from initial purchase and technology upgrades to asset disposition decisions.

Global Financing is a financial solution for clients to own and upgrade IT assets. It is also a profit center for IBM to deal with used IT assets.

What ideas did I get? I would say this tells the shape of IBM’s business model built on an on-premises IT system ownership model, which designates a hardware system as its strategic point of origin.

It is strategic in the sense that the proprietary hardware becomes a box to which IBM can attach various add-ons to boost business results. The add-ons include middleware (WebSphere), application software licenses (Cognos & Watson), Global Business Service, Global Technology Service, and Global Financing.

It might be too much to call it a master-servant relationship, but there seems to be a hierarchical relationship between product and services. At IBM, the history and the legacy make “hardware” at the top of the hierarchy.



For companies with multiple lines of businesses and services, how far the totality optimization is from the aggregate of individual optimization is critical for the leadership. It is a qualitative and vital process to consider how the leadership team can take the helm of each division that tends to have interests in different directions. A direction that works the best for a product group is only a decent choice for another. It would be fantastic if there are no strategic deviations forever and at all. In reality, every business is neither a static activity nor an action without various mutual influences.

Thus structurally, it is challenging to sort out both business performance evaluation and financial resource allocation at the company with extensive product and service portfolio.

Quite often, it is legacy, internal politics, precedent, and invisible corporate culture that force business decisions to get biased, loose balance, and fail.

In the case of IBM, the thread of strategy is woven around the hardware (mission-critical) system at the center. But is it the No 1 choice for all other groups? If not, is the business portfolio structurally a collective entity of the 2nd or the 3rd choice strategy at an individual level?

I think that has to come up in analyzing the holistic stickiness of business, which is different from the individual product/service level stickiness.


Reality and Impression | Cognitive Solutions

Lastly, there seem to be too favorable views on Cognitive solutions in the financial market. Within Cognitive solutions, Cognos Business Intelligence is a leading product as a business intelligence suite software. It competes with other BI software such as legacy ones such as Business Objects under SAP and SAS and others such as Tableau, Microstrategy, Power BI of Microsoft. The competitive landscape is getting more influenced by emerging companies and new technology.

However, IBM has put more emphasis on Watson Analytics when presenting a strategy for big data analysis. Watson Analytics is not an entirely new system, and it is not developed utilizing new data processing technology. That makes me feel that there might be a gap of perception as to how crucial Cognitive solution still is.

Anecdotally, the new version development cycle of Cognos Business Intelligence has recently been extended, and the integration plan with Watson Analytics is uncertain. New BI competitors are developing and improving products quickly on a new architecture.


IBM in Ten Years

Although the concept of cloud computing has already become widely known, most of the real impacts have not yet been played out. Infrastructure is sticky, especially for the one already in place. But greenfield investment will continue to be heavily focusing on the cloud. A new generation of companies is getting more business from the incumbents, which will gradually but inevitably force the incumbents to shift IT infrastructure to the cloud and make applications cloud-native.

IBM had survived the last major transition from the mainframe-client to the distributed computing by embracing Linux servers. But this time, things are much more complicated and uncertain, as the mainframe may be finally out due to an anticipated lack of human professional support who can understand the programming language platform that is clearly becoming a relic of the past. Additionally, IBM’s middleware is a key glue among distributed hardware and apps but it is exclusively for on-premises and no longer needed for cloud PaaS.

In my mind, the stickiness for long-term investing should be evaluated on a holistic level of company’s business. The stickiness of the individual product is not going to make it unless there is a structural tailwind. The impression of “because it looks less expensive” needs 1.5 steps deeper as there are concerns that may get bigger. Also, IBM has some financial figures include financial business, which needs adjustments like auto companies.

Those make me feel bearish and feel confused by the pair of Warren Buffett and IBM, the investor I respect and one of my least favorite in large-cap tech investing.