Resources

2026 Predictions – GPUs diversity and Ethernet in Scale-Up

DriveNets’ VP Product Marketing, Dudy Cohen, shares three key predictions for AI infrastructure in 2026, examining the shifting landscape of GPU providers, the growing importance of networking in AI deployments, and Ethernet’s expanding role across different AI infrastructure domains.

Chapters:

Read the full transcript

Hi, my name is Dudy Cohen and I’m from Drivenets and these are my three predictions for 2026.

Prediction number one
2026 will be a much more heterogeneous year in terms of GPU providers.
We will no longer see Nvidia as the sole player here. We will see other contenders like AMD, Google and others getting into this market and gaining market share.

Prediction number two
Number two, we will see networking becoming a key consideration for anyone that builds an AI infrastructure.

Prediction number three
And the third prediction? I think in 2026 we’ll see Ethernet evolving as the leading technology for any type of networking required for AI infrastructure. This is of course the case today with Scale out and Scale across Scale out thanks to the introduction of scheduling that evolve Ethernet to the performance required by AI infrastructure and Scale across because of the combination of shallow buffer and debuffer, allowing Ethernet to overcome the latency introduced when connecting remote data centers. But I think in 2026 we will also see Ethernet in the Scale up domain and this is as an alternative to proprietary technologies with the introduction of ESUN and the evolution of Ethernet as a valid technology. So 2026 is going to be a very interesting year in the AI infrastructure domain. Stay tuned for more.
Want to learn how DriveNets is reshaping AI networking infrastructure?

Explore DriveNets AI Networking Solution