Company A grew revenue 30%, has a $664 billion order book, and consumed $23.7 billion of cash last year. Company B grows barely at all, and throws off $11.6 billion of free cash flow.
Company A is Oracle. Company B is IBM. We scored them 5.2 and 6.5 — and the lower score went to the one with the growth.
That inversion is the whole article, because it is the thing statistical screens cannot see. A screen sorts on multiples. The difference between a bargain and a trap is not in the multiple at all.
Side by side, as at our reports
| Oracle · 27 Sep 2026 | IBM · 16 Jul 2026 | |
|---|---|---|
| Growth | Revenue +30%, cloud infrastructure +121% | Slow — a mature annuity |
| Order book | $664bn | — |
| Free cash flow | −$23.7bn (FY26) | +$11.6bn |
| Capital spending | $90–95bn this year | Modest |
| Debt | $156bn · rated BBB− | Investment grade, comfortable |
| Concentration | ~half the backlog is one customer | Diversified |
| Valuation at our date | — | ~14× the cash, after a 25% one-day crash |
| Our score | 5.2 | 6.5 |
Look at the free cash flow row, then at the debt row, then at the concentration row. Each on its own is survivable. Together they describe a specific situation: a company borrowing heavily to build capacity for orders that are half-dependent on one customer — a customer who is itself funded by the same capital cycle.
The backlog is real. The cash to serve it is borrowed. Those two sentences can both be true for years, and then stop being compatible in a single quarter.
One question, asked before the multiple
Does the business produce cash, or consume it?
A cheap company that produces cash has time. It can repair itself, buy back its own shares, pay you while you wait, and survive being wrong about the timing. Time is what converts a statistical discount into a return.
A cheap company that consumes cash is on somebody else's clock. Its outcome is decided in a refinancing conversation, not in an earnings report — and the multiple you paid becomes irrelevant, because the equity's value in that conversation depends on whether the lenders feel accommodating.
This is why our lower score went to the faster-growing company. Not because growth is bad, but because growth funded by debt against concentrated orders is not a cheaper version of the same thing. It is a different instrument.
If the build-out works, Oracle will look like a company that was bold while everyone else was cautious, and the 5.2 will look timid. Borrowing to build capacity into visible, contracted demand is exactly what a good management team is supposed to do, and it is how several of the best businesses in our library were created.
Our concern is not the strategy. It is the combination: negative cash flow, BBB−, and half the backlog resting on one counterparty. Remove any one of the three and the position is ordinary. All three together mean the company has very little room to be wrong about timing.
The one that looks like safety
IBM is the opposite risk, and it is worth naming because it is the one income investors actually fall into. A $11.6 billion cash machine with a 3% dividend and a decades-long increase record, available at about fourteen times its cash after a historic one-day crash — that is a genuinely attractive proposition, and it is also exactly what a slowly declining franchise looks like on the way down.
The difference between "cheap cash machine" and "melting ice cube" is not visible in the multiple either. It is visible in whether the cash flow is still there in five years — which is a question about the business, asked one layer below any ratio.
One borrowing into growth, one harvesting a mature annuity, and one whose moat genuinely eroded. Same discount, three different reasons.
Cash flow, debt, concentration and valuation, worked through.
What to do with this on Monday
Take the cheapest thing you own and find its free cash flow for the last three years. Not earnings. Cash: operating cash flow minus capital spending.
If it is positive and steady, you own a cheap business and your main enemy is impatience. If it is negative, you do not own a cheap business — you own a financing situation, and the multiple you paid is not the variable that decides how it ends.
Frequently asked
What is a value trap?
A company that looks statistically cheap on numbers that are still deteriorating, so the ratios improve on paper every quarter while the business gets worse underneath them. The tell is where the cheapness comes from: if it comes from the denominator shrinking rather than the price being wrong, the discount is correct and will keep being correct.
Is Oracle's backlog real?
Our report says yes — and that this is not the question. Revenue grew 30% and cloud infrastructure 121%, with a $664 billion backlog. The issue is how it is being funded: capital spending of $90–95 billion in a year, free cash flow of minus $23.7 billion in FY26, debt of $156 billion and a BBB− rating. The orders are real; the cash to serve them is borrowed.
What is customer concentration risk?
When too much of what a company has sold depends on one buyer. Roughly half of Oracle's $664 billion backlog is a single customer — and that customer is itself funded by the same capital cycle Oracle is borrowing into. If one party's funding stops, both sides of that contract are affected at once.
How do you tell a bargain from a trap?
Ask one question: does the company generate more cash than it consumes? A cheap business that produces cash can wait, repair itself and pay you while it does. A cheap business that consumes cash is on a clock set by its lenders, and the multiple is irrelevant because the outcome is decided by refinancing rather than by earnings.