Iterate Studio Inc. released Lifeboat, an inference engine for large language models that includes confidential computing. The company says Lifeboat can host two to six times as many concurrent AI‑agent sessions on each GPU. In its own test a single Nvidia RTX PRO 6000 Blackwell card handled 2,048 sessions, double the baseline. Lifeboat uses fair scheduling, a larger key‑value cache and per‑session security capsules. A free developer license is available, with a paid Standard tier at $49.99 / month and a Confidential Computing tier at $499.99 / month.
Why it matters
Banks, insurers and health systems can run more AI agents on existing GPUs, avoiding extra hardware purchases.
