A report by SemiAnalysis on CPO and 800V DC power architecture indicates that they expect 800V DC power to be delayed until 2028, while CPO will be delayed until 2028-2029. They believe NPO and 400V DC solutions will immediately become the biggest beneficiaries. One company mentioned optimistically in the report is Amphenol. This report directly triggered a major capital migration last night: the optical communication/CPO chain ($Lumentum(LITE.US) $Coherent Corp.(COHR.US) $Applied Optoelectronics(AAOI.US) $Corning(GLW.US) ) collectively saw outflows; $Amphenol(APH.US) rose 7.29% against the broader market to close at $154.07, touching $154.92 during the session.

The report's impact lies in the fact that it provides engineering details rather than opinions: a 95% yield for a single optical engine sounds decent, but a chip needs to integrate 32 engines, and the system yield after multiplication drops to only 19.4%. Without solving this number, CPO cannot scale profitably. On the 800V DC side, cloud providers are collectively taking a wait-and-see approach to NVIDIA's single-ended solution. Meanwhile, the two "transitional routes" of NPO and ±400V happen to be Amphenol's stronghold — technology roadmap delays never mean demand disappears, it just means the money moves to a different pocket.

Of course, NVIDIA issued an urgent refutation the same day: "No delay, mass production and delivery on schedule in the second half of 2026," with Spectrum-X photonics already in mass production. However, I believe the 19.4% yield is mathematics, the refutation is a posture, and engineering problems won't disappear because of a press conference.