The State of Frontier Neuroscience: This is the century of the brain.

Yet for decades, neuroscience was the graveyard of biotech. The biology was too complex, timelines too long, and failure rates too brutal. Most of the field stopped believing startups could reliably produce venture-scale outcomes, let alone produce any returns at all. “Always curious to back great founders but I still think we don’t understand how the F the brain works”, said a tier 1 VC.

They were right, but then everything changed.

Scientific breakthroughs, pharma demand, and a worsening global brain health crisis began converging. We believe the next decade will produce companies that do for the brain what GLP-1s did for metabolism: transform massive categories once considered intractable and unattractive into some of the most valuable markets on earth.

Three things changed simultaneously: the toolkit dramatically improved, the need became unavoidable, and pharma is swarming.

The first shift is technological.

For years, neuroscience relied on blunt tools: subjective endpoints, broad small molecules, monoclonal antibodies chasing plaques, and limited visibility into what was actually happening in the brain.

That’s changing fast:

• Blood-based biomarkers now flag Alzheimer’s-related proteins years before symptoms appear from a single draw, shifting the game from damage control to prevention.

• Shuttles are finally getting drugs across the blood brain barrier and into the brain.

• Deep sequencing is surfacing new targets.

• RNA and oligos are maturing into a predictable, programmable platform.

• Teams are building preventative therapies, even neuro vaccines.

• Materials engineering is producing BCIs that are miniaturized, wireless, and biohybrid.

• AI agents enable near-zero cost care delivery.

The result: neuro is becoming data-rich, but it’s still insight poor. The real shift is what happens when wearables, genomics, imaging, and connectomics stack on top of each other and AI starts surfacing relationships between variables no one thought to look for.

The need is obvious. Mental and neurological disorders are among the largest sources of disability and economic burden in the world. Depression and anxiety cost the global economy roughly $1 trillion annually in lost productivity. Dementia affects more than 57 million people globally, a number expected to rise dramatically over the coming decades. Suicide remains one of the leading causes of death among young people.

These trends are connected: aging populations increase rates of neurodegeneration, but environmental toxins, chronic stress, loneliness, economic and political instability, and social media overuse are contributing as well. The brain is responding to the environment society has built around it. It’s almost like we’ve actively built as hostile an environment as possible for the brain.

At the same time, pharma has reversed course on neuroscience. After the SSRI era, many large pharmaceutical companies abandoned CNS entirely. Pfizer and AstraZeneca scaled back neuroscience divisions after years of expensive failures. Neuro became known as a category where timelines stretched endlessly and mechanisms collapsed in Phase 3.

Now the buyers are back.

There’s been over $65B in neuro M&A in the past 3 years. Johnson & Johnson acquired Intra-Cellular Therapies for $14.6 billion. Bristol Myers Squibb acquired Karuna for $14 billion. AbbVie acquired Cerevel for $8.7 billion and later acquired Gilgamesh’s psychedelic program for up to $1.2 billion. Lilly acquired Centessa Pharmaceuticals for up to $7.8B. Lundbeck acquired Longboard Pharmaceuticals for $2.6 billion.

These are strategic decisions by companies facing major patent cliffs and limited pipeline replacement options.

Capital follows exits. Founders follow capital. The category is self-reinforcing now. For the first time, the field is catching up to the ambition.

The most important shift in neuro is that multiple categories are beginning to work simultaneously. Here are a few examples:

Neurotech

Brain-computer interfaces have moved into the clinic. Neuralink’s first human patient demonstrated cursor control and gameplay through thought alone. Synchron is pursuing a less invasive endovascular approach, deploying its interface through blood vessels rather than open brain surgery. Motif Neurotech is applying minimally invasive stimulation to treatment-resistant depression.

Meanwhile, non-invasive neurotech is advancing rapidly. Companies pursuing focused ultrasound, transcranial stimulation, and wearable neuromodulation are expanding what can be treated without surgery. Flow Neuroscience received FDA approval for depression. Neurode has built the equivalent of a mobile fMRI that can personalize stimulation for focus.

The companies we find most interesting are able to record and modulate the brain with precision.

Programmable Medicines

For decades, many diseases were biologically understood but therapeutically unreachable. That gap is beginning to close as medicines become more programmable.

Genetic medicines come with clearer clinical readouts and less difficult regulatory paths. Antisense oligonucleotides and RNA therapeutics allow researchers to intervene directly at the level of gene expression. Ionis and Biogen helped establish the category in diseases like Spinal Muscular Atrophy. A new generation of companies like PsyMed portcos Aerska, Sculpta, and Iris Medicine are now applying similar approaches to Huntington’s, ALS, and broader neurological indications.

The right frame isn’t which modality wins – it’s a toolbox. Multiple approaches will run concurrently. The different pieces will come together to attack each disease in the unique way it needs.

Psychedelics

Psychedelics have crossed an important threshold: they became acquirable.

AbbVie’s acquisition of Gilgamesh’s bretisilocin program marked the first major pharma acquisition of an investigational psychedelic asset. Otsuka followed with its acquisition of Transcend. Meanwhile, Spravato has become one of the fastest-growing products in psychiatry at a ~$2B revenue run rate.

The signal from these deals shows that pharma is willing to engage when the IP is defensible, the indication is defined, the trial design is conventional, and the treatment model fits existing infrastructure.

Computational Biology

The brain generates more complex biological data than any other organ – across genetics, imaging, behavior, and clinical outcomes simultaneously. For most of biotech’s history, that complexity was a liability. Now it’s becoming an asset.

A new generation of computational biology companies are building models designed to find signal in exactly this kind of high-dimensional data. Recursion uses large-scale imaging and AI-powered neural networks to identify disease mechanisms and therapeutic targets at unprecedented resolution, with neuroscience as one of its core focus areas. Isomorphic Labs, DeepMind’s drug discovery spinout, announced collaborations with Eli Lilly and Novartis worth nearly $3 billion, using AlphaFold-derived structural biology to accelerate target identification across therapeutic areas including neuro.

The brain has been hard to drug partly because it has been hard to model. The volume of data is finally growing rich enough, and the compute sophisticated, cheap, and scaled – to start finding relationships that decades of bench science missed.

Brain Foundation Models

The scaling logic that produced language models is now being pointed at the brain itself. Instead of training on text, these models train on raw neural activity to learn a general representation of how the brain encodes the world, one that transfers across subjects, regions, and tasks instead of being rebuilt for each new dataset.

The paradigm has passed its first real proof point: a foundation model trained on activity from the visual cortices of many mice could predict neuron responses to videos, generalize to new animals with little additional training, and recover neuron types and connectivity from the same model.

Companies are forming around this. Piramidal, a YC-backed startup, is building an EEG foundation model trained on brainwave data, using epilepsy diagnosis as the wedge and piloting at hospitals. On the research side, BrainLM and Brant extend the approach to fMRI and intracranial recordings, and there are now dozens of pretrained EEG transformers competing on cross-subject benchmarks.

The binding constraint is data. Most large neural data sets being passive clinical recordings collected for other purposes, not built to expose cognition. The companies we find most interesting either own a data engine that generates the right recordings, or have a downstream task where a general neural prior measurably outperforms the bespoke alternative. Preferably both.

Beyond medicine: human enhancement

The same biological systems that govern disease also govern performance. Sleep, cognition, empathy are measurable, modifiable, and increasingly understood at the molecular level.

Take sleep, for example. UCSF neuroscientist Ying-Hui Fu identified a mutation in the DEC2 gene carried by natural short sleepers – people who function fully on six hours of sleep with none of the cognitive or health deficits that affect the rest of us. The implication is that optimal sleep isn’t just the absence of a disorder. It’s a biological state that some people access naturally and that medicine might one day confer deliberately. Orexin agonists are the first serious pharmaceutical attempt to get there. Multiple programs are in active development for narcolepsy, with obvious extension into broader wakefulness optimization. Meanwhile, MDMA analogs are being studied for their effects on social cognition and empathy, not just as PTSD treatments.

The precedent is GLP-1s. Ozempic didn’t just treat obesity; it revealed a far larger market of people who simply want to lose a few pounds. The same logic applies to the brain. The collective market cap of neuro today reflects only the diseases we currently treat. It does not yet reflect what becomes possible when the target is human performance itself.

Recent exits reveal a consistent pattern: pharma pays for proof, not potential. But proof takes different forms at different stages.

The biggest prices came with promising clinical data. Bristol Myers Squibb paid $14 billion for Karuna because the Phase 3 efficacy was undeniable. AbbVie paid $1.2 billion for Gilgamesh on Phase 2 data because the program fit cleanly into an existing psychiatric framework – short duration, in-clinic model.

But pharma will move earlier when the mechanism is airtight. Biogen paid Ionis for antisense assets before human readouts because the platform itself was de-risked. Isomorphic Labs signed billion-dollar collaborations with Lilly and Novartis on platform promise alone.

The lesson is that certain categories are finally becoming legible.

Anyone can read the papers, attend the conferences, and track the FDA calendar. The investments that matter come out of years of conversations with researchers still in the lab, clinicians watching what works and what doesn’t, and founders who haven’t decided to start anything yet. You either built that network early or you’re too late to it.

But access is only part of it. The other ingredient is accumulated judgment. Specialists know which mechanisms have graveyards and why the prior attempts failed. They know which endpoints actually move the needle in the clinic versus which ones look clean in an academic paper and collapse in a pivotal trial. They know what pharma will actually pay for.

The brain has always been the hardest frontier in medicine. For decades, the honest answer to “can we treat this?” was often no. Not yet. Maybe never.

That answer is changing. We now have the tools: programmable medicines that can intervene at the genetic level, closed-loop devices that listen to the brain and respond in real time, multimodal data platforms that can find signal in biology’s most complicated system, biomarkers that catch disease before it’s irreversible. The exits are real. The mechanisms are validated. The buyers are lining up.

The field isn’t throwing stuff at the wall anymore. It’s hitting bullseyes.