Samsung Builds Neuralink's First Two-Way Brain Chip on 4nm Node as Musk Adds a Second Foundry

Samsung Foundry has reportedly begun developing Neuralink's most advanced brain implant chip yet, a project that would hand the South Korean giant its first Neuralink contract and add a second manufacturing source to Elon Musk's brain-computer interface supply chain. Multiple Korean media reports, led by Hankyung and KED Global, detail a chip codenamed O1 built on Samsung's 4nm process, with test chips already in production and mass production targeted for late 2027 . While neither company has officially confirmed the agreement, the outlines of the chip and its place inside Samsung's broader foundry revival are unusually well documented by industry leaks.

The chips themselves: O1 on 4nm

The fourth-generation Neuralink chip is being developed under the internal project codename O1, using Samsung's 4-nanometer fabrication process . Samsung began R&D late in 2025, and by May 2026 the first test chips had entered production—a sign the design is already in a mature enough state for trial runs .

Shipping and volume production follow a fast timeline. Test chips are expected to ship in the first half of 2027, and if validation succeeds, mass production could start as early as the second half of 2027 . The choice of a 4nm node—less cutting-edge than the 3nm process Samsung offers for smartphone processors—is likely deliberate: more mature nodes offer higher yields and more predictable behavior, which matters for safety-critical medical implants .

Two-way communication and the vision puzzle

The biggest architectural difference from previous Neuralink chips is bidirectional communication. Earlier generations were primarily designed to read brain signals, decoding neural activity and transmitting it out for processing. The fourth-generation chip is designed to read and write: it can both interpret brain signals and send electrical stimulation back into neural tissue .

That two-way capability is technically essential for any application that involves feeding information into the brain rather than just extracting it. Neuralink has long listed restoring vision as a goal, and the concept—stimulating the visual cortex with enough precision to produce recognizable visual percepts—requires precisely the kind of controlled stimulation the bidirectional chip would theoretically enable . The Korean reports do not explicitly link O1 to a vision-restoration product, but they describe the chip as a brain-implantable device for neurological treatment and note its therapeutic stimulation potential, which is the foundation that vision applications would need .

Why TSMC is no longer the only factory

The Neuralink order represents a diversification away from single-foundry dependence on TSMC. Reports indicate that previous-generation Neuralink chips were manufactured exclusively by TSMC, but with the fourth-generation chip, Samsung Foundry is being brought into the production mix, creating a dual-sourced supply chain that offers greater stability and resilience .

That pattern is not unique to Neuralink. Musk has publicly acknowledged that Samsung already makes Tesla's A14 chips, while TSMC handles the A15. The $16.5 billion Tesla-Samsung deal signed in July 2025 explicitly shifted production of the next-generation AI6 chip to Samsung's new Texas factory, a move Musk called "hard to overstate" in strategic importance . Across Musk's companies, the preferred arrangement increasingly looks like two foundries rather than one.

Samsung's fragile foundry turnaround

The Neuralink contract arrives at a precarious moment for Samsung's foundry division. The business has been losing customers to TSMC at an alarming rate. Google's Tensor G5 processor for the Pixel 10 moved to TSMC on a 3nm node, a loss widely attributed to Samsung's lagging yields and limited semiconductor IP . Multiple South Korean AI chip designers have also shifted or dual-sourced their most advanced designs to TSMC .

Samsung has been fighting back with a string of high-profile wins that collectively signal its manufacturing capability can still attract marquee clients:

• Nintendo Switch 2 custom processor
• Tesla AI5 chips as a joint production arrangement with TSMC
• Japan's Preferred Networks (PFN) 2nm AI chip order, poached from TSMC
• Galaxy Z Flip 7 Exynos 2500 on Samsung's 3nm node
• The Neuralink O1 chip, which would add a bio-medical implant to a portfolio currently heavy on vehicles, gaming, and data-center AI

The aggregate picture is one of a struggling foundry landing just enough strategic contracts to remain a credible alternative to TSMC on advanced nodes. The Neuralink order, if confirmed, would reinforce that narrative by opening a new application domain Samsung currently does not serve.

The Tesla deal as context: $16.5 billion and a Texas factory

The Neuralink contract does not exist in isolation. It is the latest extension of a deepening Samsung-Musk alliance that reached its most commercially significant moment in July 2025, when Samsung signed a $16.5 billion multiyear agreement with Tesla for AI semiconductors. The deal runs through 2033 and dedicates Samsung's under-construction Taylor, Texas fabrication plant to producing Tesla's next-generation AI6 chip, which is designed for autonomous driving, AI data centers, and humanoid robots .

Musk confirmed the arrangement on X and added that Tesla would contribute to optimizing manufacturing efficiency at the Texas facility, an unusually close operational relationship between a chip designer and its foundry . Samsung had previously manufactured Tesla's A14 chips, but bringing the AI6 flagship to Samsung—after TSMC handled the intermediate A15—was the inflection point that signaled a deeper strategic commitment .

The Neuralink order extends the same foundry relationship beyond Tesla's automotive and robotics businesses into medical devices, forming a near-complete chip supply line across Musk's corporate ecosystem.

Reality check: what we don't know

Everything reported about the Samsung-Neuralink contract comes from Korean industry media citing unnamed sources. Neither Samsung nor Neuralink has issued an official statement as of June 16, 2026, and until they do, the timelines, technical specifications, and even the existence of the contract should be treated as provisional .

The mass-production targets are particularly sensitive: Samsung itself has said mass production of the Tesla AI6 chip at its Texas plant is not scheduled until the second half of 2027, and bringing a second high-volume implant chip online in the same window would test the foundry's capacity .

Neuralink has separately announced plans to begin high-volume production of brain-computer interface devices in 2026 and transition to fully automated surgical procedures, which suggests the company is moving toward commercialization regardless of which chip generation or foundry partner it relies on . Whether the O1 chip is the hardware inside those scaled-up implants—or a later generation still years from deployment—is not yet publicly known.