TickerTain
TickerTain
NewsroomShortsPortfolioConvergence
NewsroomShortsPortfolioConvergence
←
▶ 1:52 · $RCAI · RCAI CTO explains pivot to pre-training for developer sovereignty
episode briefing
SemiAnalysis

Training a 400B Model on 2,048 Blackwell GPUs for $20M | Researcher Conversations at GTC

2026-07-15 · 1 company · 4 thematic
sentiment
0 bull0 bear1 neu
speakers
lucas atkins

Lucas Atkins is CTO and head of research at RCAI, an AI research company focused on pre-training open-weight models for developers and enterprises. He previously worked on post-training of open-source models like Mistral and Qwen.

quote
“for us when we were largely doing just post training on top of other open models. So Alama Mistrol back in the day and then later on Quen was a very popular one cuz we were trying to stay sub you know 20 billion parameters. We reached a po…”
—lucas atkins
now playing · $RCAI
$RCAIneutral· highlucas atkins
RCAI CTO explains pivot to pre-training for developer sovereignty
RCAI moved into pre-training to control the full stack for developers, avoiding reliance on Chinese base models and Western labs whose revenue doesn't depend on open-weight model…