Would also be very curious. I met Arista’s CFO and VP of something in a group with other investors post Q4 and someone asked about CPO. They just explained it as plug and play if it works, not a positive not a negative.
I don't think the impact is only on Arista networks. It will be broader for all network companies within datacenter (AI) plays and perhaps later also the rest of the optical transceivers market. Arista cooperates with Broadcom, and then you have Cisco (which recently signed a partnership with Nvdia).
Very interesting article! I am going to show my partial ignorance with this question: How much more flexibility does this allow regarding the distances between different server stacks/racks and entire buildings? One of the great challenges with building these enormous AI "farms" is that it can take a long time to supply enough power to one particular location. Could "going photonic" with very high bandwidth switching be a way to decentralize those operations, and bring parts of them closer to existing power plants? Or am I completely off here?
So is TSMC’s silicon photonics still bad in comparison to Global Foundry in your opinion? Or is TSMC coming to eat their breakfast, lunch and dinner there?
Been out of this for a long time. At optical MEMS company (remember those train wrecks?) into and through 'the Telecosm'. One project that was tossed to me was a team that suppose to be developing a semiconductor material to allow for electrically tunable lasers...different frequencies. 'Go find a customer!' Did that technology ever happen? We went through a buyout and the project was taken over. Never found out what happened.
Would you mind elaborating the impact on Arista? Thx!
Would also be very curious. I met Arista’s CFO and VP of something in a group with other investors post Q4 and someone asked about CPO. They just explained it as plug and play if it works, not a positive not a negative.
I don't think the impact is only on Arista networks. It will be broader for all network companies within datacenter (AI) plays and perhaps later also the rest of the optical transceivers market. Arista cooperates with Broadcom, and then you have Cisco (which recently signed a partnership with Nvdia).
So those boxes at the end? Those all slaughter Arista’s primary product line.
CPO will be mainstream in the future.
Very interesting article! I am going to show my partial ignorance with this question: How much more flexibility does this allow regarding the distances between different server stacks/racks and entire buildings? One of the great challenges with building these enormous AI "farms" is that it can take a long time to supply enough power to one particular location. Could "going photonic" with very high bandwidth switching be a way to decentralize those operations, and bring parts of them closer to existing power plants? Or am I completely off here?
So is TSMC’s silicon photonics still bad in comparison to Global Foundry in your opinion? Or is TSMC coming to eat their breakfast, lunch and dinner there?
Been out of this for a long time. At optical MEMS company (remember those train wrecks?) into and through 'the Telecosm'. One project that was tossed to me was a team that suppose to be developing a semiconductor material to allow for electrically tunable lasers...different frequencies. 'Go find a customer!' Did that technology ever happen? We went through a buyout and the project was taken over. Never found out what happened.