The most frequently cited estimate puts the human brain’s storage capacity at roughly 2.5 petabytes, or about 2.5 million gigabytes. That figure comes from a widely discussed 2016 study by researchers at the Salk Institute, who measured synapse sizes in the hippocampus and extrapolated from there. But the number papers over enormous uncertainty about how the brain actually encodes information, and some researchers think the true capacity could be far higher or meaningfully lower depending on which biological structures you count.
Where the 2.5 Petabyte Number Comes From
The headline figure traces back to a single insight about synapse size. Researchers discovered that synapses in the hippocampus come in about 26 distinguishable sizes rather than the handful previously assumed. Since each distinct size can represent a different state, each synapse could theoretically store around 4.7 bits of information rather than the 1 or 2 bits older models assumed. Multiply that by the estimated number of synapses in the entire brain, and you land somewhere around a petabyte.
The calculation depends heavily on how many neurons and synapses you start with. The figure of 86 billion neurons has been treated as settled fact for over a decade, but a recent statistical reanalysis argues the actual count is uncertain, with estimates ranging between 61 and 99 billion depending on the study and methodology used.1PubMed Central. Eighty-six billion and counting: do we know the number of neurons in the human brain? Each neuron forms thousands of synaptic connections, so the total synapse count varies widely too. Most estimates land between 100 trillion and 600 trillion synapses. Plug different synapse counts and bits-per-synapse assumptions into the same equation, and you can get answers ranging from hundreds of terabytes to multiple petabytes. The 2.5 petabyte figure is a reasonable midpoint, not a measurement.
What Mapping a Tiny Piece of Brain Actually Took
One way to appreciate the brain’s density is to look at what happens when researchers try to digitally reconstruct even a tiny fragment of it. In 2024, a team from Harvard and Google published a nanoscale reconstruction of a single cubic millimeter of human temporal cortex, roughly the size of a grain of sand. That speck contained about 57,000 cells, roughly 150 million synapses, and about 230 millimeters of blood vessels. The raw imaging data for that one cubic millimeter totaled 1.4 petabytes.2PubMed Central. A petavoxel fragment of human cerebral cortex reconstructed at nanoscale resolution
That number is the amount of data needed to describe the physical structure of the tissue at electron-microscope resolution, not the amount of information the tissue itself is “storing.” But it offers a striking sense of scale. The whole brain contains roughly a million times more volume than that single cubic millimeter. Mapping it all at the same resolution would produce data measured in exabytes, a scale that dwarfs the storage capacity of any existing data center. The brain packs an almost absurd amount of physical complexity into a very small space, and that complexity is part of what makes estimating its information content so slippery.
Why the Gigabyte Analogy Breaks Down
Asking how many gigabytes the brain holds is a bit like asking how many horsepower a river has. You can force the comparison, and the resulting number might even be useful in a rough way, but the underlying systems work so differently that the answer always comes with a long list of asterisks.
A computer stores data as fixed strings of ones and zeros in clearly defined locations. Every bit sits at a specific address, and retrieving it means going to that address and reading what is there. The brain does not work this way. Memories are distributed across networks of neurons, with the same neuron participating in many different memories simultaneously. There is no fixed “address” for a childhood birthday party; instead, fragments of the experience are encoded in patterns of synaptic strength, timing, and connectivity spread across multiple brain regions. This means the brain can lose substantial numbers of neurons and still retain a functional memory, something a hard drive absolutely cannot do if you start erasing sectors.
The brain also compresses aggressively. Research suggests that when people encode sequences of information, they rely on something resembling a compression algorithm, representing patterns using a kind of internal shorthand. The psychological complexity of a sequence tracks closely with the length of the shortest formula that captures its structure, hinting that the brain stores compressed descriptions rather than raw recordings of experience.3PubMed Central. A theory of memory for binary sequences: Evidence for a mental compression algorithm in humans This compression means the “useful” information stored in the brain could be far smaller than the raw physical capacity of its synapses would suggest, in the same way that a ZIP file takes up less space than the folder it was created from.
Storage Mechanisms Beyond Synapses
The petabyte estimates treat synapses as the sole units of storage, but there is growing evidence that other biological structures contribute to memory in ways that could change the math substantially.
One surprising avenue is epigenetics. DNA methylation, a chemical modification to DNA that does not change the genetic code itself, appears to play a role in forming and maintaining long-term memories. Researchers have found that memory formation requires lasting changes in methylation patterns within neurons, suggesting that the genome inside each neuron may serve as a form of storage medium.4PubMed Central. DNA methylation and memory formation Histone modifications, another layer of chemical tags on the proteins that DNA wraps around, also appear to regulate the gene expression changes underlying long-term memory.5PubMed. Histone lysine methylation: critical regulator of memory and behavior If each neuron can store information not just in its synaptic connections but also in chemical modifications to its own DNA and chromatin, the total information capacity of the brain could be orders of magnitude larger than synapse-only models predict.
An even more speculative idea involves the internal scaffolding of neurons themselves. Microtubules, the protein tubes that form the structural skeleton inside each cell, have been proposed as potential information-encoding substrates. In one computational model, an enzyme called CaMKII can phosphorylate tubulin proteins in microtubule lattices in patterned ways, potentially encoding information at a sub-cellular scale. Depending on the geometry assumed, a single enzyme-microtubule interaction could represent anywhere from 64 to over 5,000 unique states.6PLOS Computational Biology. Cytoskeletal Signaling: Is Memory Encoded in Microtubule Lattices by CaMKII Phosphorylation? This remains a hypothesis rather than an established finding. But if even a fraction of it holds up, it would mean storage is happening at a spatial scale far below what synapse-based estimates capture, and the brain’s total capacity would need to be revised sharply upward.
The Cells You Probably Are Not Counting
Most storage estimates focus exclusively on neurons, but the brain contains roughly as many glial cells as neurons, and a growing body of evidence suggests these non-neuronal cells are not just passive support staff. Astrocytes, the most abundant type of glial cell, actively influence synaptic transmission, neuronal communication, and even memory consolidation. They wrap tightly around synapses and release signaling molecules that modulate how strongly those synapses fire.7PubMed. Astrocytes as new targets to improve cognitive functions
Research has shown that astrocyte signaling contributes to synaptic plasticity, network-level oscillations, and memory performance. Inactivating or boosting astrocyte function directly affects cognitive abilities in animal models, and astrocytes appear to mediate the effects of arousal and vigilance on how well memories are formed.8PubMed. Astrocyte function from information processing to cognition and cognitive impairment If astrocytes are actively participating in how the brain processes and retains information, they represent a large population of cells whose contribution to total storage capacity is simply absent from conventional gigabyte estimates. Researchers do not yet have a clear way to quantify how much information an astrocyte contributes, which is one more reason the headline numbers should be taken as rough starting points.
How Much of What You Learn Actually Sticks
Even if the brain’s raw capacity is measured in petabytes, the amount of information it retains at any given moment is dramatically less. Forgetting is not a bug but a core feature of how the system works. The forgetting curve, first described by Hermann Ebbinghaus in the 1880s, shows that memory for newly learned material drops steeply in the first hours and days before leveling off. Modern replications confirm the basic shape of the curve, including a noticeable bump upward in retention around the 24-hour mark, likely reflecting the effects of sleep-based memory consolidation.9PubMed Central. Replication and Analysis of Ebbinghaus’ Forgetting Curve
This aggressive pruning means your brain is not filling up its storage the way a hard drive gradually fills a progress bar. It is constantly discarding, overwriting, and reorganizing. Two people with identical brains (if such a thing existed) who experienced the same events would hold very different subsets of those events in memory a week later. The effective storage in use at any moment is a fraction of the theoretical maximum, sculpted by attention, emotion, repetition, and sleep.
A small number of people seem to defy this pattern. Individuals with highly superior autobiographical memory, or HSAM, can recall specific personal events from decades ago with remarkable accuracy. In studies, they achieve near-perfect reliability on date-cued memory tests, with verifiable accuracy measured at about 98 percent. One participant passed HSAM assessment at ages 75 and 80, and their performance actually improved over the five-year gap between testing sessions.10PubMed Central. Highly Superior Autobiographical Memory (HSAM): A Systematic Review HSAM is extremely rare, and the underlying mechanism is not well understood. But its existence suggests that the brain’s storage limits may be far more generous than ordinary forgetting implies, and that in most people, the bottleneck is not capacity but retrieval and retention strategy.
Efficiency, Not Just Size
Framing the brain purely in terms of storage misses what may be its most impressive feature: how much it accomplishes with how little energy. The brain runs on about 20 watts, roughly the power draw of a dim light bulb. One analysis of cortical energy budgets found that communication between neurons consumes about 35 times more energy than the computation neurons perform, yet both costs are necessary to predict how many synaptic transmissions a neuron needs to fire.11PubMed Central. Communication consumes 35 times more energy than computation in the human cortex, but both costs are needed to predict synapse number The brain’s architecture is optimized not for maximum storage but for efficient, low-power information processing.
This is part of why raw neuron count does not straightforwardly predict intelligence across species. Among mammals, the best predictors of cognitive ability are a combination of cortical neuron count, how densely those neurons are packed, the distance between them, and the speed at which signals travel along their axons.12PubMed Central. Neuronal factors determining high intelligence Intelligence, in other words, depends on the quality and speed of the network, not just its size. Similarly, artificial neural networks modeled loosely on the brain do not match its capabilities simply by scaling up. One analysis noted that the brain’s intelligence depends more on neuronal type and energy supply than on the sheer number of neurons.13PubMed Central. How can artificial neural networks approximate the brain? The brain is not a hard drive with a processor bolted on; it is a system where storage, computation, and communication are deeply entangled at every level.
The Computer Metaphor and Its History
People have always described memory using the most advanced information technology of their era. Ancient Greeks compared memory to writing on a wax tablet. Medieval thinkers imagined a storehouse. The Victorians favored the metaphor of a palimpsest, a manuscript scraped clean and rewritten. Today the dominant metaphor is the computer, which is why the gigabyte question feels so natural.14New Advances in Brain & Critical Care. Memory: From Ancient Art to Synaptic Science: A Discursive Exploration of the Cultural History and Contemporary Neurobiology of Human Memory
Each metaphor shaped how scientists and the public thought about how memory works, and each turned out to be misleading in its own way. The wax tablet metaphor implied memories are stamped in permanently and decay passively, which is wrong: the brain actively maintains, edits, and reconstructs memories every time they are recalled. The storehouse metaphor implied memories are discrete objects filed away in fixed locations, which is also wrong. The computer metaphor has its own blind spots. It encourages people to think of the brain as having a fixed capacity that gradually fills up, a distinction between hardware and software, and a clear separation between storage and processing. None of these map cleanly onto the biology.
The metaphor’s persistence owes something to its partial usefulness. It is true that the brain has finite physical resources, that information can be encoded and retrieved, and that damage to specific regions impairs specific functions. But treating the brain as a very complicated computer has consistently led researchers astray when pushed too far, particularly in predicting how artificial systems need to scale to replicate biological intelligence.
How Biological Molecules Compare to Silicon
If you want to understand just how dense biological information storage can be, consider DNA. A single gram of DNA can theoretically hold about 215 petabytes of data, and its three-dimensional structure makes it roughly eight orders of magnitude denser than conventional storage media like flash drives or magnetic tape.15PubMed Central. DNA Data Storage DNA achieves a storage density of about 10 to the 19th bits per cubic centimeter, a figure that makes even the densest commercial hard drives look comically sparse.
The brain does not use DNA the way a hard drive uses magnetic platters; it is not writing and reading sequential data files. But the comparison underscores an important point. Biological molecules pack information at scales that digital technology cannot currently approach. If even a fraction of the brain’s molecular machinery, including the epigenetic modifications and intracellular structures discussed earlier, participates in encoding information, the total capacity of the organ may be vastly larger than any estimate based purely on synapse counting would suggest. Researchers working on DNA-based data storage are actively trying to harness this density for archival purposes, using synthetic DNA strands to encode digital files. The brain, in a sense, has been doing something loosely analogous for hundreds of millions of years, just not in a way that maps neatly onto gigabytes and terabytes.
What a Practical Number Looks Like
For anyone who arrived here wanting a number they can use in conversation: the most defensible statement is that the brain’s information capacity is probably in the low petabytes, somewhere between one and ten petabytes depending on the assumptions. In more familiar units, that is roughly one to ten million gigabytes. The 2.5 petabyte figure you see repeated everywhere is based on real research and reasonable assumptions, but it represents one specific model of how synapses encode data, and it ignores several biological systems that may also contribute.
If you are comparing the brain to a device you own, a petabyte is about 500 times the storage of a high-end laptop and about 1,000 times the storage of a typical smartphone. The comparison is fun but not particularly illuminating, because the brain does not use its capacity the way your phone uses its flash memory. Your phone stores every photo at full resolution until you delete it. Your brain stores a compressed, abstracted, emotionally weighted, constantly edited version of experience, throws most of it away within days, and reconstructs the rest from partial cues every time you remember. The gigabyte number captures the physical substrate. It says almost nothing about the richness, flexibility, or strangeness of what the brain actually does with that substrate.