Gotta say, I'm not exactly chomping at the bit to have this in my portfolio when the IPO drops. Some really ugly-looking numbers here.
pull down to refresh
pull down to refresh
Gotta say, I'm not exactly chomping at the bit to have this in my portfolio when the IPO drops. Some really ugly-looking numbers here.
Ouch..
Yikes it’s not looking good at all!!
https://twiiit.com/wallstengine/status/2104715031837622595
Well, does explain the call for slowing down model development...
Paywalled
The core structural problem for frontier lab IPOs is that training capex depreciates on a 12-to-18-month cycle. Unlike classic software (build once, 85% gross margin forever) or infrastructure (30-year fiber/grid depreciation), a $5B frontier model cluster is economically obsolete the moment the next open-weight or distilled generation drops inference cost by 10x.
Going public isn't an exit here — it's tapping retail/institutional equity markets because private credit and hyperscaler circular cloud-credits can't absorb the next capex step alone.