Twist Bioscience Guides to Breakeven Adjusted EBITDA This Quarter, Targets Revenue Doubling by 2031 as AI-Driven Drug Discovery Orders Surge
Investor Day, May 21, 2026 — Wilsonville, Oregon manufacturing facility
Twist Bioscience used its first large-scale investor day in years to lay out a financial roadmap that calls for adjusted EBITDA to turn positive in the current fiscal fourth quarter, gross margin to climb above 60% over time, and revenue to more than double from the fiscal 2026 exit rate of $442 million to $447 million to roughly $900 million by fiscal 2031 — a target management repeatedly described as "a floor, not a ceiling." CFO Adam Laponis told the room that the company has held operating expenses "relatively flat for the last 3 years" while revenue compounded at 22% annually since 2023, a dynamic he said will continue as the manufacturing platform scales largely within existing facilities.
AI drug discovery orders inflect sharply higher
The most closely watched new data point was the trajectory of Twist's AI-enabled drug discovery business. CEO Emily Leproust disclosed that orders from AI-native drug discovery customers totaled $25 million last year and are tracking toward triple-digit percentage growth this year, implying order volume could exceed $50 million in fiscal 2026. Chief Scientific Officer Colby Souders laid out the economics of this emerging customer cohort in granular detail: a typical AI partner starts with a pilot study of tens to hundreds of sequences costing $10,000 to $100,000, graduates to training rounds of thousands of sequences for under $1 million, and eventually moves into therapeutic discovery campaigns spanning hundreds of thousands of candidates. Twist said it has already produced over 200,000 proteins, assayed more than 130,000 of them, and generated 7 million data points for "dozens of customers" in this category alone.
The company raised its 2030 total addressable market estimate to $13 billion, up from $7 billion today and $2 billion in 2020. Of that figure, Twist is now modeling an incremental $500 million by 2030 specifically tied to AI-driven antibody discovery services, layered on top of organic growth in its existing DNA synthesis, antibody discovery, and protein expression markets. Management was explicit about the risk embedded in that number: the forecast assumes no single dominant AI model emerges and that AI-driven drug discovery continues to work as a modality. "If it didn't work, then the market will not be as big," Leproust said.
Customer testimony reinforced the model-performance narrative underpinning that bet. Joshua Meier, co-founder and CEO of Chai Discovery — a company that has raised roughly $250 million and runs no wet lab of its own — said antibody design success rates went from 0.1% in May 2025 (1 in 1,000 designed antibodies binding in the lab) to approximately 16% with its Chai-2 model published in June of that year, a more than 100-fold improvement in roughly twelve months. Meier called the resulting feedback loop between model iteration and DNA synthesis speed "a really nice flywheel," crediting Twist's shrinking turnaround times for enabling faster experimental validation.
Synthesis economics improved far more than previously disclosed
CTO Siyuan Chen provided the most granular operational disclosure of the day, quantifying three years of platform improvement that underpins the company's gross margin trajectory. Oligo synthesis cost fell 60% between 2023 and 2026, with solvent consumption per million 100-mer oligos dropping from 51 liters to 14 liters — meaning a single 500 ml bottle of chemical reagent now yields roughly 35,000 oligos. Turnaround time for the same production run fell from 26 hours to 7 hours, a 73% reduction that Chen said translated into a 4x increase in effective capacity using the same number of synthesis instruments ("riders," in Twist's internal terminology). Error rates held at roughly 1 in 3,000 base pairs, with best runs reaching 1 in 4,000 — a level Chen asserted is "really unheard of" relative to traditional column-based synthesizers.
The company has also pushed synthesis length out to 500 base pairs in production, versus an industry norm closer to 80-100 base pairs, enabling the recently launched Ultra-Complex Gene product that Chen said serves as the backbone for complex mRNA constructs. Automation consolidation has been aggressive: Twist now runs 20 integrated production systems on its manufacturing floor, up from 7 just two years ago, while cutting the footprint of its original gene assembly line by 80% and doubling fragment output in the process.
MRD panel turnaround collapsed from weeks to under five days
On the diagnostics side, SVP of Product and Marketing Jimmy Jin detailed a structural shift in molecular residual disease (MRD) testing economics. Twist redesigned its MRD panel production to run entirely on its silicon chip, cutting turnaround from one week to a single day while increasing capacity tenfold with fixed headcount — generating more than $2 million in first-year consumables savings alone. Jin argued the math favors targeted panels over whole-genome approaches even as sequencing costs fall: moving from a standard cancer panel to whole-genome sequencing requires roughly 1,500 times more sequencing data than a targeted panel, meaning a sixfold reduction in per-gigabase sequencing costs since 2021 still leaves whole genome sequencing dramatically more expensive for this application. Jin estimated the industry will need to produce 4.6 billion oligos over the next two years just to meet projected MRD panel demand, assuming roughly 2.3 million tests annually — a volume he said competitors using standard oligo synthesis infrastructure (150,000 oligos per day typical capacity) simply cannot meet without a tenfold capacity expansion.
Regeneron makes the case against whole-genome sequencing at scale
In one of the day's more pointed customer presentations, John Overton, Chief Sequencing Officer of the Regeneron Genetics Center, used UK Biobank data — a uniquely matched cohort of 500,000 participants with genotyping, exome sequencing, and whole-genome sequencing all performed — to argue that targeted sequencing remains the more efficient drug-discovery tool despite falling whole-genome costs. Comparing coding-region variant detection, Overton showed that imputed genomes (combining genotyping arrays with Twist's exome capture) and true whole genomes produced a 97% overlap in detected coding variants, with nearly identical downstream trait-association discovery rates across 100 clinical traits tested. At matched budgets rather than matched sample counts, Overton said a lab running exome-plus-genotyping at roughly 3x lower cost than whole genome sequencing generates about 5x more usable research results on the same spend — a gap that widens to 20x at a 10x cost advantage. Regeneron worked with Twist to co-develop a 600,000-probe SNP Diversity Panel that replaced legacy genotyping arrays entirely, a project Overton said no competing vendor was willing to take on at the time.
Nucleic acid therapeutics positioned as the next growth leg
Twist is explicitly targeting nucleic acid therapeutics — personalized cancer vaccines, antisense oligonucleotides, and siRNA — as a new serviceable addressable market layered onto its 2030 forecast. Chief Commercial leader Patrick Finn noted that standing up CDMO-scale nucleic acid manufacturing infrastructure typically costs on the order of $1 billion, positioning Twist's existing distribution and quality infrastructure as a structural advantage as the field shifts from bulk manufacturing toward smaller-batch personalized doses. Customer validation came from OncoRNA co-founder Christophe Huffel, whose personalized neoantigen vaccine platform uses Twist's oligo pool technology to encode over 100 patient-specific neoantigens per vaccine — more than triple the roughly 30-34 neoantigens typical of earlier platforms from Moderna and BioNTech. Huffel said the approach triggers a CD8 cytotoxic immune response in 90% of individually tested neoantigens and has secured a EUR 31 million non-dilutive grant from the European Union to advance into Phase II trials, with a stated goal of compressing vaccine turnaround to three weeks from sample to injection. "No other competitor is able to have both that speed, the right format and the price point that matches the needs for this therapy," Huffel said.
Capacity headroom and capital discipline
Laponis said Twist is running 8 synthesis riders today at roughly 50% utilization across its Wilsonville and South San Francisco facilities, with 4 additional riders on order and physical room for up to 22 riders without new construction — implying roughly 3x current rider capacity is available within existing walls. Combined with an already-signed option to expand into a fourth floor in South San Francisco, management reiterated a target of supporting over $1 billion in annual revenue capacity by 2030 without major new capital outlays, with sustaining capital expenditure expected to run in line with depreciation. On M&A, Laponis described a narrow mandate: acquisitions must bring incremental volume onto the silicon chip platform and improve unit economics, citing the recent Invenra bispecific antibody licensing deal (the B-Body platform) as the template rather than a shift toward larger, dilutive transactions.
Scientific advisory perspective on AI's limits
Nobel laureate Frances Arnold, a long-tenured Twist scientific advisor, offered a tempering counterpoint to the AI enthusiasm dominating the day. Asked whether AI-driven drug discovery is durable, Arnold said flatly it is not a "flash in the pan" but cautioned that structure prediction tools such as AlphaFold "are not good enough unless you combine it with these optimization methods" — the directed evolution techniques for which she won her Nobel Prize — and that data to train functional, as opposed to structural, models "we don't have" yet. She also pushed back on the idea that AI will reduce demand for synthetic DNA, arguing instead that imperfect models will require more experimental iteration, not less: "DNA is the point at which you translate your computation into the real world... I think it always will be."