Ten Billion Dollars, Due 2029

On July 22, 2026, IBM told the market two things in a single document. Its revenue would grow more slowly this year than it had promised in April. And it would spend more than $10 billion on machines that do not yet work.

The guidance cut was small in print and large in meaning: full-year constant currency revenue growth of four-to-five percent, down from the more than 5 percent Arvind Krishna had reaffirmed three months earlier. The quantum commitment came with a date attached. IBM remains on track, the release said, to deliver the first large-scale fault-tolerant quantum computer by 2029.

Three years. One number the company controls, and one it does not.

Ninety Days Between Numbers

April had gone well. First-quarter revenue of $15.9 billion rose 9 percent, 6 percent at constant currency. Software grew 11 percent, infrastructure 15 percent, and operating pre-tax margin expanded 140 basis points to 13.4 percent.

"The first quarter was a strong start to the year with broad-based revenue growth across our segments," Krishna said, and held the year's guidance where it was: more than 5 percent constant currency growth, free cash flow up about $1 billion.

Then the second quarter arrived at $17.2 billion - up 1 percent. GAAP pre-tax income fell 5 percent. Diluted earnings per share fell 2 percent. James Kavanaugh, IBM's chief financial officer, attributed the miss to revenue headwinds late in the quarter.

Forty-Two Percent, One Line

The damage sat in one line of the segment table. IBM Z revenue fell 42 percent.

Mainframes sell in product cycles; a quarter deep into a cycle trough looks like collapse whether or not it is one. But the fall was steep enough to drag all of Infrastructure down 7 percent, to $3.8 billion, even as Distributed Infrastructure inside the same segment grew 37 percent and Power and Storage built an order backlog of nearly $500 million. Transaction Processing software - the code that runs on those mainframes - fell 8 percent alongside it.

Set against that: Software revenue of $7.8 billion, up 5 percent, with Red Hat up 11 percent and the Data business up 19 percent. Consulting, at $5.3 billion, did not move.

A company growing in the parts it is buying and shrinking in the part it invented.

Thirty-Six Years At Armonk

Krishna, 63, joined IBM in 1990. He built and led the company's data businesses, ran the Systems and Technology Group, directed IBM Research from 2015 to 2020, and led Cloud and Cognitive Software from 2017 - the seat from which he was, in the board's own description, a principal architect of the Red Hat acquisition, the largest in IBM's history. He became chief executive and a director in April 2020 and was elected chairman that December. He holds an undergraduate degree from the Indian Institute of Technology, Kanpur, and a PhD in electrical engineering from the University of Illinois at Urbana-Champaign. Since 2022 he has been a director of the Federal Reserve Bank of New York.

A researcher who bought the open-source company, then took the chair. The bet he is making now is the one a researcher would make.

The Machine That Does Not Exist

In May 2026, IBM announced a letter of intent with the U.S. Department of Commerce to build Anderon, described in the company's own filing as the world's first pure-play quantum wafer foundry. Behind it sits the spending commitment: more than $10 billion over five years, and a fault-tolerant quantum computer at large scale by 2029.

Fault tolerance is the threshold that separates a physics demonstration from a computer. No one has crossed it. IBM has now put a public year on doing so, in a document filed with the Securities and Exchange Commission, in the same quarter it lowered what it expects to earn.

"We fundamentally believe that we are in the early innings of a structural shift for business," Krishna said in the July release, describing a portfolio positioned "to help our clients tap the value, and manage the challenges, of an AI-driven future."

What Pays For The Bet

The financing of a moonshot is unglamorous and visible in the cash line. IBM generated $2.5 billion of free cash flow in the second quarter, down $0.3 billion year over year. It ended the quarter with $8.2 billion in cash and marketable securities - $6.3 billion less than at the end of 2025 - having spent $10.5 billion on acquisitions during the year. Total debt stood at $62.0 billion.

None of that touched the dividend. The board declared $1.69 a share; with the September payment, IBM will have paid consecutive quarterly dividends every year since 1916.

Krishna's answer to the pressure is to spend into it. The company is expanding sales coverage across thousands of additional clients, hiring specialised technical staff it calls Forward Deployed Engineers, and shipping fast: Lightwell, an open-source security capability launched in the quarter, had made more than 7,500 patches available in its first two weeks. He described the posture as "heavily investing in commercializing innovation at speed and scale."

The July release carried both halves of that posture in adjacent paragraphs - a lowered growth range and a raised ambition - and asked shareholders to hold them at once.

What arrives first is another quarter, and the mainframe line will still be there to read. What arrives in 2029 is either a machine or an explanation. Krishna has three years, and unlike revenue guidance, that number cannot be revised in April.