OCP Global Summit 5/6 | Copper Retreats into the Package, and 1–3 m Goes to Active Copper First
OCP Global Summit paid series 5/6: where 448G electrical interconnect meets slow-and-wide optics at 1–3 m — how far copper retreats, where optics come in, and who ships first
Free section
Part 2/6 of this series showed that the same footprint can take either optics or copper, and 4/6 covered the reliability gap between copper and optics. This piece asks about distance: at 448G per lane, how far does copper retreat, where do optics come in, and who ships first?
The market's standard answer is that after 224G passive copper has only about 1 m left, 448G is shorter still, and optics should take over inside the rack. For why 448G is the dividing line, see ECOC2026 | Fast-and-Narrow vs. Slow-and-Wide: 448G Is the Watershed.
At the 2026 OCP Global Summit, both the copper side and the optics side brought new numbers.
First, TE Connectivity simulated the same 350 mm 448G backplane channel: with the connector mounted directly on the package (CPC) it is about 6 dB, and with it on the mainboard next to the package (NPC) about 12 dB. Luxshare-Tech measured a 1.6 m cable-backplane segment with clean insertion loss out to 100 GHz.
Second, the OCP SROI workstream's copper reach table: 200G/lane with a retimer reaches about 2.5 to 3 m, and beyond 3 m optics are required; 400G PAM4 is projected at about 1 m, with CPC probably the only option left.
Third, slow-and-wide optics brought three ways to cut fiber count: LightXcelerate carries 57 channels on just 3 multicore fibers, the OCI MSA uses bidirectional single fiber with four wavelengths in each direction, and Credo showed a blind-mate optical backplane with 1 million fibers.
STT's view: at 448G the 1 to 3 m range will go to active copper first; for optics to win this range, the key is the gearbox that connects to the SerDes.
▲ Public reach figures for each medium in the 448G generation on one axis: passive copper at about 1 m and needing CPC, active copper covering 1 to 7 m, slow-and-wide optics starting at 10 to 50 m (Image source: Simple Tech Trend original; data from TE Connectivity, Luxshare-Tech, OIF, Lightelligence, AMD, Molex, Samtec, Astera Labs, Credo, Avicena and AttoTude 2026 OCP Global Summit presentations)
Why slow-and-wide was once blocked by fiber count: It Is Not Physics That Kills Slow-and-Wide, It Is 60,000 Fibers (in Chinese); how copper and optics divide scale-up: 2026 OCP APAC Summit | SemiAnalysis | Scale Up Sophistry.
In the paid section
Reach breakdown: published lengths and approaches for seven 448G copper links
Bandwidth breakdown: connector bandwidth against the two thresholds set by PAM6 and PAM4
Conversion breakdown: three modulation formats against one SerDes, and where the gearbox sits
Latency and power breakdown: the cost of retimers, gearboxes and SerDes between 1 and 3 m
Fiber breakdown: three answers to the 60,000-fiber problem and the challenges each leaves behind
Map: every vendor's published reach on one axis
Timeline breakdown: who ships first between 2026 and 2030
Testimony from the other side: what the copper sellers and the optics sellers each concede
Differences from our ECOC calls: one revision, two partial revisions, one reinforcement
Taiwan supply chain: Taiwanese names on the slides, where they sit, and the two real bottlenecks
Counterarguments and eight signals to watch
──────────────────────────────
That is the summary of this piece.
STT's full analysis — how far 448G copper retreats, where slow-and-wide optics come in, who ships first, and four differences from our ECOC series — is in the paid section, with 7 original charts.
👉 Subscribe to the STT paid section to read the full version:
──────────────────────────────
This article is for technology and industry trend analysis only and does not constitute investment advice.
Comments