The 39 trillion: what we actually mean when we say “the microbiome”

For a long time the headline number was “ten to one” — ten bacterial cells for every human one. It was repeated so often it became folklore. When researchers finally recounted, the ratio collapsed to roughly one to one: about 38–39 trillion microbial cells against 30-odd trillion human ones. Not ten times us. About the same as us.

That correction matters less for the arithmetic than for what it revealed: a science still establishing its basic measurements. The microbiome is the collection of microorganisms — bacteria, archaea, fungi, viruses — living on and inside the body, together with their genes and the chemistry they produce. The densest neighborhood is the colon, but there are distinct communities on the skin, in the mouth and elsewhere, each as different from the next as a rainforest is from a tidal flat.

It’s an ecosystem, not a scoreboard

The most useful shift in thinking is to stop counting and start mapping. A gut community isn’t “good” because it has more bacteria; a clogged drain has plenty of bacteria. What characterizes a resilient microbiome is closer to what characterizes a healthy ecosystem: diversity of species, redundancy of function, and the ability to bounce back after a disturbance such as a course of antibiotics.

The right mental model isn’t a supplement you top up. It’s a forest you tend — and forests are defined by their relationships, not their headcount.

This is why two healthy people can share almost no species. The field has moved from asking which microbes are present to asking what they do — which metabolic jobs get done, regardless of which organism does them.

Why now

Three things changed. Sequencing got cheap enough to read whole communities instead of culturing one bug at a time. Computation got good enough to make sense of the data. And the questions got sharper — from “is this bacterium present” toward “is this specific, defined strain doing this specific, measurable thing.” That last move, from genus-level guesses to strain-level precision, is where most of the interesting work now lives.