A Stroke of the Pen With No Shovel

The credits in this course share a feature: they attach to a physical act. Build a wind farm, generate a kilowatt-hour, tool a fab — the credit follows a thing that happens in the world. Section 174 of the Internal Revenue Code is different, and its difference is the point. It changed nothing physical, sited nothing, built nothing, and it still reached into the accounts of a wide swath of technology-intensive firms and made them worse. If you want to feel the fragility thesis in your bones rather than your notes, this is the provision to study.

Here is what happened, stated carefully. For decades, a business could deduct its research and experimental expenditures immediately — the money spent on developing new products, processes, and software was subtracted from taxable income in the year it was spent. This is called expensing, and it is generous: a dollar of research reduces this year's taxable income by a dollar. The 2017 tax law contained a provision, scheduled to take effect for tax years beginning after 2021, that changed the treatment. Beginning then, specified research or experimental costs had to be capitalized and amortized — spread out and deducted gradually over a period of years rather than all at once. Domestic research costs were to be amortized over five years; research performed abroad, over fifteen. Crucially, software development costs were swept explicitly into the definition of research or experimental expenditures, which is why this landed so hard on exactly the software-intensive firms that populate the compute economy.

Notice what did and did not change. What changed was timing. The total amount a firm may eventually deduct for its research is largely the same; it is the schedule that moved, from immediate to stretched-out. But timing is not a cosmetic detail in finance — timing is cash. A firm that could once deduct a dollar of software development this year must now deduct a fraction of it this year and the rest over subsequent years. In the near term, its taxable income is higher than it would have been, so its tax bill is higher, so its cash is lower — sometimes dramatically, for a firm whose costs are overwhelmingly the salaries of engineers writing software. Companies that were reinvesting nearly everything into research, and had been paying little tax because their research zeroed out their income, suddenly owed real money on income that was, in an economic sense, plowed straight back into the business. The rule did not touch a single building. It re-timed a deduction, and re-timing a deduction moved cash off balance sheets across an industry.

That is why I call it a stroke of the pen with no shovel. There was no construction, no siting decision, no permit, no interconnection. A sentence in the tax code took effect on a date, and the economics of research-heavy firms deteriorated on that date. Compare this to everything else in the course. A clean-electricity credit is fragile, but at least it is tethered to a physical project you can watch. Section 174 demonstrates the fragility in its abstract, undiluted form: fiscal policy can change a firm's economics with zero physical mediation. Nothing you could photograph would tell you it had happened. Only the cash-flow statement knows.

Draw the connection to the second-power-grid frame explicitly, because it is the reason this provision earns a module. The first power grid — physical infrastructure — changes slowly and visibly. You can see a substation being built. The second power grid — fiscal policy — can change instantly and invisibly, and Section 174 is the cleanest proof, because it changed instantly and invisibly and hit hard. A practitioner who has internalized this stops treating the tax code as a stable backdrop and starts treating it as a live variable that can move without any warning sign in the physical world.

A measured word on where the story went, because it sharpens rather than softens the lesson. This provision proved unpopular and disruptive enough that there were repeated, bipartisan legislative efforts to reverse or soften it, and by my read the treatment of domestic research costs was subsequently revised again to restore something closer to immediate expensing for research performed in the United States — with the treatment of foreign research left less generous. I will not stake the lesson on the precise contours of the fix, because the contours are exactly the kind of thing that keeps moving. Stake it instead on the pattern: a stroke of the pen imposed the change, and a stroke of the pen substantially undid part of it, all within a few years, all without a shovel touching the ground. That two-way volatility is not a bug in my argument. It is the argument. The second grid can be re-wired in both directions, quickly, by whoever holds the pen — and a firm's fortunes swing with it while nothing physical moves at all.