AI model training

3 statements across 2 episodes · 0 bullish · 0 bearish · 2 people on the record · first statement Feb 6, 2025 by Nat Bullard · across every show →

Everything said about AI model training, oldest first

Feb 6, 2025 neutral
Assertion Supported
Bullard: Energy accounts for only 1.7% to 6.3% of AI training costs
“The energy component of that comes way after the AI accelerator chips, the other server components, the interconnection costs, and it's somewhere between 1.7 and 6.3% of the cost of training and experimentation.”
Nat Bullard Feb 6, 2025 ▶ 45:56 More 2025 trends: DeepSeek, plug-in hybrids, and curtailment
Feb 6, 2025 neutral
Assertion Supported
Bullard: R&D and equity represent up to 50% of AI training costs
“R&D staff, and also in this case, they include, as they should, the cost of equity is between, like, 30 and almost 50% of running of training a model and of doing experiments with it.”
Nat Bullard Feb 6, 2025 ▶ 45:36 More 2025 trends: DeepSeek, plug-in hybrids, and curtailment
Dec 18, 2025 neutral
Insight
AI training compute cycles cause massive power fluctuations unlike inference workloads
“And some of the people in the energy space may know that there are massive energy fluctuations or power fluctuations we will see in data center usage when the GPUs go from this computational intensive phase where you're learning the model weights to this commu…”
Dr. Ben Lee Dec 18, 2025 ▶ 18:09 Will inference move to the edge?
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 200 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.