IBM entered Think ’26 with a different posture than the market is used to. In this AI cycle, the company is leaning hard into openness and ecosystem – and doing it without the usual proprietary overtones. According to CEO Arvind Krishna, that choice is purposeful. It’s also a sign that IBM’s CEO has a clear vision of the puck is going and the approach IBM must take to win in this era. According to Krishna, foundation models are in a game of leapfrog and must become interchangeable over time. Value is shifting to the platform layer to ensure governance, integration, and the ability to move workloads and data without getting trapped by one vendor’s propriety.
This Special Breaking Analysis is based on an analyst AMA with IBM Chairman and CEO Arvind Krishna at IBM Think 2026. The session covered where IBM is constrained, what metrics matter, how IBM is allocating resources, the workforce implications of AI, and why Krishna chose not to show a “stack slide” in the keynote – a deliberate choice to stay anchored on business value, not tech specs. Our take is that IBM’s open posture is directionally correct, but it raises a strategic question in that IBM’s greatest historical strength has been deep vertical integration. The mainframe proves what integrated, transaction-grade engineering can achieve for customers. The opportunity now in our view, is to combine that integration with an ecosystem that’s motivated to build, sell, and profit alongside IBM.
Key question: Where is IBM still structurally constrained – and what changes fastest?
Krishna’s answer started with a blunt reality commenting that IBM doesn’t have enough market share and he wants teams to do twice as much in half the time. He connected that to a portfolio view that is not only reliant on M&A but also involves predicting where clients will find value around the corner; then concentrating credibility and investment into a small number of arenas where IBM can win. He pointed to Red Hat as the by far the best bet. He said the Red Hat acquisition looked “weird” to many observers in 2019, but it has proven to be the default delivery model.
He also challenged the stale perception that IBM is inherently proprietary, arguing IBM is the “most open source company in the world,” and that has doubled since the Red Hat deal.
Key Message: IBM is trying to win by narrowing focus, accelerating execution, and leaning into openness as a first principle, while still picking a few spots to go deep and lead.
Key question: What metrics will prove IBM’s AI strategy is winning over the next 24 months?
Krishna’s answer was unequivocal: “software growth.” He tied that directly to the idea that AI success is not just about the best AI model – it’s where AI runs, how data is unlocked, and whether IBM’s hybrid and data strategy are pulling workloads into IBM software. He also referenced the scale of IBM’s GenAI book of business (noting it has surpassed $12.5B).
Key Message: IBM is effectively telling investors and customers to watch software growth as the proxy for AI traction, because that’s where repeatable value flows once AI moves into production workflows.
Key question: How do you prevent the mainframe’s gravitational pull from slowing the AI-era portfolio shift?
Krishna pushed back on the premise that the future is only cloud native, arguing the market is still meaningfully split between on-prem and cloud and that IBM doesn’t care where you run it – the priority is resilience and long-term cost. He acknowledged the temptation of high-margin, slower-growth franchises, but described two counterweights: 1) Resource allocation from the top (IBM’s R&D spend growth); and 2) Sales specialization and incentives geared to newer portfolio areas.
Key Message: IBM is managing the portfolio with incentives and capital allocation – treating mainframe as an advantage to extend, not a reason to underinvest in the next stack.
Key question: How is AI changing IBM’s workforce strategy – and what’s the real job impact?
Krishna said IBM is planning to triple entry-level hiring in 2026 based on two convictions: 1) The business will grow, and 2) AI productivity can make a person with one year of tenure perform like someone with five years of experience. He also said the shape of hiring is changing, with fewer clerical/back-office roles, more engineering, sales, and digital roles. He offered an anecdote from his daughter who claims roughly 25% of work that took eight hours three years ago now takes less than an hour.
He added a concerning observation about education saying universities are moving too slowly, though he sees a meaningful minority leaning in. He shared an observation that the educational system is polarized with one end of the spectrum forcing students to lean into AI and the other end banning the use of the technology in their studies. Krishna, like most folks in tech favors the former approach.
Key Message: IBM is leaning into AI as a productivity engine and using it to justify growth hiring, while acknowledging roles are materially changing, especially where repetitive work can be streamlined.
Key question: What should CEOs do differently when AI becomes the digital workforce?
Krishna rejected the idea of AI becoming autonomous in the consciousness sense, describing it as a powerful tool under human control. His advice to other CEOs was focus on the operational:
- Stop treating AI as a generalized “risk-based” science project;
- Pick 2–3 high-impact areas where benefits are provable;
- Over-invest to make those succeed – instead of running 100 experiments;
- Tie success to leaders who want the change, not just technical experts.
Key Message: The operating shift is cultural more than technical – and Krishna’s prescription is focus, commitment, and leadership accountability.
Key question: Why didn’t IBM show a “stack slide” in the keynote – and does IBM favor integration or a more dis-integrated model?
Krishna said the decision was intentional – he wanted to stay on business value rather than “tech up,” and he also wanted to keep AI and quantum discussions distinct.
On the deeper integration question, he argued enterprises of scale should keep the discipline to work with two or three model options because foundation models will become a commodity in the sense that you can substitute one for another (admittedly, with some work). He stressed the discipline of insulating end users from idiosyncrasies of model analysis, and keeping portability inside the platform layer.
Key Message: While IBM has an end-to-end stack, the company is betting that choice expands and that sustainable value shifts to the platform layer – where portability, integration, and governance live.
Key question: Where does IBM fit in agent infrastructure – especially in a multi-cloud, multi-vendor enterprise world?
Krishna described agent infrastructure as early, saying it will require rough consensus rather than complete consensus, and implied it’s still too soon to declare a single winning approach. He suggested IBM’s role is neutral plumbing across heterogeneous environments. He referenced learnings from history with initiatives like message queuing (MQ) and service-oriented architecture (SOA), stressing the importance of things working across multiple agent providers rather than trying to be the unique agent provider.
Key Message: IBM’s agent strategy aligns with Krishna’s philosophy – i.e. integration across heterogeneity. The near-term test is whether IBM can operationalize that neutrality into an enterprise-grade control layer and scale that across industries and clients while building an open ecosystem.
Key question: How soon does quantum move from science to engineering – and what are the practical workloads?
Krishna argued quantum is shifting from a science problem into an engineering problem, with progress coming faster than many expected. He cited molecular modeling (chemistry/biology) as the obvious near-term category, referenced protein simulation progress (12,000-atom scale), and emphasized hybrid approaches where quantum tackles pieces and classical systems assemble results. He also pointed to differential equations and real-world physics as the next category – aerodynamics, fluids, reservoirs – problems that are currently too hard to solve classically.
He suggested commercialization as late 2028/2029, tied to both technical progress and the economics of building systems at scale.
Key Message: IBM is positioning quantum as less of a platform shift, and more of a hybrid complement by nature – with near-term traction in molecules and a credible path into physics-based simulation as systems improve.
Our take: A different IBM – and the opportunity to fuse openness with “mainframe-class” integration
We believe this (and other Krishna AMA’s) show a materially different IBM than the one many still carry in their heads. Krishna repeatedly stressed openness, and he challenged the “IBM is proprietary” narrative and tying IBM’s identity to open source scale post–Red Hat. That posture is sensible in an era where developer affinity and ecosystem economics decide who wins.
We were struck when Krishna referenced a 2018 interview with John (Furrier), where they discussed whether Kubernetes and containers would win, and he connected that to a broader point about the need for standards and rough consensus rather than one company keeping tight control. He used that example to frame his answer about how the agentic infrastructure layer may evolve similarly.
At the same time, we believe one of IBM’s greatest advantages has been deep integration – the mainframe and its continued success is the proof point; and frankly the one and only product where IBM is an unquestioned number one. It’s still the gold standard for secure, high-performance transactions, and that level of integration creates real customer value that DIY approaches struggle to match. The trick in this cycle is to build a tightly integrated stack, while still recruiting an ecosystem that can make money alongside IBM – AWS has proven the power of ecosystem, but it has not delivered the same kind of integrated transaction-grade stack end to end.
That’s where IBM has an opening in our view. If foundation models become commoditized, the sustainable value shifts to the integration layer. We see a key opportunity for IBM in data and process harmonization, workflow integration, governance, and the ability to operationalize AI into end-to-end processes. Krishna chose not to show the stack slide in the keynote because business value is the right emphasis at Think. But we also believe IBM can press its advantage by articulating a fuller end-to-end integration story – one that looks more like an ontology-driven harmonization layer enterprises can trust and operationalize (our system of intelligence – SoI – model). IBM can strike a posture of “open by default” plus “integrated where it differentiates” is a viable path to success in the AI era.
Key Message: In our view, IBM is doing the right thing by leading with business value and openness – now it has to connect that to an integrated, enterprise-grade story that turns hybrid + governance + data into a repeatable operating model customers can scale.
Action item for CIOs: In the next 60–90 days, pick one cross-functional workflow that has real business value (cycle time, cost, compliance risk) and run a disciplined platform reality check against IBM’s premise. The deliverable isn’t a model demo or a RAG-based chatbot. It’s a governed, end-to-end agentic workflow that includes systems of record, applies consistent identity and policy, and produces an auditable, repeatable outcome. The goal should be to begin to shape an AI operating model. and move beyond pilots and point tools.
The gotcha to avoid is treating open as an architecture by itself. Openness helps ecosystem expansion and choice, but it doesn’t replace deep integration where it matters most – i.e. transactions, security, and deterministic execution. CIOs should stress test whether IBM has an intention of participating in earnest to own the harmonization layer across data and process semantics while still keeping the ecosystem motivated to build and profit alongside IBM. If IBM chooses not to go there, it will affect your relationship with the company, as we see this layer as the highest value piece of real estate in the emerging AI software stack.
If IBM is not delivering that value, you’ll need to find a partner that will, and understand how and where your IBM stack fits.
