The setup
Texas Instruments spent the first half of the 2020s on a roughly $20 billion, six-year capital expenditure program, building 300mm wafer fabs in Texas and Utah. On the surface, the company absorbed this comfortably. Across FY2023 to FY2025, its cash balance barely moved: $2,964m, $3,200m, $3,225m. A stable, healthy cash line through the heaviest investment period in the company's recent history.
That stability is an illusion. And the way I found out was by building a model, not by reading the income statement.
The accidental discovery
While constructing a three-statement model for TXN, I set two forecast assumptions that seemed conservative and reasonable: hold short-term investments flat, and assume no net new borrowing (a constant-debt base case). With those two assumptions in place, the model's projected cash balance went negative almost immediately.
My first instinct was a formula error. It wasn't. The model was correctly showing that TXN's operating cash flow alone could not cover its capital expenditures plus its full dividend plus its share buybacks. Something else had been funding the gap. To find out what, I went back to the historical statements, this time looking not at the income statement but at the balance sheet and the financing lines.
What the balance sheet reveals
The cash line held steady. The liquidity behind it did not.
| US$ millions | FY2023 | FY2024 | FY2025 |
|---|---|---|---|
| Cash and equivalents | 2,964 | 3,200 | 3,225 |
| Short-term investments | 5,611 | 4,380 | 1,656 |
| Total liquidity | 8,575 | 7,580 | 4,881 |
Short-term investments fell from $5,611m to $1,656m in three years. Total liquidity (cash plus short-term investments) nearly halved. On top of that, the financing section shows roughly $5,300m of net debt raised over the same period.
So the fab expansion was funded by two taps that the cash balance conceals: drawing down the investment portfolio, and borrowing. The flat cash line was held in place by both. Reading only the cash balance, an investor would conclude the buildout was self-funded. It was not.
Why this reframes management's message
On the May 2026 Bernstein call, TXN's CEO said the investment cycle was at its "tail end," that capex would come down, and that free cash flow would begin to return. Read against the balance sheet, the subtext is sharper: the expansion was financed by spending down liquidity and taking on debt, the investment portfolio is now largely depleted, and the company needs capex to fall in order to fund itself from operations again. "Free cash flow will return" is not just an upside story. It is a description of a company that had been leaning on its reserves and now needs the spending to ease.
The modeling lesson
There is a second, more general lesson here, and it is really the point.
In my first forecast, I assumed a flat $1,500m of annual buybacks, continuing the recent trend. On its own, that assumption looked entirely reasonable. It was only after linking all three statements, so that cash, capex, dividends, and buybacks all had to hold at once, that the assumption revealed itself as unaffordable: it drove cash negative. An assumption can be individually plausible and still be impossible when every other assumption has to be true at the same time.
That is what a three-statement model enforces and what a standalone income statement or a quick multiple cannot. It is also why the exercise of judging whether a company needs external financing, something I had studied in the abstract, only became concrete once the statements were wired together and the cash simply would not balance.
The income statement tells you whether a company is profitable. Only the linked statements tell you whether its plan is fundable.