$LITE CEO just said something the market needs to hear
Over the last two months, non-NVIDIA customers have shown a marked surge in interest in NPO, not CPO!!!
This matter a lot rn because at 1.6T and 200G per lane, copper hits its physical limits even with retimer technology. You simply canât cover reasonable distances anymore. Optics becomes mandatory, not optional.
But the more interesting dynamic is the ASIC vendors.
They canât compete with NVIDIA on CUDA, NVLink, and NCCL.
So theyâre leapfrogging, adopting more aggressive optical backplane architectures to try to overtake on the curve.
The results are obviously gonna be insane for these smaller vendors because they may actually deploy more lasers per rack than NVIDIA customers.
Fewer total racks but more optical links per rack.
The optical channel volume could be larger than the headline market share suggests.
This reframes the entire optical interconnect opportunity.
NPO is the architecture that accelerates opticalization across the entire AI server, every rack, every server, eventually every chip.
For
$LITE,
$COHR,
$MRVL, and
$GLW, the market size ultimately depends on how many optical links get deployed inside each rack. From that angle, NPOâs TAM could actually exceed CPOâĶbut letâs see