Turiyam’s inference chip promises 10‑20% lower AI costs than Nvidia
Bengaluru startup Turiyam is shipping a full‑stack inference processor that charges by finished output, not tokens, claiming 10‑20% cheaper AI compute than Nvidia GPUs. If its pricing holds, a firm spending ₹2 cr annually on AI could cut spend to about ₹40 lakh, reshaping enterprise AI budgets.