Both Sterlite and HFCL continue to get big orders from MNC's/Hyperscalers.
Why are optics replacing copper in DC build out?
A few compounding factors are pushing optics into places copper used to own in AI data center builds:
Copper hits a physics wall as speeds rise. At today's high per-lane rates, passive copper simply can't span the distances it used to — at 200+ Gb/s per lane, traditional passive copper can no longer span beyond a single server rack, and sometimes not even within a rack. Higher data rates mean more signal loss, more power burned just driving the signal, and more heat. All getting worse as clusters demand faster links.
The traffic problem got much bigger. AI training and inference generate enormous "east-west" traffic — GPU-to-GPU chatter within and across racks — as clusters scale into the hundreds of thousands of accelerators. The primary data-center bottleneck has shifted from transistor performance to interconnect bandwidth and latency, sometimes called the "I/O wall": compute scales through packaging and process tricks, but chip-to-chip I/O doesn't keep pace the same way.
Power efficiency is now existential, not just nice-to-have. In modern AI clusters, moving data can consume up to half of total system energy, and unlike copper, light doesn't suffer from electrical resistance. So, every watt saved on interconnects goes straight back into compute headroom or lower cooling costs, which matters enormously at hyperscaler power budgets.
September 1, 2026 2K 1
