So abandon AI and now quantum is the new moonshot? AK is tossing-up a Hail Mary.
https://www.wsj.com/tech/ibm-stock-crash-quantum-computing-mainframes-7cc6c18f
While other tech giants spend on AI data centers, Big Blue aims to keep selling mainframes and pour profits into quantum computing
By: Christopher Mims | July 23, 2026 7:37 pm ET
IBM stock has been languishing since July 14, when it pre-announced disappointing earnings. But the current AI bo-m might only present a temporary disruption to its core business. And it will drive opportunity for the 115-year-old company, says Chief Executive Arvind Krishna, who spoke to me in an interview.
To understand where IBM is going, you have to understand where it’s been, and how it makes money today.
Nearly every time you pay for something with a credit card, that authorization is processed by an IBM system. IBM’s mainframes, servers and software are no less dominant in the core functions of insurance companies, banks, airlines and other members of the Fortune 500. “When we look at that sort of financial backbone, 70% by value of all transactions flow through the mainframe on a daily basis,” says Krishna.
If IBM’s hardware and software suddenly ceased to exist? “The economy as we know it would kind of grind to a halt,” he adds.
IBM is infrastructure. And infrastructure might be less exciting than the generative-AI models changing our lives (and running amok), but it’s safe—and profitable.
The company reported $4.8 billion in free cash flow in the first half of 2026. Alphabet, meanwhile, just announced that it is cash-flow negative for the first time since its stock-market debut.
The tech industry might have abandoned powerful-but-specialized mainframe computers and shifted to the cloud, but IBM’s customers—what you might call the “real” economy—still lean heavily on them. Krishna says the company is counting on that demand to continue for years to come, until the company can debut its next trillion-dollar business: quantum computing.
The stock-market scare came when Krisha told investors that a swath of valuable orders were delayed, suggesting that its customers were moderating their spending because their other IT costs had risen on account of the AI bo-m. He’s now saying he’s confident that most of those deals will close within six months.
IBM has been making mainframes since 1964. The company has never given up its core advantage in computing: designing its own processors. Its custom Power architecture still drives IBM servers that handle critical business functions. Another custom line, called Tellum II, powers IBM’s latest mainframes.
In the archipelago of computing architectures—which includes Intel’s x86, still found in PCs and servers, and the ARM chips that power everything from smartphones to AI clusters—IBM’s stands alone. It’s an island of design constraints and engineering decisions that evolved across decades to serve the company’s demanding, deep-pocketed customers.
IBM’s mainframe processors have huge amounts of on-chip memory, yet relatively few very fast computing cores. These chips are all about processing a stream of data (say, bank transactions) very quickly. By contrast, Nvidia’s latest systems spread massive AI workloads across thousands of processors.
Hardware sales are only about 20% of IBM’s revenue, and mainframes are about a quarter of hardware sales, but its software and consulting services are heavily dependent on those sales.
For every dollar a customer spends on mainframes, they will typically spend an additional $3 to $4 across IBM, says Krishna.
IBM employs close to 300,000 people, a vast workforce 50% larger than that of Google parent company Alphabet. One of the original multinational corporations, only 20% of its employees are in the U.S.
About 160,000 of IBM’s employees are in its services business. “That has to be local to where the service is being provided, because people want language skills, people want locality, people want you in their office,” says Krishna. “That piece is labor-intensive, let me acknowledge that.”
Krishna anticipates the AI bo-m will also boost IBM’s revenue, because most customers need someone to come in and hold their hand. Only the very biggest non-tech companies have in-house tech expertise to assess and incorporate the latest AI systems, he adds.
Hardware being a linchpin of IBM’s revenue is one reason investors found the company’s recent sales decline so alarming.
The good news, says Krishna, is that of the large deals IBM failed to close by the end of last quarter—numbering in the “low tens”—about a third closed within the first three weeks of the current quarter. That is much faster than what’s typical for deferred deals, he adds, and he expects 80% of those deals will close within six months.
IBM’s near-total dominance of an obscure but important corner of tech hasn’t made it immune to customers redirecting their IT budgets toward memory and storage chips. Both have skyrocketed in price as the entire silicon supply chain reorients to feeding the AI infrastructure bo-m. Some analysts believe IBM might miss even its more modest revenue targets for the rest of the year.
Krishna pushes back. He expects the dip in revenue to be a one-time event. He also says that, while IBM itself prepurchased $600 million in hardware to get ahead of price increases, it isn’t financially prudent for companies to continue buying hardware, only to have it languish in inventory.
IBM is an odd duck in the technology industry, in that it continues to have positive cash flow. Giants including Google and Oracle are reporting staggering negative cash flow, and the value of every publicly traded AI neocloud startup has dropped significantly in the past six months as investors worry about their mounting debt.
IBM is investing its profits in its future, says Krishna, particularly as a leader in quantum computing.
By 2029, the company projects, it will offer the world’s first truly useful quantum computer. In June, it said the Trump administration would spend $1 billion to help it set up a quantum-chip manufacturing facility. On Thursday, it announced it will acquire HRL Laboratories, which brings in complementary research on the fundamental units of quantum computing, called qubits.
In conventional chip manufacturing, IBM continues to do pioneering research. It recently showed off what it calls the world’s first sub-nanometer chip technology—although demand for this architecture is unclear. IBM has also sunk $5 billion into a project called Lightwell, which uses AI to find and help fix vulnerabilities in open-source software. With powerful AI models that can hack into systems rapidly proliferating, this, too, could be a growth opportunity for the company.
In 1985, IBM was, briefly, the world’s most valuable company. Then came the PC. Since then, it’s been slow-growing at best. It isn’t clear which of the company’s bets will pay off. But tech companies are spending trillions on AI that has yet to benefit end customers with appreciable returns. So perhaps IBM is the tortoise in this race.
Krishna himself is, naturally, bullish on the future of his company. He believes quantum computing will begin to power IBM’s growth in a matter of years, even if critics say it could be more like decades. The company’s first-mover advantage in quantum computing could, he believes, boost its fortunes the way being a first mover in GPUs did for Nvidia.
When it comes to the total size of the opportunity for quantum computing, “I have actually upped my estimate,” says Krishna. “I think quantum will be a trillion-dollar value-adder by the end of the ’30s.”