OCP Global Summit 3/6 | Laser Specs Are Now Written by System Builders
OCP Global Summit paid series 3/6: Laser sources — external sources, integrated lasers and multi-wavelength sources: where each goes, whether to count power or wavelengths, and who sets reliability and the thermal interface
Free Section
Part 1/6 of this series left one question open: does the laser have to sit outside the package? Part 2/6 listed the ELSFP as the last layer of the interface. This part asks: where does the laser actually go, and who decides?
Around ECOC 2026 the market was still comparing watts: CPO uses a front-panel external source, and the laser chip has to reach 300–400 mW. For the integrated-laser alternative, see CPO Lasers Don't Have to Chase 400 mW: NPO Integrated Lasers. A month later, at the 2026 OCP Global Summit, the question on stage changed.
First, Lumentum's Christian Urricariet split "where the laser goes" and "where the optical engine goes" into two independent architecture axes — and Lumentum itself has ELSFP in volume production.
Second, Nokia laid out laser-to-modulator loss as a ladder: about 0.25 dB for monolithic integration, about 2 dB for a front-panel ELSFP.
Third, Ayar Labs and Wiwynn presented a liquid-cooled ELSFP and stated plainly that Wiwynn first set the coolant flow, temperature and thermal resistance, and Ayar Labs then designed the laser.
STT's view: laser specs are now written by the system builders.
▲ Loss, thermal zone and serviceability for four laser placements: monolithic, on-PIC, in-package and front-panel ELSFP (Image source: Simple Tech Trend original; data from Nokia 2026 OCP Global Summit presentation)
For what the OCI scale-up optical interconnect spec asks of external sources, see Tech Analysis | Meta, Broadcom and AMD Define the OCI 200G Line-Side Spec; for the reliability results of third-generation CPO, see 2026 OCP APAC Summit | Broadcom | CPO's Third-Generation Report Card.
In the Paid Section
Placement: where twelve designs shown at OCP put the laser
Power: how much more output each step away from the modulator costs, and where NPO and CPO power needs differ
Wavelengths: how many laser channels per Tbps two public designs need, and how many ELSFPs the compute side of a 1,024-ASIC pod needs
Reliability: the root cause of Meta's ELS failures, and how far down mode-hop screening must go
Thermal interface: who defines the 5°C between 45°C coolant and a 50°C laser die
Testimony from rivals: the fallback each camp keeps — the ELSFP seller, the integrated-laser camp and the InP fab owner
Differences from our ECOC calls: three revised, one maintained
Taiwan supply chain: the Taiwanese names that appeared at OCP, their roles, and the two real bottlenecks
Counterarguments and eight signals to watch
──────────────────────────────
That's the summary of this piece.
STT's full analysis — who decides laser placement, power, wavelengths, reliability and the thermal interface, and four differences from our ECOC series calls — is in the paid section, with 6 original charts.
👉 Subscribe to the STT paid section for the full version:
──────────────────────────────
This article is for technology and industry trend analysis only and does not constitute investment advice.
Comments