Jonathan Rothberg's whole career basically runs on one question: if a life-science tool is expensive, bulky, and centralized, why hasn't this been moved onto silicon yet?
He's already answered it twice.
With 454 Life Sciences and later Ion Torrent, he pulled DNA sequencing onto a semiconductor platform. Gordon Moore ended up being the first individual sequenced on a semiconductor chip, which is one of the moments that opened the NGS era and a big part of why Rothberg got the National Medal of Technology under Obama. Then at Butterfly Network, he did it again with ultrasound — replacing the piezoelectric crystal architecture with a silicon chip, getting FDA clearance across 13 applications, and bringing the iQ down to roughly $2,000. [Yale Medicine](https://medicine.yale.edu/profile/jonathan-rothberg/)[Wikipedia](https://en.wikipedia.org/wiki/Jonathan_Rothberg)
What's interesting is how identical the playbook is. A legacy medical device is expensive, specialized, and hand-assembled. Move the core function to CMOS and the cost collapses, the form factor shrinks, the tool moves from centralized institutions out to broader access. And then the second-order effect kicks in: the device keeps generating data, and the data eventually becomes more valuable than the hardware itself. Razor, blade, data.
The third target is protein.
DNA's been decoded, but the actual machinery of biology runs at the protein level. \~20k genes, but the protein universe is orders of magnitude bigger once you account for isoforms, modifications, PTMs. And the dominant tool for analyzing all of it is still mass spectrometry. Powerful, sure, but the high-end systems run hundreds of thousands to millions of dollars, require PhD-level operators, and are mostly built around bulk averaging rather than true single-molecule readouts. Structurally this looks a lot like where DNA sequencing was sitting before the NGS shift, and Rothberg has already lived through that exact transition.
That's why Quantum-Si is on my radar.
The core asset is Proteus. Investor Day last November laid out the roadmap, but what changed this year is that the thesis stopped being theoretical and started being something you can actually track through data releases.
On April 8, Quantum-Si ran its first customer samples on a Proteus prototype. Versus Platinum Pro, Proteus showed significant improvements across amino acid coverage, read length, peptides identified, and sequencing output. Three weeks later, on April 28, they reported automated end-to-end sequencing on integrated Proteus instruments, detecting 17 amino acids — up from 15 just four months earlier. [Quantum-Si](https://www.quantum-si.com/press-releases/quantum-si-announces-first-customer-samples-tested-on-the-proteus-prototype-system/)[Stock Titan](https://www.stocktitan.net/news/QSI/quantum-si-announces-successful-sequencing-on-integrated-proteus-tm-ugkn9wy17z10.html)
The pace is the thing I keep coming back to. The company currently uses six recognizers to detect 14 amino acids and believes an eight-recognizer set can cover all 20. The Proteus optical system has already demonstrated it can distinguish eight dyes. The hardware is basically waiting on the chemistry. That's a meaningful shift in risk profile — from "can this work at all" to "how fast do they close out the recognizer set." Those aren't the same risk and they don't deserve the same valuation. [GenomeWeb](https://www.genomeweb.com/proteomics-protein-research/quantum-si-targeting-late-2026-launch-proteus-protein-sequencing)
Guided timeline from management: early access customers this summer, pricing announcement in Q2 (range previously guided at $300k–$500k), launch capabilities and Platinum Pro upgrade program in Q3, commercial launch end of 2026 with 18 AA coverage, and full 20 AA coverage to market in 2027. Q2 earnings likely land in early August based on reporting cadence.
I want to be careful with the BFLY parallel because the easy version of it is wrong. Chip-based disruption takes time, and Butterfly's stock has been painful even when the underlying technology was directionally right. So that's not the lesson. The lesson is that once the chip version starts working, the category eventually has to deal with it. When Butterfly raised $250M in 2018 from Fidelity, the Gates Foundation, and Fosun Pharma at a $1.25B valuation, the piezo vs CMUT debate had already shifted. The market timing was a different question, but the direction of the technology wasn't really in dispute anymore. [Wikipedia](https://en.wikipedia.org/wiki/Jonathan_Rothberg)
Proteus may be approaching a similar moment. If single-molecule protein sequencing on a semiconductor actually works at production scale, I don't see how legacy mass spec wins long-term on cost, form factor, or decentralization. Orbitrap will probably remain more accurate for a while — but accuracy gaps narrow with better reagents, chemistry, and algorithms. Unit economics and distributed access don't narrow. Those are structural advantages.
There's also a layer most people aren't pricing in. Just like Butterfly was never only about hardware, QSI isn't only about instruments and consumables. Platinum and Proteus generate sequence data that flows into QSI Cloud, and over time that can become a proprietary peptide identification database. The 23andMe comparison gets thrown around here but I think it cuts the other way. 23andMe didn't fail because biological data has no value. It failed because the data wasn't embedded tightly enough into clinical or drug-development workflows. If Proteus gets adopted inside pharma proteomics pipelines, that trap is avoidable.
Balance sheet is fine too. As of March 31, QSI had $190.4M in cash, with operations funded into Q2 2028. 2026 guidance is roughly $1M in revenue, adjusted opex below $98M, total cash usage below $93M. Most pre-revenue medtech dies or dilutes badly right before launch, and QSI looks like it can reach the Proteus launch window without slamming into that wall. [BriefGlance](https://briefglance.com/companies/quantum-si-incorporated/pulses/11907)
The thesis can break, obviously. The things I'm watching: amino acid progress stalling over the next six months, especially around the 17 → 19 transition (suggests recognizer engineering is plateauing). Early access customers going silent or peer feedback turning negative (workflow fit problems). Nasdaq minimum bid moving into a real deficiency stage without a reset (separate but real risk). Until one of those triggers, I think the bigger picture holds.
Rothberg has run this pattern twice. Take a centralized, expensive life-science tool and move the core function onto silicon.
BFLY was the first act. QSI looks like the next one.
*Position: long QSI.*
