EN
Back to the archive

The archive · Hardware & Devices · Strategic decision · 2025-2026

Athos Silicon winds down: Mercedes safety-chiplet spinout fails to raise

Mercedes-Benz spun out its safety-chiplet team as Athos Silicon in 2025; a year later the board dissolved the startup after fundraising failed.

Athos Silicon

The betAutomakers would buy chiplets that deliver fail-safe redundancy inside one package at 10-20x lower power, starting with drones and scaling to autonomous driving.No longer exists

What the business is

Athos Silicon was developing functional-safety chiplets for autonomous driving and other autonomy systems: multiple chip instances in a single package monitor one another through a patented voting mechanism, and workloads can move between chiplets so a hardware failure does not interrupt service. After Arm acquired chiplet-hub maker DreamBig, Athos dropped its plan to buy third-party chiplets and switched to taping out its own SoC chiplet from third-party IP, using several instances of one tape-out on an organic substrate to control cost.

Starting capitalMercedes-Benz made an undisclosed but 'significant' investment in the spinout and held a minority stake; Athos said in 2025 that it would raise venture capital from other investors, but no external round was ever disclosed before the board wound the company down.

How it started

Engineers at Mercedes-Benz Research & Development North America spent about five years developing chips meant to be safe enough for cars while using less energy. In September 2025 Mercedes-Benz spun the group out as Athos Silicon, based in Santa Clara, California, transferring the team's intellectual property and making an undisclosed 'significant' investment while keeping a minority stake. CEO Charnjiv Bangar said independence mattered because Athos needed to reach other carmakers, competitors of Mercedes, and to raise venture capital of its own.

What happened

Athos's architecture attacked the redundancy problem: cars often run critical self-driving functions on two or more separate chips so a failure has a backup, but chips communicating across a circuit board cost power - 10 to 20 times more, by Athos's claim, than multiple chiplet instances bound in a single package. The startup planned a patented voting scheme in which chiplets monitor one another and workloads move between them if one fails, targeting autonomous driving in the long term and smaller systems such as drones in the near term. Its original plan was to buy and integrate third-party chiplets, including DreamBig's chiplet hub; after Arm acquired DreamBig, Athos redrew the roadmap to tape out its own SoC chiplet from third-party IP, with multiple instances of a single tape-out on an organic substrate to cut cost.

How it ended up

In early September 2026 former CEO Charnjiv Bangar told EE Times that Athos Silicon had ceased operations after its board decided to wind down and dissolve the company. 'Despite extensive efforts, we were unable to secure the financing required to continue the productization and commercialization of our technology,' he said, adding that the technology was consistently well received by the technical community and prospective partners, and that 'ultimately, the challenge was financial and cap-table related rather than technical.' Under the licensing agreement with Mercedes-Benz, the original IP licensed from Mercedes reverts to Mercedes-Benz, while IP developed after the spinout remains with its inventors.

Background

Mercedes-Benz spent about five years incubating a Silicon Valley team that designed fail-safe, power-efficient chips for automated driving. In September 2025 it spun the engineers out as Athos Silicon, a Santa Clara startup that received the group's intellectual property and an undisclosed 'significant' investment from Mercedes-Benz, which remained a minority shareholder.

Athos's bet was that automakers would buy a neutral supplier's safety chiplets: multiple chip instances bound in a single package, watched over by a patented voting mechanism, could replace the separate backup chips cars use today at a claimed 10-20x power saving. It targeted autonomous driving in the long term and smaller systems such as drones first, and it initially planned to integrate third-party chiplets - including DreamBig's hub - before Arm's acquisition of DreamBig pushed it to tape out its own SoC chiplet from third-party IP.

The startup said in 2025 that it would raise outside venture capital, but no round was ever disclosed. In early September 2026 former CEO Charnjiv Bangar confirmed that Athos had ceased operations: the board voted to wind down and dissolve the company after failing to secure financing for productization and commercialization.

'The technology was consistently well received by the technical community and prospective partners,' Bangar said. 'Ultimately, the challenge was financial and cap-table related rather than technical.' Original IP licensed from Mercedes-Benz reverted to the carmaker, while IP developed after the spinout remains with its inventors.

What has to be true

  • Athos sat between two funding realities: Mercedes-Benz was only a minority shareholder by design, and the outside venture capital Bangar promised in 2025 never materialized.
  • Chip productization needs long, capital-heavy tape-outs, leaving a young startup little room to bridge from favorable technical reviews to revenue.
  • After Arm acquired DreamBig, the roadmap shifted from integrating third-party chiplets to taping out an in-house SoC, raising the cost and lengthening the timeline of the bet.
  • Neutrality meant no Mercedes purchase order could carry the company while it courted rival automakers as customers.
  • Bangar's own account separated the two failure modes: the technology was well received, but the board based its decision on the company's financing position and cap table.

What can be applied

Technology that is 'consistently well received' still needs investors willing to finance productization; a cap table can kill a company whose engineering works.

Aftermath

As of early September 2026 Athos Silicon was winding down and dissolving, with former CEO Charnjiv Bangar attributing the closure to financing rather than technology. Under the licensing arrangement with Mercedes-Benz, original IP licensed from the carmaker reverted to Mercedes-Benz, while IP developed after the spinout remains with its inventors. No customers, pilots or external funding round were disclosed during the startup's roughly one-year independent life.

Sources

spotted an error? The archive wants to know.

Your turn

You just read one. Describe what you are building, and see who is betting on the same thing.

Free account · 3 free questions · no card

Related cases