A sixteen-hour problemwith a six-hour answer.
GenoAccel is an FPGA-accelerated genomics analysis platform from Three.Group. This page covers the market thesis, where defensibility actually comes from, what capital each phase requires, and the one decision that moves every number on it.
01 · The market
Same computation. Radically different buyers.
The problem is not that genomic analysis is impossible. It is that it is slow.
A modern sequencer produces raw reads, and those reads are worthless until secondary analysis turns them into an answer: align to a reference genome, call the variants, assess quality, interpret. On conventional CPU infrastructure a 30× whole human genome takes roughly sixteen hours of compute.
That number is the entire business, because it means different things to different people. A research core facility experiences it as cost: compute is a line item, throughput caps how many samples a grant can process, and the queue is a planning constraint. A neonatal intensive care unit experiences it as a clinical deadline — a critically ill newborn has a window in which a genetic diagnosis can still change treatment, and sixteen hours of compute plus queue time plus interpretation can fall outside it.
One engine serves both markets. The research buyer pays for efficiency; the clinical buyer pays for time. The willingness to pay differs by more than an order of magnitude.
02 · Defensibility
The FPGA is not the moat.
Timing matters too: AWS F1 reached end-of-life on 31 December 2025, F2 offers materially better price-performance, and the commercially maintained bitstream path runs on F2 — Illumina ships DRAGEN there. Every cost model built on F1 is void.
Anyone can rent an F2 instance, and we say so plainly because an investor will work it out anyway. Defensibility compounds in a specific order, and the order is the strategy:
1. Validation evidence. A documented concordance record against Genome in a Bottle truth sets, maintained across every pipeline version, is roughly eighteen months of work a competitor cannot skip or buy.
2. Compliance posture. SOC 2 Type II, then CLIA and CAP. Each is a gate that removes competitors from the buying conversation entirely rather than merely outranking them.
3. Clinical workflow depth. Laboratory information system integration, ACMG variant classification, sign-out workflow. Unglamorous, and extremely hard to displace once a laboratory's procedures are built around it.
The FPGA gets you in the door. The other three keep you there.
03 · Market phases
Each phase funds the next.
Revenue is staged behind regulatory milestones rather than optimism. Each gate is a real artifact — a signed contract, an auditor's report, a certification — not a launch announcement.
| Phase | Market | Compliance gate | Timeline | Target ARR |
|---|---|---|---|---|
| M1 — Research | Academic labs, core facilities | Terms + baseline security | Months 0–12 | $200–500k |
| M2 — Pre-clinical | Biotech, CROs, pharma | HIPAA + SOC 2 Type II | Months 12–24 | $2–5M |
| M3 — Clinical | Hospitals, clinical labs | CLIA / CAP | Months 24–48 | $10–20M |
| M4 — Device | National labs, IVD | FDA 510(k) | Months 36–72 | $20M+ |
The discipline the plan imposes on itself: do not start a phase whose exit conditions cannot be funded from the prior phase's revenue plus a raise already closed. M3 costs $500k–1.5M; M4 costs $300k–1M. Targets are planning models, not forecasts.
04 · Capital required
What it costs to get to research launch, and the one line that dominates.
Approximately 48 weeks of engineering to research launch, assuming focused effort with heavy AI-agent leverage. The compliance work does not compress, because an audit observation window is wall-clock time rather than effort.
| Line item | Estimate | Note |
|---|---|---|
| FPGA bitstream licensing | $0–500k / yr | The swing factor. Three routes differing by ~12 months. |
| Baseline cloud infrastructure | $1,250–2,970 / mo | Pre-revenue. F2 compute is variable and scales with analyses run. |
| HIPAA gap analysis (consultant) | $40–60k | Build Phase 4 |
| SOC 2 Type II audit | $30–60k | Observation window is 3–12 months wall clock |
| Penetration test + remediation | $15–25k | Also a SOC 2 expectation |
| Legal (terms, BAA, DPA, SLA) | $15–30k | Pre-launch |
| Translation, 20 locales | $15–40k | Interface is already internationalised |
| Bioinformatics advisor (fractional) | $30–80k / yr | Scientific credibility |
| Accessibility audit / VPAT | $5–15k | Public-sector procurement requirement |
| Reference data licensing | $0–5k | COSMIC commercial licence |
Read the top line first. Bitstream licensing spans zero to half a million dollars a year depending on which of three routes is chosen, and it also moves the timeline by about a year. Until that decision is made, any single headline number for total capital required would be false precision. We would rather show the range and the reason than a confident figure we cannot defend.
Cost per genome is deliberately not stated. It is a formula with two unverified inputs — verified F2 pricing and measured pipeline hours. Every cost-per-genome figure from the earlier plan was extrapolated from the retired F1 generation and has been voided rather than carried forward.
05 · Build status
What exists today, verified rather than asserted.
The built half is concentrated deliberately in architecture, security posture, tenancy, identity and data management — the layers most expensive to retrofit, and the ones everything else sits on.
| Phase | Scope | Status |
|---|---|---|
| 0 — Decisions | Repo governance, accounts, bitstream decision | Partial — bitstream decision open |
| 1 — Foundation | Domain, persistence, logging, i18n, API hardening | Partial — auth and the web shell are built; no AWS infrastructure yet, so nothing has run against a live identity pool |
| 2 — Core product | Projects, samples, uploads, pipeline, billing | Partial — everything that does not need AWS is built; the pipeline, queue workers and spot handling wait on a live account |
| 3 — FPGA & validation | AFI, FPGA stages, GIAB concordance | Not started |
| 4 — Compliance | HIPAA, SOC 2, pen test, DR, GDPR | Not started |
| 5 — Launch | Marketing, docs, CLI, load test, beta | Not started |
Quality signals as of today: 1,171 automated tests passing, zero failing. Clean release build with compiler warnings treated as errors. Zero known-vulnerable dependencies across all fourteen projects. CodeQL static analysis on every change, plus weekly, with a build-breaking severity gate. Every change verified in continuous integration before it lands.
06 · Risks we would raise ourselves
The four things a diligence process would find.
A pitch that hides these is a pitch that wastes everyone's second meeting.
Bitstream licensing is unresolved. A $0–500k/year, ~12-month swing. Nothing downstream is fully costed until it is decided.
No measured performance data yet. The 3–5× end-to-end target is derived from AWS's general FPGA figures, not from our own runs. Until the benchmark suite executes, the speedup is a design target.
Clinical revenue is gated on certification we do not hold. CLIA is held by the laboratory reporting a result. Serving clinical customers at scale means either supporting their certification or obtaining our own, which is a $500k–1.5M, 12–24-month programme including a board-certified laboratory director as a senior hire.
Timeline assumes focus. 48 weeks presumes near-full-time attention. If GenoAccel competes with other Three.Group platforms for attention, the calendar stretches or the scope shrinks.
07 · Questions
Investor questions
What stage is GenoAccel at?
Pre-revenue and pre-launch. The control plane is built and continuously tested: domain model, persistence, tenant isolation, upload pipeline, project and sample management, job submission and metering, a role-scoped members area, enterprise single sign-on logic, internationalisation and security middleware, with 1,171 automated tests passing in continuous integration. The FPGA execution plane is specified and not yet built, and nothing has yet run against live cloud infrastructure: the AWS account, hosted identity pool and payment processing are ahead.
What is the single largest open decision?
FPGA bitstream licensing. There are three routes — license a commercial bitstream, buy from the AWS Marketplace, or build a custom one — and they differ by roughly half a million dollars a year and about twelve months of calendar time. The engineering plan treats everything downstream of that decision as provisional until it is made. We would rather surface that in the first conversation than have it discovered in diligence.
Why is now the right time?
AWS F1 instances reached end-of-life on 31 December 2025 and customers were migrated to F2. F2 is faster, offers materially better price-performance, and the commercially maintained bitstream path runs on it — Illumina deploys DRAGEN secondary analysis on F2. Every cost model built on the retired F1 generation is void, which resets the competitive landscape for anyone who had one.
What stops a larger competitor simply renting the same hardware?
Nothing, and that is the point — the FPGA is not the moat. Defensibility compounds in a specific order: a documented concordance record against Genome in a Bottle truth sets maintained across every pipeline version, then SOC 2 Type II followed by CLIA and CAP, then clinical workflow depth. The first is roughly eighteen months of work a competitor cannot skip. The second removes competitors from buying conversations entirely. The third is very hard to displace once a laboratory's standard operating procedure is built around it.
Are performance claims independently verifiable?
They will be. Validation runs against public Genome in a Bottle reference materials, and the methodology and raw outputs are published so a third party can reproduce the comparison rather than take a score on trust. No benchmark figure has been measured yet, and none is published until it has been.
How is the business structured?
GenoAccel is a Three.Group platform. The recommended structure is GenoAccel as its own Delaware C corporation with Three.Group holding founder equity — pure-play exposure for investors, liability isolation for a business that will handle protected health information, a cleaner regulatory scope for CLIA and SOC 2, and a cleaner acquisition path. Final structuring is with corporate counsel.
08 · Next step
Start a conversation.
Information, not an offer — see the note below.
GenoAccel is a Three.Group platform. Three.Group is the software development company behind it and is the entity to talk to about the portfolio; GenoAccel is the platform this page describes.
If the thesis is interesting, the most useful next step is a conversation rather than a document. Technical diligence material — architecture, the validation methodology, the full cost model and the bitstream decision analysis — is shared under NDA with qualified parties, because the implementation specifics are the part that is genuinely worth protecting.
Contact GenoAccel Visit Three.Group
Important — this page is information, not an offer. Nothing on this page is an offer to sell or a solicitation of an offer to buy any security, and it is not investment, legal, tax or accounting advice. Figures shown are internal planning estimates drawn from the engineering cost model; they are forward-looking, subject to change, and several inputs are explicitly unverified. Any securities offering would be made only to qualified investors through formal offering documents and in accordance with applicable law. Past or projected performance is not a guarantee of any outcome.