What a map of the brain costs today, and why the price has not fallen
Dan Siroker, October 18, 2026
How much does it cost to map a brain? Most people I ask guess a few million dollars. The real number is a thousand times larger, and the reason is not the microscope.
The one piece that exists
The largest piece of mammal brain ever mapped at the level of synapses is one cubic millimeter of mouse visual cortex, finished in 2025 by the MICrONS consortium. It came out of a five-year, $100 million program. It contains more than 200,000 cells. The computer traced them, and then people corrected the computer: about a million corrections by human proofreaders, which fully traced only 85 of its neurons, with proofread axons for about 1,400 more. The proofreading by hand is still going on.
A whole mouse brain is about 500 cubic millimeters (Badea et al., 2007, the figure Wellcome's report uses). In 2023 the Wellcome Trust put the cost of mapping one with today's methods at $7.5 to 21.7 billion. Almost all of it is wages: the people correcting the computer's tracing. That is more than the Human Genome Project cost, for one mouse.
Why the price has not fallen
Two precedents say costs collapse when the tool changes.
Reusable rockets cut the cost of a kilogram to orbit about 20 times below the Space Shuttle, and Starlink became possible. A constellation of thousands of satellites was never worth building at Shuttle prices.
The sequencer changed in 2008, and a human genome fell about a million-fold in twenty-two years: from the $300 million draft of 2000 to $525 per genome at NHGRI's centers in 2022. Sequencing became routine medicine, and a sequencing business of about $4 billion a year at Illumina alone. None of it made sense at $300 million a genome.
A cubic millimeter of brain map has fallen less than 10 times since the first measured tracing rate in 2013. Why so little? Because its cost is people tracing by hand, and wages do not fall. Better electron microscopes made the images cheaper. They did not make the correcting cheaper, and the correcting is the bill.
| Today | Our target | |
|---|---|---|
| Microscope | Electron, to resolve the 20 nanometer gap at a synapse | Light, after expanding the tissue sixteen-fold |
| The receptors | A second, correlated imaging step | Proteins imaged directly in the same gel |
| Correcting the tracing | People, by hand | Models trained on human proofreaders' corrections |
| One cubic millimeter | $15 to 43 million | About $200,000 |
| One mouse brain | $7.5 to 21.7 billion; none exists | Under $100 million |
The right-hand column is a goal. Nobody has reached it.
Where the 100x would come from
Two changes, and both are measurable.
Change the instrument. A synapse is a 20 nanometer gap, about ten times finer than a light microscope can resolve. That is why the field used electron microscopes. Expand the tissue sixteen-fold in a gel and the gap becomes 320 nanometers, which an ordinary light microscope can see. Light microscopes cost about a tenth of the electron microscopes used so far, by E11 Bio's estimate, and they can image the proteins, a drug receptor among them, in the same picture. LICONN, published in Nature in 2025, showed this works on small volumes of mouse cortex.
Change the labor. The corrections people made to the public maps were logged: three million edits on the fly brain, a million on the mouse cubic millimeter. About 700,000 of them have been curated into a public benchmark of proofreading decisions. A model trained on those decisions can learn to make them. We trained one and measured how fast it gets better with more data; Teaching a small model to proofread has the numbers. We also ran a repair pipeline on nine held-out cubes of the public mouse map; it closed about a fifth of the breaks, and Repairing the public mouse map says exactly what it did and did not do.
That said, a model that scores well on a benchmark is not a finished brain, and a fifth of the breaks is not all of them. Closing that gap is our core work, and the scaling law says roughly how much data it takes.
Why the cost matters beyond the brain
At $10 billion a brain, one gets made, maybe, by a national program. At $100 million, a company can make one, and the things worth doing with a map become affordable: a study of one compound in one circuit, then whole-brain comparisons, then a map for each disease model. The genome did not change medicine because one existed. It changed medicine because the second one, and the millionth, were cheap.
What would you do with a brain map if it cost $100 million instead of $10 billion? That question is the Market page. If you have an answer, I would love to hear it!
Dan