International Business Machines

On the 14th of July 2026, this 115-year-old company had the worst single trading day in its history — a 25% collapse, worse than the crash of 1987, roughly seventy billion dollars of value gone before lunch. Six weeks earlier it had touched an all-time high on quantum-computing euphoria. The trigger was mundane and terrifying at once: IBM warned that its software revenue grew only 5% last quarter instead of the ~10% Wall Street expected, and explained that enterprise customers, scrambling to buy AI hardware before prices rose, had raided the very budgets that buy IBM's software. The whole software sector fell in sympathy on a single word: cannibalisation. So we arrive, as value investors so often do, at the scene of a wreck — to ask whether the market has broken something, or merely a narrative.

What the wreckage exposes is a business most people misremember. IBM is no longer the declining hardware dinosaur of the 2010s. After spinning off its shrinking services arm as Kyndryl in 2021, it is now, at its core, a software-and-consulting company built on the most durable annuity in technology: the mainframe. Roughly 44% of revenue is software, ~31% consulting, ~23% infrastructure (the mainframes themselves). It earned $10.6 billion of net income on $67.5 billion of revenue in 2025, and — the number that matters most — threw off $11.6 billion of free cash flow, which after the crash the market now values at roughly 14 times. It has paid a dividend every quarter since 1916 and raised it for 31 straight years. This is not a broken company. Whether it is a mispriced one is the question.

"I was wrong… IBM is a big strong company, but they've got big strong competitors too." — Warren Buffett, CNBC, August 2017, on exiting his IBM stake

No company on our shelf carries a heavier Buffett ghost. In 2011 he broke his own rule against technology and bought $10.7 billion of IBM, praising its predictability and its buybacks. By 2018 he had sold it all at roughly breakeven, admitting flatly that he was wrong — and rolled the money into Apple, which became Berkshire's greatest win. So the question this report circles is irresistible: has IBM finally become the durable, cash-gushing software franchise Buffett thought he was buying in 2011 — just as the market, on the 14th of July, decided all over again that it isn't?

IBM's history is the longest in modern technology, and it matters because the mainframe you must value today is the direct descendant of a bet made in 1964 — and because this company has burned two full generations of narrative investors, a fact any buyer must stare at squarely.

Two lessons for an owner, pulling in opposite directions. The bull's lesson: a company that has reinvented itself across a century — tabulating machines to mainframes to services to software to AI — and whose 1964 mainframe bet still prints cash in 2026 has proven, more than almost any firm alive, that its core is durable. The bear's lesson: this is the company that spent a lost decade disguising decline with buybacks, over-marketed the last AI (Watson) into embarrassment, and has left two generations of investors — including Warren Buffett — poorer for believing its turnaround stories. The crash of 14 July is, in that light, either the third act of a genuine refounding or the market refusing, once again, to be fooled. Your job is to decide which.

The machine, in five steps:

Be precise about the two halves, because the crash confused them. The durable moat is the mainframe-and-Red-Hat annuity — the thing that generates the $11.6 billion of free cash flow, protected by switching costs no competitor can breach in under a decade. That half is knowable, and it is what you are really buying at 14 times cash. The contested part is everything the 2024–2026 rally was about: that AI bookings convert to high-margin revenue, that the mainframe benefits from AI, and that quantum becomes a business. On the 14th of July the market discovered that IBM sits on both sides of the AI trade — its customers' AI-hardware binge came out of the budgets that buy IBM software — and repriced the optimism accordingly. A Buffett-style buyer underwrites the boring annuity at a fair price and takes the AI and quantum as free options. The danger is paying up for the options and discovering the annuity itself is what's shrinking.

FY2025 revenue by segment — the shape of the post-Kyndryl IBM:

The engine, properly understood. Ignore the hardware headlines: IBM is a software-margin business anchored by a hardware moat. The mainframe (Infrastructure) barely grows, but it is the gravity well — more than seventy percent of the Fortune 500's core operations touch an IBM mainframe, and once they do, the high-margin Transaction Processing software that rides on it renews almost automatically, year after year, at ~80% gross margins. That annuity is the profit engine, and it is precisely what stumbled in Q2 — not because customers left, but, IBM says, because they deferred. Meanwhile Red Hat is the genuine growth star: Broadcom's brutal VMware price increases (renewals up hundreds of percent) are driving a wave of enterprises to Red Hat's OpenShift, and it was the one line accelerating even inside the July warning. The whole investment case reduces to one question the numbers can't yet answer: is the software annuity merely paused, or is the AI boom quietly draining it?

IBM's moat is genuinely wide where it matters and genuinely contested where it grows. Three walls:

1 · Mainframe lock-in — switching costs measured in decades. The world's banks, card networks, airlines and governments run on sixty years of accumulated COBOL and CICS logic on IBM Z. Migration projects fail publicly and expensively; the realistic alternative to a mainframe is not another mainframe but a ten-year, nine-figure re-platforming most CIOs will never start. IBM holds >90% of the installed mainframe base. This is the closest thing in software to Buffett's beloved railroad — a right-of-way no one will rebuild — and it funds everything. Trend: shrinking slowly, but sticky, and cash-rich.

2 · Red Hat / hybrid cloud — a widening wall. RHEL and OpenShift are the de-facto enterprise Linux and Kubernetes standards, and Broadcom's VMware price shocks handed IBM the strongest organic tailwind Red Hat has had since containers — enterprises are fleeing to OpenShift by the thousand. Trend: widening — the one unambiguous bright spot.

3 · Deep enterprise incumbency & trust. Decades-deep relationships in the most regulated industries; consulting as the distribution arm for the software. A slow, unglamorous, durable wall. Trend: stable.

Where it cracks — and where it's merely an option. The mainframe wall is durable but slowly eroding as workloads drift, over decades, toward the cloud. The growth the market paid for — that AI bookings become high-margin revenue and that the mainframe rides the AI wave — is precisely what the July warning called into doubt. And quantum, which drove the June melt-up to $329, is a real research program with a concrete roadmap and a US-government-backed foundry — but near-term revenue is negligible; the market capitalised an option as if it were a business, and just repriced it. I score the moat a 7: a genuinely wide annuity, with a growth leg that is half real (Red Hat) and half faith (AI, quantum).

Everything — the crash, the value, the verdict — turns on a single interpretation of one quarter's software number. IBM guided investors to expect software growth accelerating toward ~10%. It delivered ~5%. The stock lost a quarter of its value in a day because of why: management said enterprise customers, in the last weeks of June, reprioritised their quarterly budgets toward AI infrastructure — servers, storage, memory — to lock in supply-constrained hardware ahead of price increases, and deferred the software and IT projects that buy IBM's products. Two readings of that sentence are worth about a hundred dollars a share apart.

Here is the honest, uncomfortable truth an owner must sit with: the definitive answer arrives on 22 July 2026 — six days after we publish this — when IBM reports full results and, crucially, tells the market whether its full-year guidance survives. If management reaffirms the year and frames the miss as pure timing, the deferral reading wins and today's price will look like a gift. If it cuts the year, the trend reading gains, and the stock has further to fall. We cannot resolve it for you, and we will not pretend to. What we can say is that the crash has priced in a great deal of the bad case already — at 14 times free cash flow you are paying an annuity multiple for a business the market five weeks ago called a growth stock — and that the mainframe cash and the 31-year dividend protect you while the question is answered. This is the rare analysis where the calendar, not the thesis, holds the trigger.

The competitive story is one of a company that lost the war everyone watched and quietly won the one that pays. IBM lost the public-cloud land grab to Amazon, Microsoft and Google a decade ago — that is precisely the disruption Buffett failed to foresee in 2011. But it stopped fighting the hyperscalers head-on and instead sells the hybrid thesis: Red Hat's OpenShift as the neutral management layer that runs across all the clouds and the mainframe at once. And it caught an extraordinary break — Broadcom's acquisition of VMware and the savage price increases that followed have turned into the biggest organic tailwind Red Hat has ever had, as enterprises flee to OpenShift. In the arena that generates the cash — the mainframe and its attached software — IBM has no competitor at all. The July warning was not about losing to a rival; it was about IBM's customers spending on AI hardware from Nvidia instead of on IBM software. That is a subtler, and possibly more temporary, problem than losing share.

Management is coherent and disciplined; the balance sheet shows the cost of its ambition.

The good, and the ledger it's written on. The good is real: a 31st consecutive annual dividend increase, a payout that consumes only ~55% of free cash flow (comfortably covered even in a soft year), and a dividend paid without interruption since 1916 — through two world wars, the Depression, the 1993 near-death and the lost decade. Krishna's strategy — shrink to grow, buy the platforms that matter, manage to cash — is coherent and, until 14 July, was working. The ledger it's written on carries a warning of its own: three large software acquisitions in two years have pushed total debt to ~$67 billion and left ~$79 billion of goodwill and intangibles — more than twice the company's $33 billion of tangible-adjacent equity. Net debt runs ~3.4× EBITDA. This is not distress — the FCF services it easily — but it means the reported returns are inflated by the thin equity base (ROE of 35% sits atop a ROIC of just ~10%, the honest number), and it constrains the buyback firepower precisely when the crashed stock would most reward it. I score management a 7: a credible refounder with real cash discipline, carrying an acquisition-heavy balance sheet and the burden of a company that has disappointed believers before.

The whole equity case is one row: $11.6 billion of free cash flow, from a business that spends almost nothing on capital equipment because software and mainframes don't need factories. That cash covers the ~$6.3 billion dividend nearly twice over and funds the acquisitions besides. But read the balance sheet with clear eyes, because it is the honest counterweight: this is a company that bought much of its software growth — $79 billion of goodwill and intangibles against $33 billion of equity, and $67 billion of debt — which is why its dazzling 35% return on equity collapses to a merely-decent ~10% return on the capital actually invested. That gap is the truth about IBM: a genuine cash machine, but one assembled through acquisition and leverage rather than pure organic compounding. At 14 times that cash flow, you are being asked to decide whether the annuity that produces it is durable enough to justify the borrowed, goodwill-heavy foundation beneath it.

For a business that manages itself to free cash flow, free cash flow is the lens — and on it, the crash created something rare on our shelf: a wide-moat cash machine at ~14 times cash, a 6.4% free-cash-flow yield, on top of a 3% dividend raised for 31 straight years. The models are emphatic: the automated DCF reads $406, an implausible-looking +85%, and Wall Street's average target sits at $308, +40% (from a divided panel — 24 buy, 21 hold, 5 sell). Treat both with caution, because they largely predate the full digestion of the July warning and may be cut on the 22nd. But even halving the DCF's optimism leaves a stock trading well below fair value if the software annuity is merely paused. The honest valuation verdict is therefore conditional, and we will state the condition plainly: at 14× free cash flow, the price already assumes a good deal of the bear case — a durable-annuity multiple, not a growth one — so the margin of safety is real, resting on the mainframe cash and the dividend, and the upside is a call option on the 22 July guidance confirming 'deferral.' The zone deepens toward the ~$205 post-crash low, where the FCF yield pushes past 7%.