Neuralink designs its implant in-house and manufactures it at its own California and Texas sites — but the silicon comes from outside, and in 2026 that supplier changed. Samsung Foundry took the fourth-generation chip from TSMC, which built generations one through three. Here is the full component map, the manufacturing footprint, and the reason components are not actually what limits Neuralink’s scale.
👉 The framing that separates this from every other Musk company: Neuralink’s bottleneck is not a factory. It is operating-room hours and the biology of scar tissue. A supply chain analysis that only counts vendors will miss the entire story.
- The chip: Samsung Foundry is developing the fourth-generation implant chip on a 4nm process, internal codename O1. TSMC built the first three generations.
- The implant: a hermetically sealed unit roughly 23×8 mm with 1,024 electrodes across flexible polyimide threads 4–6 µm wide, plus a custom low-power ASIC and an inductively charged battery.
- The factories: Neuralink’s own cleanroom device manufacturing in California and an expanding Austin footprint, including a $8.2M fit-out at its Del Valle site.
- The real constraint: not parts. Surgical throughput, thread reliability and regulatory stage gate the ramp far more than any supplier does.
- The scale gap: roughly 20 trial participants as of January 2026 against a stated goal of high-volume production in the same year.
Who Makes the Neuralink Chip?
In June 2026 it emerged that Samsung Electronics’ foundry division had secured its first order from Neuralink, developing the company’s fourth-generation brain implant chip on a 4nm process under the internal codename O1 (SamMobile on the Samsung order).
The reported timeline is specific: development began in late 2025, pilot wafer fabrication started in May 2026, initial test samples are due in the first half of 2027, and mass production could follow in the second half of 2027 if biocompatibility and laboratory validation pass (detail on the Samsung–Neuralink roadmap).
Why Neuralink moved off TSMC
Two reasons, both structural rather than technical. TSMC’s advanced nodes and packaging capacity are heavily committed to AI and mobile customers, which makes securing predictable production windows hard for a low-volume medical device. And single-sourcing the most critical component in an implantable product is a risk profile no medical manufacturer wants permanently.
Samsung also offers turnkey capability — foundry, memory and advanced packaging from one supplier — which matters more for a sealed implant than raw process leadership does (analysis of the foundry switch).
What the fourth generation actually changes
Generations one through three read neural signals and converted them into device commands. The fourth-generation chip is described as bidirectional — able to write signals back into the nervous system as well as read from it. That is the technical prerequisite for vision restoration and sensory feedback, not an incremental performance bump.
💡 Overlooked engineering point: for an implant, the design target is not speed. It is heat dissipation and zero-defect reliability inside a sealed device sitting against brain tissue. A 4nm node is being used to reduce power draw, not to win benchmarks — which is why "advanced node" headlines misread this deal.
Inside the N1: A Component-by-Component Supply Map
| Component | Specification | Sourcing status |
|---|---|---|
| Implant ASIC | Custom low-power chip, signal processing to ~20 kHz per channel | External foundry — TSMC (gens 1–3), Samsung 4nm (gen 4) |
| Electrode threads | 64 flexible threads, 4–6 µm wide, 1,024 electrodes total | Neuralink in-house thin-film fabrication |
| Thread materials | Polyimide substrate with thin gold or platinum conductors | Specialty materials — externally sourced |
| Hermetic enclosure | Coin-sized sealed casing, roughly 23×8 mm, biocompatible | Medical-grade packaging, tightly controlled |
| Battery | Small rechargeable cell, inductively charged | External medical battery supply |
| Wireless link | Bluetooth Low Energy to phone app | Standard RF components |
| R1 surgical robot | Automated thread insertion avoiding vasculature | Neuralink in-house |
Neuralink describes the power system plainly: the implant is "charged wirelessly from the outside via a compact, inductive charger" that allows use from anywhere (NPR on the N1 hardware). Thread composition — polyimide with gold or platinum conductors — is documented in independent technical references (N1 implant technical profile).
✔ Notice the pattern in that table: everything biological-facing is in-house, everything electronic is bought. That is the inverse of Tesla, where Tesla builds the electronics and buys the materials. It reflects where the defensible IP actually sits in neurotech — in the threads and the robot, not the silicon.
The Manufacturing Footprint
Neuralink has been steadily building physical capacity. Filings show an $8.2 million tenant improvement at its Del Valle site in Austin, converting 37,607 square feet into office and manufacturing space, following earlier filings for a 112,000-square-foot facility with offices, a machine shop and cleanroom device manufacturing (reporting on Neuralink’s manufacturing build-out).
Cleanroom capacity is the tell. Thin-film thread fabrication and hermetic sealing are cleanroom processes, and they are the steps Neuralink cannot outsource without giving away its core advantage. The Austin expansion is a bet on doing more of the implant internally, not less.
Four Constraints — Only One Is a Supply Chain Problem
This is where most coverage of Neuralink goes wrong. Analysts count vendors and factories, then conclude the ramp is a manufacturing question. It is not. Rank the constraints honestly:
- Component supply — the easiest. At current volumes, chips, batteries and materials are not the limiting factor. Neuralink builds in the low tens of devices, not millions.
- Biocompatibility qualification — hard. Every material change in an implant triggers revalidation. This is why the Samsung chip must pass biocompatibility testing before mass production, not just electrical testing.
- Surgical throughput — hardest. Each implant requires a craniectomy, a trained neurosurgical team and an operating room. Operating-room hours, not factory hours, set the ceiling.
- Regulatory stage — binding. Neuralink has FDA Breakthrough Device designations, but reporting indicates it has not registered the pivotal trial that commercial approval requires. No pivotal trial, no commercial volume, regardless of factory capacity.
👉 This is why the automated-surgery announcement matters more than the manufacturing announcement. Musk paired "high-volume production" with an almost entirely automated procedure inserting threads through the dura without removing it. The surgical change is the actual scaling lever; the factory is the easy part.
The Thread Retraction Problem
In the first patient, Noland Arbaugh, electrode threads retracted from brain tissue in the months following implantation, degrading signal quality. Neuralink compensated through software and the device kept working, but the episode exposed a genuine biocompatibility limit: brain tissue reacts to foreign objects, and holding signal quality across years remains unsolved (investor analysis of Neuralink’s technical risks).
For a supply chain, this has a specific consequence most write-ups miss. If thread geometry, coating or insertion depth changes to address retraction, the material specification changes — and every material change restarts biocompatibility validation. Design iteration and supply stability are directly in tension here in a way they never are in automotive or aerospace.
Scale Math: What "High-Volume" Actually Means
On December 31, 2025, Musk said Neuralink would begin "high-volume production" of its devices and move to an entirely automated surgical procedure during 2026 (Reuters on the 2026 plan). Set that against the record.
| Metric | Stated ambition | Documented reality |
|---|---|---|
| Patients implanted | Over 1,000 achievable in 2026 (2024 statement) | 12 by September 2025; roughly 20 trial participants by January 2026 |
| Production mode | "High-volume production" during 2026 | Clinical-scale build; Samsung mass production targeted for late 2027 |
| Surgery | "Almost entirely automated" | Robot-assisted, clinician-supervised; dura-sparing insertion in development |
| Regulatory | Commercialization | Breakthrough designations held; pivotal trial not registered |
| Valuation | — | $9B at the Series E; secondary-market implied figures reported near $40B in early 2026, unconfirmed |
Note the internal inconsistency: mass production of the fourth-generation chip is targeted for late 2027, yet "high-volume production" was promised for 2026. Both statements can be true only if 2026 volume runs on third-generation silicon — which means the 2026 ramp, whatever its size, is not built on the new chip.
⚠ For anyone modelling this: treat 2026 as a clinical-scale year with an automation pilot, and 2027–2028 as the first plausible window for anything resembling volume. Company projections circulating in investor material point to roughly 2,000 procedures annually by 2029 — a useful reality anchor against the four-figure 2026 talk.
Blindsight: A Second, Different Supply Chain
Blindsight is not a variant of the N1 — it is an inverted product. Where Telepathy reads from motor and speech cortex, Blindsight writes into the visual cortex, stimulating neurons to produce the sensation of sight. It received FDA Breakthrough Device Designation on September 18, 2024, with first human trials expected in 2026 (Blindsight trial overview).
The hardware is a step up: the S2 implant is reported at 3,072 electrodes, paired with camera-equipped glasses and a phone doing image processing before wireless transmission to the implant (Blindsight hardware breakdown).
Supply-chain consequence: tripling electrode count roughly triples thin-film fabrication load per device, and stimulation hardware carries different safety qualification than recording hardware. Blindsight does not reuse the Telepathy supply chain so much as fork it.
How Neuralink’s Supply Chain Compares
| Neuralink | Synchron | Precision Neuroscience | |
|---|---|---|---|
| Approach | Penetrating threads in cortex | Stent-based, via blood vessels | Thin-film array on brain surface |
| Electrode count | 1,024 (N1); 3,072 (S2) | Around 16 | Thin-film surface array |
| Surgery required | Craniectomy, robot-assisted | Endovascular, no open surgery | Less invasive than penetrating |
| Supply chain implication | Cleanroom thin-film + custom ASIC + robot fleet | Leverages existing stent manufacturing | Thin-film focused |
| Scaling bottleneck | Operating-room hours | Lower surgical barrier | Lower surgical barrier |
💡 The strategic trade in one line: Neuralink chose the highest-bandwidth approach and therefore inherited the hardest scaling problem. Competitors with fewer electrodes need far less surgical infrastructure. Bandwidth per device and devices per year pull in opposite directions in this industry.
Checklist: How to Evaluate Any BCI Supply-Chain Claim
- Separate device generation from production claim. A 2026 volume statement built on 2027 silicon is two different things.
- Ask whether the constraint named is manufacturing or surgical. Operating-room capacity is the real ceiling for penetrating implants.
- Check regulatory stage, not designation. Breakthrough Device status accelerates review; it is not approval, and it is not a pivotal trial.
- Watch for material changes. Any change to thread, coating or enclosure restarts biocompatibility validation and resets timelines.
- Treat secondary-market valuations as indicative only. Reported figures near $40 billion in early 2026 are unconfirmed and sit far above the last priced round.
FAQ
Who makes the Neuralink chip?
TSMC manufactured the first three generations. Samsung Foundry is developing the fourth-generation chip on a 4nm process, with pilot wafers started in May 2026 and mass production targeted for the second half of 2027.
Does Neuralink manufacture its own implants?
Partly. Neuralink fabricates the electrode threads and builds the R1 surgical robot in-house, with cleanroom device manufacturing in California and an expanding Austin footprint. The chip, battery and specialty materials come from outside suppliers.
What is the N1 implant made of?
A hermetically sealed biocompatible casing roughly 23×8 mm containing a custom low-power ASIC, an inductively charged battery and 64 flexible polyimide threads with gold or platinum conductors carrying 1,024 electrodes.
How many Neuralink implants have been done?
Twelve patients had received implants by September 2025, rising to roughly 20 trial participants by January 2026. That is clinical scale, not commercial volume.
What limits Neuralink from scaling?
Surgical throughput and regulatory stage, not component supply. Each implant needs a craniectomy and a trained neurosurgical team, and no pivotal trial has been registered.
Is Blindsight using the same supply chain?
No. Blindsight’s S2 implant is reported at 3,072 electrodes and stimulates rather than records, which means different fabrication load and different safety qualification.
The Bottom Line
Neuralink’s 2026 supply chain story has one clean headline — Samsung taking the fourth-generation chip from TSMC — and one uncomfortable subtext: the chip was never the thing holding this company back.
The implant is a small number of well-understood parts wrapped around two genuinely hard problems: threads that stay put in living tissue, and a surgical procedure that currently requires an operating room per patient. Neither is solved by a better foundry partner.
Judge progress in 2026 on three things only — whether dura-sparing automated insertion works in humans, whether a pivotal trial gets registered, and whether thread signal quality holds past the two-year mark. Everything else is packaging.
💡 We track the Musk hardware ecosystem end to end — Neuralink, Tesla, SpaceX — with confirmed and reported figures kept in separate columns. Join the community to get the breakdown as each disclosure lands.
Sources: Reuters, SamMobile, Korea Economic Daily reporting, NPR, bionic-vision.org, Neurofounders, Acquinox Capital, Neurapod. Supplier relationships and secondary-market valuations are reported by third parties and not confirmed by Neuralink. Last fact-check: August 5, 2026.
NEVER MISS AN OPTIMUS UPDATE
We track every Tesla Optimus development — specs, deployment milestones, pricing and competitive moves — updated as news breaks.