Political and economic uncertainty is a constant challenge for International Business Machines, which does business with enterprises in 185 countries around the world. Add in high interest rates and it becomes difficult to make a profit.
Luckily for Big Blue, there are product cycles that it must follow because sooner or later its systems customers will need more computing power and storage capacity for their businesses, and lucky for IBM, there are always new things coming along that need to be mastered and that IBM can sell as software or a service.
Thus, IBM’s emerging generative AI consulting and software business, along with Power10 and z16 processor-based servers that can run light AI training and moderate AI inference on the same platforms that run databases and applications, are benefiting from the GenAI wave, filling the conservative spending gaps seen elsewhere in the data center.So, on the surface, things look pretty good for IBM and its hundreds of thousands of enterprise customers, even at a time when there is a lot of change and disruption going on.
For the second quarter ended in June, IBM’s revenue increased 1.9% to $15.77 billion, gross profit increased 5.3% to $8.95 billion, and net income increased 15.9% to $1.83 billion, representing 11.6% of revenue. That was a pretty good second quarter considering its consulting business was struggling as clients stalled some projects, like application and database modernization, to focus on others, like AI.
This phase change is no different from when IBM had to pivot from pitching outsourcing services to selling hardware and web infrastructure software during the dot-com boom. It remains to be seen whether IBM can build a GenAI software stack that rivals the WebSphere middleware giant of the late 1990s and early 2000s. WebSphere still benefits the company today, but IBM has certainly accumulated many pieces and is adding more. Both Red Hat Enterprise Linux and the OpenShift Kubernetes platform ship with IBM’s own Granite pre-trained AI models, which, like Meta Platforms’ Llama model, are open source and range in size from 8 billion to 36 billion parameters, with test results for many cognitive tasks that match or exceed popular models from OpenAI, Mistral, Inflection, and others.
IBM offers a variety of routes to market for new AI products: You can buy IBM models to run on RHEL, you can containerize IBM models and applications to run on OpenShift, or you can buy these models and dozens of others as part of the Watsonx development stack, which is the closest parallel to WebSphere that Big Blue has for GenAI.
It’s an important aspect. Then we’ll go back to calculating the money.
The Watsonx stack has three homegrown families of AI models, all named after metamorphic rocks.
The Slate model is an encoder-only model, suitable for classification and entity extraction tasks, not intended for GenAI. The Granite model is a decoder-only model, and is therefore used only for generative tasks. IBM’s Granite models are small, ranging from 8 billion to 34 billion parameters, but perform comparably to other models in cognitive tests, are open source, and can be expanded to larger parameter sets if needed. The Sandstone model employs a combined decoder-encoder AI approach, and can be used for both generative and non-generative tasks.
IBM is committed to training its own models and, at least for Granite, keeping them open source so that companies can see how they work and, in theory, avoid having to rely on a single vendor with a closed-source model. While that does provide more transparency, no one has been able to deterministically show how a GenAI model derives a response, so accountability and reproducibility remain issues.
To this we say: This is what happens when you base your own determinism on a statistical system rather than a deterministic one.
Either way, here’s the takeaway from Big Blue and the GenAI wave: After IBM’s Watson QA system beat humans on the game show Jeopardy! in February 2011, and IBM’s conquered humanity, the company now has a different, more practical mission after accomplishing that feat. IBM now has a different set of things to prove.
To some extent, everyone is excited about GenAI today in the same way we were excited 15 years ago about the prospect of IBM turning its Watson QA system into something useful. Now everyone is trying to build a QA system that is much more scalable, much more powerful, and dare I say, omniscient, than Watson. IBM could do it, but what’s the point when Sam Altman seems fixated on it? What IBM needs to prove above all else is that it has a firm grasp on the utility of GenAI, and that it will pay for the modernization of GenAI by engaging existing clients with consulting, software, and systems to modernize their applications and retool their workforces as automation advances.
This is IBM’s game, played on many levels. It’s not as flashy as Microsoft’s or OpenAI’s, but it may be more effective for a more conservative company looking to cut costs and boost sales, and not trying to create an artificial general intelligence to replace our childhood friends. (I’ll admit that in the late 1970s and early 1980s I wanted my own R2D2 units and X-Wing fighter planes, and I even tinkered around with coding Eliza in basic code on my Commodore 64, but I quickly grew tired of all the typing and nonsensical answers, and I had no desire to build a semantic tree of all knowledge, so I went fishing and camping instead.)
In the long term, and perhaps in the wake of its HashiCorp acquisition, IBM will have an opportunity to help companies build out the software stack for GenAI, keeping existing customers of Power and mainframe platforms happy and helping them ride the GenAI wave. It seems unlikely that IBM will suddenly start selling loads of Power and mainframe hardware to new companies using X86 and Arm platforms to run GenAI. And IBM is not foolish enough to think it can.
Big Blue is tending its own garden and will likely benefit from GenAI’s efforts.
Based on numbers from the past three quarterly reports and the current report, here’s how the GenAI boom has helped Big Blue so far.
It’s important to note that the above table is cumulative bookings to date, not revenue for each quarter. IBM doesn’t discuss revenue recognized to date, which is only a portion of bookings. Much of the company’s GenAI software is sold under a subscription license and recognized over time. Revenue recognition for consulting contracts is also spread over time.
Although this market is still too young to operate smoothly, IBM’s GenAI services and software (including but not limited to Watsonx) are clearly showing an upward trend in cumulative bookings. Initially, the ratio of these bookings was approximately one-third software and two-thirds services, but by Q1 2024, it was one-quarter software and three-quarters services. Both GenAI software and GenAI services are experiencing very high sequential growth, and in the long term, we believe that GenAI software and GenAI services will not only reach tens of billions of dollars in bookings over the past 12 months, but also billions of dollars in actual sales each quarter.
But it will likely take several years to get to this point. Lower inflation and economic and political stability will likely help accelerate this transformation. But uncertainty may backfire as companies like IBM are able to better leverage GenAI for specific industries and customer use cases. IBM remains a leader in the field, helping hundreds of thousands of businesses adopt internet technologies in the 1990s, transforming ERP applications in the 1990s and 2000s, and data analytics and HPC in the 2010s. GenAI is just the next opportunity. And it’s a big one.
Meanwhile, IBM’s “real” systems business, which consists of servers, storage, operating systems and other middleware but not databases and applications, is doing well. Our model predicts that the core systems business brought in revenue of just over $7 billion in the second quarter, up 3.4%, and that real systems pre-tax profits rose 4.1% to $3.62 billion, accounting for 51.7% of revenue.
This is the ballast that keeps the IBM ship straight and provides the momentum to keep moving forward, regardless of the shifting seas of IT.
