Caffeine in Coffee Calculator

The whole calculation is one multiplication

Volume times a caffeine density, times the number of servings. Everything interesting about this page is in that density, because it is the part that varies enormously and the part a calculator has to freeze.

caffeine mg = fluid ounces × mg per fluid ounce × servings

Where each density comes from, and which ones are ours

Five drinks, five different provenances. Three come from USDA FoodData Central, one is inside USDA's published precision, and one has no published source at all.

• Brewed drip, 12 mg per fl oz. USDA FoodData Central entry 2710375 gives 40.0 mg per 100 g with an 8 fl oz portion weighing 240 g, which is 96 mg a cup and 12 mg an ounce exactly.
• Espresso, about 64 mg per fl oz. USDA entry 2710378 gives 212 mg per 100 g with a fluid ounce weighing 30 g, which is 63.6 mg. The 64 is rounded, not exact.
• Instant, 7.5 mg per fl oz. USDA gives 26 mg per 100 g for reconstituted instant coffee, which works out to 7.8 mg an ounce. This page is about 4 percent low, and since USDA publishes 26 as a whole number the true band is roughly 7.65 to 7.95, so 7.5 sits just outside it.
• Decaf, 0.4 mg per fl oz. USDA gives 1 mg per 100 g, a single significant figure, so anything from 0.15 to 0.45 mg an ounce is consistent with it and 0.4 is inside. The FDA independently puts decaf at 2 to 15 mg in an 8 oz cup, and 0.4 falls in that too.
• Cold brew, 18 mg per fl oz. This one is ours. USDA has no cold brew entry at all: a search of every food in the 2021-2023 and 2019-2020 FNDDS releases returns zero matches, against 22 for other brewed coffees. The nearest published measurement, from a 2019 study in Food Research International, is 1.25 mg/mL for undiluted cold drip, which is about 37 mg an ounce, roughly double this figure. That preparation was a concentrate and commercial ready-to-drink cold brew is usually diluted, so 18 is not refuted so much as unsupported. It is this page's estimate.

A fixed average is the wrong shape for this quantity

This is the part worth reading before you trust the number above. Caffeine in coffee varies more than almost anything else a calculator on this site handles.

The sharpest finding comes from McCusker and colleagues in the Journal of Analytical Toxicology in 2003, who bought THE SAME DRINK FROM THE SAME OUTLET ON SIX CONSECUTIVE DAYS and measured 259 to 564 mg per dose. That is a 2.2-fold swing with the shop, the drink and the order held constant. Across different specialty coffees they found 58 to 259 mg per dose.

Crozier and colleagues in Food & Function in 2012 bought one espresso from each of twenty outlets and found 51 to 322 mg a serving, and put the consequence well: "one cup of high-caffeine coffee could cause as much difficulty to these susceptible consumers as six cups of coffee to another".

The FDA publishes its own range rather than a single figure: regular brewed non-specialty coffee at 113 to 247 mg per 12 fluid ounces. This page's 12 mg an ounce works out to 144 mg for the same 12 ounces, which sits inside that band about a quarter of the way up. So the default here is defensible and it is also toward the low end of what the FDA itself publishes.

Bean, roast and method, and which of those folk beliefs survives

Species is the biggest single lever. Roasted arabica runs about 1.2 to 1.5 percent caffeine by dry weight and roasted robusta about 1.5 to 3.0 percent, so robusta carries roughly twice as much. A blend's robusta fraction is rarely on the bag.

Method matters, measurably. A 2019 study brewed eight methods from the SAME raw material, which isolates method from bean: 42.8 mg a cup at the low end up to 149.5 for cold brew, a 3.5-fold spread, and an eight-fold spread per millilitre. Read those per-cup figures with care, because the authors adopted a typical volume for each style rather than measuring a serving.

Roast is the one everybody gets wrong, and the crux is how you measure. By weight the literature does not agree there is an effect at all: a 2024 paper in Scientific Reports notes that across more than twenty studies "the variety and disagreement among published results is staggering" and that the most commonly reported result is that roast makes no difference, and a 2022 study in Foods found "neither processing nor roasting affected caffeine content". By SCOOP light roast wins clearly, and not because dark roast loses caffeine: roasting drives off water and puffs the bean, so a scoop of dark roast holds fewer beans. It is a density effect. If you weigh your coffee the question mostly disappears.

What the FDA actually says about 400 mg

The wording matters and it is routinely upgraded in the retelling. The FDA "has cited 400 milligrams a day, that's about two to three 12-fluid-ounce cups of coffee, as an amount not generally associated with negative effects" for most adults. That is a statement about the absence of observed harm. It is not a recommendation, the agency does not call it a limit, and this page will not either.

Two details that usually go missing. The coffee equivalence is two to three 12-ounce cups, not the four or five 8-ounce cups often quoted. And the FDA attaches a caveat in the same breath: "there is wide variation in both how sensitive people are to the effects of caffeine and how fast they eliminate it from the body".

How long it stays, and why the same cup lands differently on two people

EFSA's draft opinion on caffeine puts the plasma half-life at "about 4 hours with range of about 2 - 8 hours". Most of the clearing is done by one enzyme: CYP1A2 accounts for about 95 percent of it, which is why the things that change that enzyme change everything.

• Pregnancy lengthens it substantially. EFSA cites half-life rising to a range of 6 to 16 hours in pregnancy, returning to 2 to 8 within weeks of delivery, and notes it is three to four times longer at the end of pregnancy than in the non-pregnant state.
• Smoking shortens it, because the polycyclic aromatic hydrocarbons in smoke induce the enzyme; one review reports the half-life falling from 6.0 to 3.5 hours.
• Combined oral contraceptives roughly halve clearance, one study reporting a 55 percent reduction after a treatment cycle.
• Advanced liver disease slows it, with elimination reduced by about half in Child C cirrhosis.
• Genotype contributes but explains less than people assume: the much-discussed CYP1A2 variant accounted for 14 and 22 percent of the variation between people in two cohorts. It is a factor, not a verdict.

Working one out by hand

A 16 oz brewed coffee and a double espresso.

1. The coffee. 16 × 12 = 192 mg.
2. The double. 2 × 64 = 128 mg.
3. Total. 192 + 128 = 320 mg, which is 80 percent of the 400 mg the FDA cites.

Now the honest part. Put McCusker's same-shop range around that 16 oz coffee and it could plausibly be anywhere from well under 192 to several times it. The arithmetic above is exact; the input it multiplies is the estimate.

Edge cases

• Decaf is not caffeine-free. The FDA puts a decaf cup at 2 to 15 mg. Several a day is not nothing if you are sensitive.
• Espresso per ounce versus per drink. Espresso has more than five times the caffeine of drip per ounce and comes in an ounce or two, so a single shot usually carries LESS total caffeine than a mug of drip. Per ounce and per drink point opposite ways.
• Cold brew. Concentrate and ready-to-drink are different products by a factor of two or more, and this page has one number for both.
• Pregnancy. The FDA publishes no number and defers to your clinician. ACOG's 200 mg figure is a negative finding rather than a ceiling: moderate consumption below it "does not appear to be a major contributing factor in miscarriage or preterm birth", and above it ACOG explicitly declines to make a recommendation because two large studies disagree. EFSA lands on the same 200 mg independently.
• A shop's own figures beat this page. If the chain publishes caffeine per drink, use that. It is measuring its own coffee; this page is multiplying an average.

Your figures stay in this page

The drink, size and servings you choose are computed by this page in your browser. Nothing is sent to a server, nothing is stored after you close the tab, and there is no account.