Ross Doerr needed an MRI. He had a brain tumour, and his doctors wanted a clear picture before deciding what to do about it. He also had a metal and electronics package surgically attached to the back of his eye, and nobody could tell the radiologist what was inside it, what it was made of, or whether a magnetic field would heat it, move it or tear it.
The manufacturer had stopped answering. So he got a CT scan instead, with less detail, and no answer to the question that mattered.
That is what a dead device company looks like from the inside of a patient's body.
What Happened to Argus
Second Sight Medical Products built the Argus II, a retinal prosthesis for people blinded by retinitis pigmentosa. A camera on glasses sent signals to an electrode array on the retina, and users learned to read rough patterns of light as shapes. It received European approval in 2011 and US approval in 2013. More than 350 people were implanted worldwide, at a total cost of surgery and upkeep reported at around US$500,000 (≈RMB 3.6 million / HK$3.88 million) per patient.
In February 2022, the company announced it would merge with a drug-delivery firm and pivot away from the retinal business. Support for the Argus hardware ended. Patients found out through news reports, not letters: no email, no call, no service line.
The consequences were mundane and brutal. One user's video processor cracked and he crowdfunded for spare parts. Another had a device fail after four years of use and never got it repaired. Others had to decide whether to have the implant removed — a second surgery, back into darkness, and a bill — or keep a working device whose failure they could no longer predict.
The Part of the Industry That Rarely Gets Modelled
Every brain-computer interface (脑机接口) pitch deck contains a technology roadmap. Almost none contains a support roadmap.
The two are not the same document. An implant is not a phone. It cannot be replaced by next year's model, because replacing it means opening the skull or the eye again. It has a battery with a finite number of cycles, firmware that needs updating, electrodes that scar over, and a patient who will outlive the company's enthusiasm for the product.
Medical device regulation in most countries assumes someone will hold the registration for as long as the device is inside a person. In China, the Medical Device Supervision and Administration Regulation (医疗器械监督管理条例), in force since June 2021, puts post-market obligations — quality management, adverse-event monitoring, recalls, technical support — on the registration holder (注册人). The logic is sound: the entity that sold the device stays responsible for it.
The logic breaks when the entity stops existing. A startup that runs out of money does not file a plan for the patients still carrying its electrodes. The registration lapses, the engineers disperse, and the firmware source code sits on a server that gets switched off.
Why China's Brain-Interface Sector Is Exposed
China now runs one of the world's busiest pipelines of implanted brain-interface companies — many private, many venture-funded, and many younger than the clinical follow-up periods their own regulators demand.
The pattern of the sector makes the orphan risk structural rather than hypothetical:
- Long timelines, short runways. A registration trial in China for an implanted neural device typically runs several years. A Series B round typically buys two to three years of runway. Those two clocks do not line up.
- Single-product companies. Many firms have one device and one indication. If the indication disappoints clinically or commercially, there is no second product to carry the company.
- Hospital-anchored deployment. Devices are often placed through research agreements at large tertiary hospitals. When the study ends, who maintains the implant is frequently undefined in the paperwork.
- No maintenance revenue yet. Because brain interfaces are not routinely reimbursed, companies cannot bill for the programming sessions and repairs that would fund a service organisation.
The United States and Europe have the same problem, which is why the Argus case keeps resurfacing in policy discussions. China's version is likely to be larger if, as expected, the number of implanted patients grows faster.
The Questions That Survive a Bankruptcy
Patients and clinicians can price this risk before surgery, even if they cannot eliminate it. The useful questions are unglamorous:
- Who holds the registration, and who is contractually on the hook if they dissolve? Ask for the entity's legal name, not the brand.
- What is the battery's expected service life, and what happens at year eight? Implants with non-rechargeable batteries have a hard expiry date written into them.
- Is there a documented service commitment with a duration? "Lifetime support" in a brochure is marketing. A warranty with a term and a named third party is not.
- Can an independent hospital program the device if the maker disappears? If the firmware and clinical programmer are locked, the answer is no.
- What is the exit plan for the implant itself? Removal is a second surgery with its own risks and costs, and it should be priced and discussed before the first one.
Some of these have begun to appear in Western procurement contracts, particularly escrow arrangements for device software. In China they are still rare. A regulator, a hospital group or a patient advocate could close that gap without waiting for a new technology.
The Argus II still works in some heads. That is the uncomfortable part. The hardware outlived the company's willingness to answer the phone.
Honest limitations
The account of the Argus II shutdown rests on reporting by IEEE Spectrum and follow-up coverage, not on company filings that I examined directly. The patient details — the tumour, the crowdfunded processor, the four-year failure — come from that reporting and I have not independently verified each case.
I have not seen a Chinese court case or regulatory action involving an abandoned neural implant, and I am not claiming one exists. The exposure described here is inferred from the structure of the sector — early-stage companies, long trials, no maintenance revenue — not from a documented Chinese failure. The description of registration-holder obligations paraphrases the general duties in the 2021 regulation; specific obligations vary by device class and should be checked against the current text.
Finally, this article treats "the company dies" as the main risk. It is not the only one. A company can survive, abandon a product line, and leave patients in the same position.
What readers can do now
- Before any implant, ask what happens at year eight. Not year one. The support question is a longevity question, and it belongs in the consent conversation, not the sales brochure.
- If you are building or funding a neurotech company, budget for a service organisation from the first commercial implant. The cost of maintaining 300 devices is small compared with the cost of being the company that left them.
- If you regulate or procure, treat software escrow and transferable programming rights as standard terms for any device that cannot be removed without surgery.
