IBM
Arvind Krishna forecasts 75% to 90% of all token usage will go to open-weight models, with the Chinese-Western open-weight gap at most 6 to 12 months.
Krishna says a Chinese open-weight model (transcribed 'Kimikay3', ~800B parameters) on own GPUs costs about $1 to $2 per million tokens versus ~$20 for a prior-generation frontier model.
Krishna says agents write probably half the code inside his company, and that it is superior to what was written before.
Krishna says DARPA's quantum benchmarking program is down to about 13 companies from a couple of hundred, with ~2030 the midpoint for quantum breaking today's cryptography.