The average PhD stipend in the US is a true number that will not help you. That sounds contradictory, but sit with it. An average mashes together computer science and English, private universities and public ones, Boston and Tuscaloosa. You do not live in an average. You will live in one specific apartment, in one specific city, on one specific stipend.
What averages are good for is the shape of things. STEM pays more than humanities. Private schools usually show bigger numbers on paper. The gap between the top and the bottom of the market is enormous, and I mean genuinely enormous, not the kind of enormous people say when they mean noticeable. Knowing the average gives you a tripwire: if your offer sits way below it, that is worth a hard question before you sign.
Then comes the monthly math, which is where averages go to die. Divide by twelve. Or by nine, if the money only covers the academic year, which changes everything. A stipend that sounds respectable annually can feel thin monthly once rent, food, and the emergency you did not plan for take their share. A free grad school budget tracker app makes this concrete in about ten minutes. Type in the monthly take-home, type in rent, and look at what is left. It is the most useful ten minutes of your application season.
Watch what the average hides, too. Tuition is usually waived on top of the stipend, but not always, and fees are a different story. Some programs leave students paying thousands in fees every year. Summer money might exist or might not. The headline number is never the whole paycheck.
Use the average as a sanity check, then move on to real offers. The per-program figures at phdstipend.fyi show stipend amounts next to guaranteed funding years and benefits, so you are comparing total packages, not headlines. And run your own numbers through a grad school stipend calculator before you fall for a program. If the math goes negative, that is not a budgeting failure. That is a funding failure.