At a glance
A large, well-paid specialty with a starved software layer
Eye care has an unusual shape: enormous, well-reimbursed clinical demand sitting on top of almost no venture-scale technology spend. Capital pours into molecules; the delivery layer is nearly unfunded.
People with sight loss
Living with the consequences of vision loss worldwide — mostly for want of access, not want of a treatment.
WHO fact sheetAnnual productivity loss
Against an estimated $25bn cost to close the unmet-need gap. A 16× spread.
WHO fact sheetOphthalmology VC · 2025
▲ 112% vs 2024 — $934m in 2024. Concentrated in retinal therapeutics.
Market Scope funding trackerDigital eye care VC · 2025
Across 2 rounds through October. Roughly 0.3% of the sector's capital reached software.
Tracxn — digital eye careDiabetics imaged for retinopathy
Received ophthalmic imaging over a five-year window in a US cohort study. Of those, just 2.2% were screened by an AI system.
PMC — US DR imaging cohortAutonomous AI systems cleared
LumineticsCore (2018), EyeArt (2020), AEYE-DS (2022) — plus a further clearance in H1 2026. A billing code has existed since 2021.
Optometry Times — H1 2026 FDA recapAnnual exam adherence
Share of diabetic patients who attend their recommended annual eye exam. The rest are the market.
PMC — DR screening adherenceThe two numbers to hold together: a screening technology cleared eight years ago, with its own reimbursement code, still reaches roughly one diabetic patient in a thousand. Regulatory clearance and a CPT code are necessary here, and demonstrably not sufficient. Retina Specialist — AI DR screening 2025
Capital
Where the money actually went
2025 was a genuine boom — venture funding more than doubled. But it was a therapeutics boom: four rounds account for most of it, and every one of them is a drug or an implant.
Ophthalmology venture funding by year
US$ · equity roundsMarket Scope counts venture capital narrowly; Tracxn counts all equity funding. They differ ~2× on 2024 and are shown here on Market Scope's basis. 2026 is a partial year. Market Scope
Table view
The rounds that made 2025–26
| Company | Amount | What it is | Source |
|---|---|---|---|
| Belite Bio | $350m | Oral tinlarebant for geographic atrophy & Stargardt | Market Scope |
| Ollin Biosciences | $330m | VEGF/Ang2 bispecific (OLN324) for retinal disease | Market Scope |
| Science Corporation | $230m | Prima subretinal implant for dry AMD | Company |
| Orasis Pharma | $126m | Presbyopia-correcting eye drop | Market Scope |
| ForSight Robotics | $125m | ORYOM robotic cataract surgery platform | Fierce Biotech |
Where exits happen
| Deal | Value | Buyer type | Source |
|---|---|---|---|
| Retina Consultants of America → Cencora | $4.6bn | Distributor | OBN 2025 recap |
| EyeBio → Merck | up to $3bn | Pharma | Merck |
| PRISM Vision → McKesson | undisclosed | Distributor | OBN 2025 recap |
| Cimerli franchise → Sandoz | $170m | Pharma | OBN 2025 recap |
| Belkin Vision → Alcon | $81m upfront | Device strategic | Alcon |
Note the buyer list. Two of the five largest are drug distributors buying practices, not technology. The exit path for eye-care software runs through pharma, device strategics (Alcon, Bausch + Lomb, J&J, Zeiss, Topcon), or the practice roll-ups — not through IPO.
Segments
Eight places to build, and what each one is really like
Maturity is judged by whether a new entrant is joining a fight, a build-out, or an empty room.
Surgical robotics · deep dive
Robots are finally operating on the eye
The eye is one of the hardest organs to operate on — micrometer tolerances, a surgeon's tremor, and a fraction of a millimetre of working space. That is exactly what robots are for, and 2024–26 is the moment it moved from papers to patients: first-in-human robotic cataract surgery, robot-assisted retinal surgery on real eyes, and the first approved endoscope-holding robot. The limiting factor is no longer the engineering; it is reimbursement and workflow.
ForSight Robotics — JASPER / ORYOM First-in-human done
What works: world's first fully robot-assisted in-human cataract procedure, filmed end-to-end. The JASPER platform pairs robotic arms (14 degrees of freedom, micron-level precision, tremor-free) with 3D image guidance, computer vision and software "no-fly zones". Proven by: 300+ fully robot-assisted porcine-eye procedures, ISO 13485 certification, and now human cases + early FDA-track data collection. Next: US regulatory pathway, then India/China/Middle East, then glaucoma & retina. $195m total; ~$500m valuation.
Preceyes (Carl Zeiss Meditec)
The system that did it first: robot-assisted surgery inside a human eye at Oxford in 2018, removing a membrane a hundredth of a millimetre thick. Joystick telemanipulation, <20 µm precision, tremor-free — now CE-marked and heading toward FDA via New York Eye and Ear.
Horizon Surgical Systems — Polaris
An AI-enabled robotic microsurgery platform for cataract surgery, combining robotic precision with intraoperative imaging. Founded by UCLA retina surgeon Jean-Pierre Hubschman; preparing for first-in-human studies. $30m Series A (2024).
Riverfield — OQrimo
The world's first approved ophthalmic endoscope-holding robot (Japan, 2023) — a robotic "third hand" that stabilises the intraocular endoscope so the surgeon keeps both hands on instruments. Built on the Eye Explorer research robot from Kyushu University.
Acusurgical — LUCA
French robotics (LIRMM) spin-out with a dual-arm telemanipulated robot for full vitreoretinal surgery at up to 10 µm precision. First-in-human results — six vitrectomies for macular pathology — presented at the 2025 Retina World Congress with no device-related adverse events. European trial recruiting.
The research frontier
IRISS (UCLA) — first robot to complete an entire cataract surgery mechanically. Ophthorobotics — head-mounted robot for automated intravitreal injections (the highest-volume procedure in retina). RAM!S (TU Munich) — a 306 g palm-sized robot with 5 µm accuracy for subretinal injection. OctoMag (ETH) — magnetic-field steering of a wireless microrobot inside the eye, achieving retinal vein cannulation in animal models.
Why this matters for a new entrant. Every serious platform here is a hardware company with a regulatory path measured in years. But the software surface is wide open and immediately addressable: surgical planning, intraoperative AI (depth sensing, tremor-free target tracking), simulation/training for a field where surgeons train on eyes, and the reimbursement + workflow layer that every one of these robots needs. Note also the buyer dynamic — Zeiss already folded Preceyes in; the rest are acquisition candidates for Alcon, Bausch + Lomb, Zeiss, Topcon and J&J.
Who was first? Two companies claim first-in-human robotic cataract surgery and both are technically right — the claims differ in degree. Horizon Surgical announced (Oct 2025) the first in-human cataract procedures using a robotic and AI-enhanced platform (Polaris), where the robot assists standard steps. ForSight Robotics published (2026) the world's first fully robot-assisted in-human cataract procedure, where the robot performs the surgery with a surgeon in the loop. Watch the distinction: assistive robotics (Horizon) and surgical robotics (ForSight) are different products with different reimbursement and training dynamics. Devgan on Horizon's Polaris · ForSight first-in-human
Cell & gene therapy
The biotech layer — where the science is genuinely moving
Ophthalmology is one of the most active categories in cell and gene therapy: the eye is compartmentalised, easy to dose and image, and has measurable endpoints. This is where the majority of sector capital actually lands — the counterweight to the starved software layer. Companies link to their profiles; each carries its own sources.
IRDs · gene therapy. OPGx-LCA5 AAV therapy for LCA5-associated Leber Congenital Amaurosis in a registrational Phase 3; enrollment completed Aug 2026. Formed via Ocuphire's acquisition of Opus.
LCA1 · XLRS. ATSN-101 for GUCY2D-associated LCA1 published 12-month Phase 1/2 data in The Lancet; pivotal Phase 3 expected 2H 2026. ATSN-201 for X-linked retinoschisis.
XLRP · gene therapy. laru-zova (AGTC-501) for RPGR-linked X-linked retinitis pigmentosa in a registrational Phase 2/3; VISTA enrollment completed Jul 2025, topline expected 2H 2026.
Stargardt · protein splicing. SB-007 — the first dual-AAV gene therapy FDA-cleared for Stargardt disease (ABCA4) — in Phase 1/2.
RP · optogenetics. RTx-015 in Phase 1 (ENVISION); FDA RMAT designation (Apr 2026). Series B funds the pivotal trial.
RP · gene-agnostic optogenetics. MCO-010 works regardless of the underlying RP mutation; rolling BLA submission initiated Jul 2025.
GA · platform gene therapy. Broad AAV platform across ophthalmology, neurology and metabolic disease; KRIYA-825 (complement inhibitor, suprachoroidal) in Phase 1/2 for geographic atrophy.
Cornea · cell therapy. Allogeneic corneal endothelial cell therapy plus ROCK inhibitor; launched in Japan (2024), Phase 1/2 in the US.
AI/data platform. FDA-cleared, CE-marked imaging AI for AMD, DR and glaucoma. The segment's biggest 2025 exit: acquired by EssilorLuxottica.
Context: ~75% of ophthalmology investor capital consistently goes to retinal-disorder companies, and the pipeline skews heavily to later-stage assets — investors are paying down clinical risk, not betting on moonshots. Gene therapy faces persistent market-access and manufacturing-cost questions (retinitis pigmentosa funding actually declined in 1H 2025). For a software entrant this layer is competitor, customer and benchmark all at once. Outcome Capital
Who's building
The incumbent set
Who you would be competing with, partnering with, or selling to. Funding figures are latest disclosed.
Scorecard
Segment conditions, scored
Each segment rated 1–5 on how favourable it is to a small, software-first new entrant. Higher is better in every column. These are judgements, not measurements — argue with them.
Entrant conditions by segment
What the grid says. Retinal therapeutics is the only column-topping segment on payment clarity and the worst on everything a small team controls. Clinic operations is the inverse: trivially reachable, immediately payable, and unglamorous. The interesting rows are the ones that are mixed — oculomics and home monitoring both have real whitespace paired with a payment problem, which is a solvable kind of hard.
Barriers
Why good eye-care technology stalls
The graveyard in this space is not full of things that didn't work. It is full of things that worked, got cleared, got a billing code, and still didn't get used.
Idea log
Ideas under evaluation
This section grows. Each idea analysed gets an entry with its thesis, the wedge, what would have to be true, and the honest objection.
OPENING Three gaps the data points at
Not proposals — just the three places where the landscape data contradicts itself, which is usually where a business is. Bring me an idea and it gets logged below this.
1. The adoption layer, not the algorithm. Autonomous retinal AI is solved, cleared and billable, and reaches almost nobody. The unbuilt product is whatever converts a cleared algorithm into a screening event inside a primary-care, pharmacy or endocrinology workflow — capture quality, staff time, referral routing, and the billing itself.
2. Oculomics without the payer problem. Retina-as-systemic-biomarker has strong science and no reimbursement. That combination usually means the first buyer is not a payer at all — life insurance, employers, pharma trial enrichment, or health systems with capitated risk.
3. Capacity, not diagnosis. The leading causes of global blindness are capacity-limited, not knowledge-limited. Anything that multiplies scarce ophthalmologist minutes — triage, pre-op, post-op follow-up, surgical throughput — sells into a demonstrable shortage.
Method
How to read this
Funding totals come from Market Scope and Tracxn, which disagree because they count differently — Tracxn tracks equity rounds only, Market Scope includes public offerings and debt. Where they conflict, both are shown. 2026 is a partial year and is marked as such; do not read the drop as a trend yet. Every figure on this page carries an inline source link; hover-tap the small superscript-style links to go straight to the primary source.
Scorecard values are editorial judgements about entrant conditions, not measured quantities. They exist to be argued with, and they will change as ideas get tested against them.
Company profiles live at companies/index.html — one page per company with a description, funding where disclosed, and links to the company website and LinkedIn. Sources are listed per claim.