10+ hours of podcasts, in 5 minutes.
7 quotes from 2 episodes on All-In Podcast, each with a timestamped link to the source.
Andrew Feldman states that current AI hardware cannot keep up with demand, forcing the industry to invent new computing architectures. The global buildout now requires individual data centers the size of football fields that consume more power than midsize cities.
Surpassing Nvidia on traditional processor designs is nearly impossible, prompting companies to invent entirely different hardware.
“…the odds that you're better than Nvidia in our view are approximately zero.”
Establishing communication networks between computing clusters in space remains an unsolved engineering challenge.
“We we're not super good yet at building the clusters in space necessary for the communication between Exactly.”
From Marshall: Space Compute Cheaper Than Earth Within a Decade
The industry cannot build facilities fast enough to meet current compute requirements. Individual data centers now span the area of football fields and draw more power than midsize cities.
“The demand is way outstripping our ability to build data centers and to fill them with hardware.”
“We're talking about individual buildings the size of football fields that have more power coming into them than midsize cities…”
Achieving 20 times better performance demands fundamentally different systems because established competitors have already consumed all the easy architectural improvements.
“Our view as computer architects is if you want to be 20 times better than somebody, your architecture can't look like them. They have enjoyed and eaten all the lowhanging fruit.”
From Cerebras's Feldman: Outpace GPUs 18x by Rewriting AI Silicon
The central problem in AI processing is transferring data between memory and compute components. Faster data movement directly results in quicker answers and a better user experience.
“…the hard part is moving data from memory to compute. This is the fundamental problem in AI.”
“That means your answers are delivered more quickly. It means your engagement with the AI is more enjoyable.”
From Cerebras's Feldman: Outpace GPUs 18x by Rewriting AI Silicon
How we attribute quotes. Every quote was matched against the episode transcript, so the words and the timestamp are real (we trim filler words like "um", nothing else). The name comes from our written summary of the episode. YouTube gives us no voice-by-voice transcript, so open the timestamp to hear who is talking. See a wrong name? Tell us and we fix or remove it.
10+ hours of podcasts, distilled into one 5-minute read. Free, every Sunday morning.
One email a week. Unsubscribe with one click.