Brain-Computer Interface Sub-Micron Advanced Interconnect

1. Micro-Assembly Target: What Are We Actually Assembling?

In January 2024, Elon Musk posted on X: "The first human patient has received a Neuralink implant and is recovering well." On screen, a completely paralyzed patient was controlling a computer cursor, typing, and even playing video games — using only their thoughts.

To date, 21 people worldwide have participated in Neuralink clinical trials. In early 2026, Musk further announced that Neuralink would begin mass production of BCI devices that year.

But behind the miracle of "thought-controlled typing" lies a critical engineering challenge rarely discussed — how do you precisely connect thousands of flexible electrodes, thinner than a human hair, onto a chip the size of a fingernail?

Below, we explore the emerging field of brain-computer interfaces (BCI) from a micro-assembly engineering perspective[1].

A typical implantable brain-computer interface (iBCI) system integrates a variety of technical components:

Table: Core Components of an Implantable BCI System

Component	Function	Assembly Challenge
CMOS Chip	Signal acquisition, processing, wireless transmission	High-density I/O, sub-micron alignment
MEMS Structures	Micro-sensors, actuators	Co-integration of precision mechanical structures and electrical interconnects
Flexible Electrode Array	Neural tissue interface — recording or stimulating neural signals	Ultra-thin flexible substrate, easily deformed, difficult to fix; requires "soft-to-hard" interconnect with rigid chips
Thin-Film Substrate	Interconnect routing carrier	Heat-sensitive material, requires low-temperature processing
Biocompatible Encapsulation	Long-term implant protection	Hermeticity requirements, thermal management challenges

▲ Neuralink BCI exploded view — multi-level interconnect structure of flexible electrodes and chips (Source: Neuralink)

2. Three Engineering Challenges: Accuracy Is Only the Beginning

The introduction of flexible materials creates new engineering difficulties — how do you reliably connect such thin, easily deformable flexible electrodes to rigid chips? This is not just a material interface problem, but a convergence of three challenges: accuracy, force control, and thermal-power management.

▲ Flexible electrode — microelectrode contacts densely arranged on an ultra-thin polymer substrate

A typical high-throughput BCI chip requires thousands of micron-scale contacts to be precisely aligned: each Neuralink thin-film device features 3,072 electrode contacts, using flip-chip bonding to attach the integrated circuit to the electrode contacts in the thin-film sensor area, packaging 3,072 channels within an area of 23 × 18.5 mm²[3]. Stepper lithography can create metal films at sub-micron resolution, but flip-chip bonding onto these tiny contacts demands sub-micron placement accuracy — otherwise signal channels are lost.

▲ Novel polymer probe flexible electrodes with 50 μm and 75 μm contact spacing

BCIs use flexible electrodes precisely because their mechanical properties more closely match soft brain tissue, significantly reducing immune rejection and inflammatory response. However, the introduction of flexible materials brings new process challenges:

In short: Flexibility solves tissue compatibility but creates chip connection difficulty.

Implants directly contact brain tissue and face extremely strict thermal budgets — even a 1°C temperature rise can cause irreversible neural damage.

Higher power means more heat, forcing chip power consumption to ultra-low levels: Neuralink's single-chip power is only 6 mW, and Samsung's use of 4 nm process technology for its next-generation chip is driven primarily by energy efficiency considerations. Yet high-throughput recording (thousands of channels in parallel) and wireless data transmission continuously consume power — 60% of power is consumed by data transmission. While the academic community has improved communication energy efficiency nearly 10-fold over the past decade and boosted data rates to Gbps levels, the conflict between power and performance remains far from resolved.

This contradiction places higher demands on chip interconnect methods and integration density — shorter interconnect paths mean lower transmission power, while higher integration requires connecting more channels in a smaller area, both pointing toward high-density, short-interconnect advanced packaging solutions.

3. Three Challenges Overview

Facing these challenges, the global academic and industrial communities are pursuing a dual-track approach: improving chip energy efficiency and integration on one side, while breaking through micro-assembly process bottlenecks on the other.

▲ Continuous improvement in BCI energy efficiency is the foundation for industrialization

4. From Lab to Clinic: Domestic and International Progress

Taking implantable BCIs as an example, overseas development is led by Neuralink. As of 2025, Neuralink has completed a cumulative 21 human implantations. Samsung Electronics has secured the order for Neuralink's fourth-generation brain implant chip, using a 4 nm process, internally codenamed "O1," targeting mass production in 2027.

On the domestic front, China's independent BCI R&D is progressing from multi-point breakthroughs toward industrialization: On June 1, 2026, China's National Medical Products Administration (NMPA) officially approved the clinical trial application for the world's first invasive BCI chip — a landmark decision that could give China an early advantage in the BCI race[4].

Table: Key Domestic BCI Institutions and Progress

Institution	Stage / Role	Achievements
Westlake Lingxi Technology (Incubated by Westlake University)	Chip design → industrialization phase	Self-developed 40 nm BCI chip, 256-channel tape-out verified, 1024-channel chip architecture showcased at ISSCC 2026; targeting Chinese-language neural prosthesis, system latency < 10 ms, decoding ~30 Chinese characters per minute; selected for MIIT "Open Competition" program 2025; completed tens of millions RMB Angel+ funding round in June 2026[2]
Neurohope Technology (Shanghai)	Invasive product approved → commercialization	World's first invasive BCI medical device NEO-ONE SCI approved for market in March 2026; as of June 2026, 36 implant surgeries completed, cumulative safe implant days exceeding 14,000, 68.8% of patients showing significant manual function improvement 6 months post-surgery; STAR Market IPO accepted, proposed fundraising of RMB 2.5 billion, aiming to become "BCI First Stock"
NeuroTiger Technology (Shanghai)	Entered Class III medical device registration clinical stage	Fully implantable, fully wireless, fully functional BCI system officially launched GCP registration clinical trial at Huashan Hospital in July 2026, the first domestic subdural surface-implantable BCI product to enter national Class III medical device registration clinical trials; collaborating with 15 Grade-A hospitals in multi-center study; Jiangxi "Super Factory" with planned area of 14,300 m² targeting mass production in H2 2026, achieving 10,000-unit stable delivery capacity

These developments show that BCI is moving from "lab-feasible" to "repeatably manufacturable." And this leap cannot happen without the support of precision bonding technology.

5. Equipment Support: From "Connecting Electrodes" to "Building the Entire Micro-System"

The "miniaturization" of BCI means making the entire system small — fitting electrodes, chips, wireless modules, and power management into a coin-sized implant. This involves three levels of interconnect challenges:

Table: Three Interconnect Levels of BCI Miniaturization

Interconnect Level	Core Requirement	Key Process
Level 1: Chip ↔ Electrode (Front-end acquisition)	Connecting thousands of contacts on flexible electrodes one-to-one with ASIC chip pads	Sub-micron flip-chip bonding
Level 2: Multi-Chip System Integration (Processing & communication)	Interconnecting multiple chips (ASIC, power management, wireless communication) within a tiny implant	System-in-Package (SiP), heterogeneous integration
Level 3: Implant Encapsulation (Long-term reliability)	Hermetic sealing of the entire module to ensure years of stable operation in the body	Thermal management, stress control, material compatibility

The three levels correspond to chip flip-chip bonding, system-in-package, and hermetic sealing, each independent yet interdependent. This means micro-assembly equipment must simultaneously meet requirements for accuracy, force control, thermal management, and process flexibility.

In BCI development, early prototyping stages often use temporary adhesive fixation for chips, quickly establishing functional verification — low cost, fast iteration. However, adhesive bonding is unsuitable for human implantation environments; the final solution must transition to biocompatible bonding (such as flux-free eutectic soldering).

This means that in the implantable medical device field, the process choices made during early R&D are often not the final answer; equipment must have sufficient flexibility to accommodate mid-course changes. This imposes clear "process upgradeability" requirements on bonding equipment: