A microprocessor for biology - millions of tiny experiments, all running at once.
Somewhere in Emeryville, a tube feeds into a chip no bigger than a playing card. Inside it, tens of millions of droplets - each its own private microreactor, each holding a slightly different cell - stream past a sensor that judges them thousands at a time. By the end of the day Triplebar has watched more biological generations unfold than a traditional lab might see in a year. Nobody pipetted them one by one. Nobody could.
This is the company as it stands in 2026: roughly a dozen people, around $45M raised across its history, and a single conviction that biology has been waiting far too long for a faster feedback loop. Triplebar calls its system a "microprocessor for biology." The phrase is doing a lot of work, and most of it holds up.
Discovery in the life sciences runs on a brutal arithmetic. You want the one microbe in ten million that makes a protein cheaply, or the one cell line that grows fast enough to make cultivated meat affordable. Finding it means testing - and testing, the old way, means plates, robots, weeks, and budgets that swallow startups whole. The needle exists. The haystack just refuses to cooperate.
Most of the industry responded by building bigger haystacks-sorting machines. Triplebar's founders looked at the same bottleneck and asked a quieter question: what if you shrank each experiment until you could afford to run all of them? Not a faster search through the haystack - a haystack small enough to read end to end.
Jeremy Agresti founded Triplebar in 2019, betting that droplet microfluidics - the art of turning a fluid into millions of identical, controllable droplets - could be aimed at the messy business of cells. The bet was unfashionable. Microfluidics had promised revolutions before and delivered mostly clogged channels. Agresti, who stayed on as CTO, kept building anyway.
The wager has since widened. Maria Cho led the company through its Series A and public-facing growth; in March 2025 the board named Shawn Manchester, who had joined in 2021, as CEO. The through-line across the leadership changes is the same hunch Agresti started with: that the data the chip produces - genotype matched precisely to phenotype, at enormous scale - is the rarest commodity in modern biology, and the thing AI has been starved of.
The platform is the engine; the products are what come off the line. Triplebar packs picoliter microreactors onto a palm-sized chip, runs cells through directed evolution, and reads the winners. The output is a dataset most labs can't generate at any price - and increasingly, AI models trained on it.
Tens of millions of microreactors on a chip; thousands tested per second to optimize strains and cell lines.
An engineered CHO cell line aimed at higher-performance biologics production.
"CAD for cells" - AI models trained on application-specific datasets for microbial design.
TCR-based T-cell engaging therapies pointed at cancer treatment.
Skepticism is the correct default for any company promising to speed up nature. So here is the evidence, such as it is. Investors put $20M into the Series A in October 2023 - Synthesis Capital leading, with Essential Capital, Stray Dog Capital, iSelect Fund, and The Production Board alongside. Two named commercial partners signed on for very different products. And the throughput claim, the one everything rests on, is large enough to be worth scrutinizing.
Triplebar frames itself as a B2B product design engine "bringing abundance to problems of scarcity." Translated: the same machine that helps make a cancer therapy can help make protein cheaper to grow, and the company would like to do both rather than pick. The longer-range ambition is more audacious - generative AI genome models trained on its data, what it openly calls "a predictive model for life itself."
Return to that tube feeding into the palm-sized chip in Emeryville. A year ago the droplets streaming through it were, in a sense, just data being born. Today that data trains models, and the models propose the next batch of droplets to make. The loop has started closing on itself - experiments suggesting experiments, biology beginning to behave a little more like the software Triplebar always insisted it could be.
Whether that loop produces a cheaper protein, a better cancer drug, or a genuinely predictive model of living systems is still unsettled, and a dozen people in California will not decide it alone. But the bottleneck they set out to break - biology's stubborn, expensive slowness - is, on that one chip, measurably looser than it was. The haystack got smaller. The needle got easier to find. Everything else is just scale.