As OFC passes, it is very clear that Nvidia is the gorilla in the industry. Just a few years back, Nvidia had little presence, if any. With Lightcounting predicting that Nvidia purchasing 4-5 million unites of 800G transceivers from their suppliers over the course of 2024, it's also clear that the road to 1.6Tbps is coming sooner rather than later. You can thank AI for all of this.
Power consumption (and power density) has become one of the most significant concerns in the industry. With AI clusters being built so dense, even the most marginal power savings can become highly relevant. Lots of talk on LPO, LRO, and CPO. Pros, cons, ecosystem disruption, etc.
800G pluggable optics are now deployed on a very large scale (especially when we're talking Hyperscalers). According to Lightcounting, 800G will overtake 400G optics shipment volume in 2024. Most of the 800G optics are focused on AI cluster and data center fabric deployments. Eventually 800G WAN use cases will emerge, but still early.
Hyperscalers, to date, have had to deploy 100G/lane architectures (in large volumes). It's clear that newer 200G/lane is now being considered, especially when you look at the likes of Broadcom Tomahawk and Marvell Nova/Nova2. However, reliability testing held this back for some time, as you'd expect.
I don't post much on here, but I would not lose sight on the fact that Lebby has publicly stated they are now engaged with more prospects than ever. The timing is right here, especially when you see the path to 400G/lane coming sooner than expected - especially with 1.6Tbps and 3.2Tbps.
Food for thought, especially as everyone bickers about the PR today about ISPR-I. Lebby and team are skating to where the puck will be... I still have firm belief.
