NVIDIA just committed $4 billion to two photonics companies that most people in tech have never heard of. That bet is worth understanding, because it is a signal that the entire AI hardware stack is about to change.
The part that surprised me when I went deep on this research: the real constraint holding back the next generation of AI is the wires. Not the chips. The copper cables carrying data between GPUs inside the world’s most powerful training clusters are hitting a hard physical ceiling. Heat, resistance, energy cost.
The physics of electricity is running out of room at exactly the moment AI needs more of it. What engineers are racing toward is photonics. Moving data as light.
No resistive heat, no signal degradation, and in some configurations, over 90% less energy for the same computation. MIT published that number and I had to reread the paragraph twice.This video breaks down how photonic computing actually works, why NVIDIA just put $4 billion behind it, and what the transition means for engineers building systems in this industry.