OCP Global Summit 1/6 | NPO Wins on Buying Power, Not Volume

OCP Global Summit paid series 1/6: NPO becomes the main battleground — who is buying, when, and how far the CPO timeline slides

Free Section

When ECOC 2026 wrapped up, the NPO-versus-CPO debate was stuck on "should we jump now?": Huawei's scorecard had NPO ahead 4:2, and the floor vote put NPO first and CPO second. Most of the people on stage that round were sellers and Chinese vendors. For the backstory, start with NPO Is Not CPO's Appetizer.

A month later, at the 2026 OCP Global Summit, the buyers took the stage.

First, Meta sent its technical sourcing manager, who said NPO goes into scale-out first as a pilot, because that layer has pluggables to fall back on.

Second, another Meta team published the closing numbers of its 51.2T CPO comparison test: 175 million port-hours, zero optical-engine failures, and an annualized interruption rate of 0.2% versus 1.22% for the pluggable control group.

Third, NVIDIA closed the NPO workshop as the contrarian, saying that anyone hoping to get multiple suppliers out of NPO is "kidding yourself"; LightCounting's forecast shows combined CPO and NPO ports only reaching the 40-million level in 2029.

STT's view: NPO is winning this round on purchasing, not physics; and the claim that "CPO has been delayed" needs to be taken apart and reread.

▲ Where each buyer at OCP stands on NPO versus CPO: position, use case and earliest timing (Image source: Simple Tech Trend original; data from 2026 OCP Global Summit presentations)

For our previous timing call, see Goldman Raises, Eoptolink Delays; for LightCounting's 2031 crossover, see ECOC2026 | 4.7 pJ/bit and 2031.

What's in the Paid Section

• Buyer map: which side nine kinds of buyers back, at which layer, and how early

• Power breakdown: four companies' power ladders, and how many watts actually separate NPO from CPO

• Reliability breakdown: Meta's 175 million port-hours of interruption data row by row, and how solid the statistics are

• Timeline breakdown: eleven dated statements on one axis

• How far the CPO timeline slides: NVIDIA's and Broadcom's own roadmaps

• Where we differ from our ECOC calls: five judgments we revise or keep

• Taiwan supply chain: the Taiwanese names on the slides and the two real bottlenecks

• Counterarguments and seven signals to watch

──────────────────────────────

That is the summary of this piece.

STT's full analysis — who is buying NPO, when, how far the CPO timeline really slides, and five differences from our ECOC series — is in the paid section, with 6 original charts.

👉 Subscribe to the STT paid section for the full version:

👉 Subscribe to STT Premium to read the full analysis on vocus (in Chinese): https://vocus.cc/article/6ad5b168fd897800013636b3

──────────────────────────────

This article is for technology and industry trend analysis only and does not constitute investment advice.