Nvidia, SPIL plan Taiwan optical engine expansion as Spectrum-X CPO ramps
In brief
- Nvidia and SPIL plan to expand optical engine capacity in Taiwan, according to Crypto Briefing.
- SPIL handles chip-scale packaging and testing; TSMC fabricates silicon photonics through its COUPE platform.
- Spectrum-X CPO switches began shipping in June 2026 and reached full production by mid-August.
- CoreWeave, Lambda and Oracle Cloud Infrastructure have adopted the Spectrum-X platform.
Two Taiwanese partners, two production stages
SPIL's role in Nvidia's supply chain covers chip-scale packaging and testing. TSMC handles silicon photonics fabrication through its COUPE platform, so Nvidia has two major Taiwanese partners working on different stages of production.
According to Crypto Briefing's analysis, a larger optical engine footprint would tie SPIL more tightly to the Spectrum-X product line. For Nvidia, it adds packaging capacity in a region where it already leans on TSMC.
Why switches are moving off copper
Co-packaged optics (CPO) puts optical engines directly onto switch silicon instead of relying on separate, pluggable optical transceivers. The reason comes down to power.
As bandwidth climbs, copper eats more power and struggles to carry signals far without degradation.
Pluggable transceivers aren't a clean fix either, since they bring their own power draw, heat and failure points. Nvidia's CPO approach reportedly delivers a 5x improvement in power efficiency over traditional optical modules. It also uses 4x fewer lasers, offers 10x higher mean time between failures and supports high-density, liquid-cooled builds.
Spectrum-X shipments and what comes next
Nvidia began shipping select Spectrum-X CPO switches in June 2026. In larger configurations such as the SN6800, the platform reaches throughput of up to 409.6 Tb/s. The switches support AI factory networking on Nvidia's Vera Rubin platform, and cloud providers including CoreWeave, Lambda and Oracle Cloud Infrastructure have adopted it.
Nvidia executives have said they're optimistic about ramping capacity in the second half of 2026. At SEMICON West in October 2026, the company unveiled prototypes of next-generation optics-on-interposer (OOI) CPO, which aims to push bandwidth density higher and cut power requirements further.
The approach has a catch.
Packaging optics directly onto switch silicon raises the stakes on manufacturing yield. A single defect can ruin an assembly that costs far more than one pluggable module, which isn't a small consideration at this scale. Crypto Briefing described the wider picture as Taiwan's role in AI hardware widening beyond chip fabrication, with advanced packaging and testing becoming a strategic chokepoint.
Frequently asked questions
What is co-packaged optics (CPO)?
Co-packaged optics puts optical engines directly onto switch silicon instead of using separate, pluggable optical transceivers. Optical engines convert electrical signals into light so data can travel between chips and switches. Nvidia's CPO approach reportedly delivers a 5x power-efficiency improvement over traditional optical modules.
What do SPIL and TSMC each do for Nvidia's optical switches?
SPIL covers chip-scale packaging and testing in Nvidia's supply chain. TSMC handles silicon photonics fabrication through its COUPE platform. Together they give Nvidia two major Taiwanese partners that work on different stages of production.
What are the risks of packaging optics directly onto switch silicon?
It raises the stakes on manufacturing yield. A defect can affect an assembly that's far more expensive than a single pluggable module, so a production error costs more than it would with traditional transceivers.
