Staera launches with $4.6 million to link AI chiplets with light instead of copper
San Francisco startup Staera Semiconductor says its Lightwire optical links, built into the chip, will move data between chiplets faster and with less power than copper.
Staera Semiconductor Inc., a San Francisco startup, launched on 10 November 2026 with $4.6 million in pre-seed funding to develop optical links between the chiplets inside AI accelerators, SiliconANGLE reports. Singapore's Openspace Capital led the round, joined by SGInnovate, Black Kite Capital, 28 Holdings, In Group and the financial software company Decimus Technologies.
In high-end AI accelerators, several chiplets are connected through an interposer or a smaller, cheaper bridge, and both carry data over tiny copper wires. Staera wants to replace that copper with light. Chief executive Andrew Kim says copper performs far worse once a link has to span anything from millimetres to a metre, the distances that count most inside AI accelerators, and that add-on optics bring extra power and complexity. The company says optical links are faster, use less power and keep their speed over longer distances, which would let chipmakers build larger accelerators.
Staera's first product, Lightwire, sends data through waveguides made of silicon nitride, a transparent material that SiliconANGLE says loses less light than many alternatives. The light comes from tiny LEDs made of gallium nitride. Kim says Staera builds the light source into the chip, so the whole optical link sits inside it, and that the link shrinks, runs cooler and speeds up with every new generation.
Staera says Lightwire moves data between an accelerator's chiplets at up to 8 gigabits per second and fits more than 600 optical links in each square millimetre. On those figures, SiliconANGLE reports, an HBM4 memory stack using Lightwire could transfer up to 24 terabits per second, against 2.8 terabits per second for current technology.
Staera says the new money will go toward bringing Lightwire to market: improving the product, growing its U.S. team and building a manufacturing supply chain. It hopes to start commercial shipments in 2030.
For information only. For the full story, check the sources: