Jensen Huang told Lex Fridman in a recent podcast that agent models will not replace ERP or CRM systems; they will simply add an automation layer on top of existing licensed software. The implication is clear: software spend will remain, and the only new cost will be the compute required for agents.
OpenClaw has launched the Vera Rubin platform, which turns server capacity into a token‑oriented farm. According to OpenClaw’s own figures, one million tokens can generate up to $1,000 in revenue. That moves token farming out of hobbyist territory and positions it as a potential profit center for any data centre.
Nvidia is redesigning its hardware strategy. Legacy racks are being phased out in favor of five new Vera Rubin‑specific towers, each built expressly for running AI agents rather than traditional large language model inference.
What does this mean for business right now? Companies must re‑evaluate their IT budgets and allocate a portion to hybrid solutions where GPU platforms operate alongside token farms. Skipping investment in these specialized towers could cost a firm up to $1,000 per million tokens of missed revenue and leave it trailing competitors who are already monetising agent compute.
Why this matters: Ignoring the shift means forfeiting a measurable new income stream while rivals capture it. Allocate budget now to integrate token‑farm capable hardware and secure that incremental profit.