It seems that the “AI-and-RAN” pitch, which posits that MNOs can monetize excess compute capacity at their RAN sites by selling inferencing tokens for very low-latency applications, may have gone through the wringer just enough that the industry has arrived at a more reasonable sales pitch. At Nvidia’s most recent annual gathering, GTC 26, the chip giant pitched its version of the AI Grid, which is positioned as the next step in the AI infrastructure buildout, after the AI data center, or factory. It provides an opportunity for MNOs and networking firms to provide AI services, but largely sidesteps the AI-RAN idea. “While AI factories are optimized for manufacturing intelligence centrally, an AI grid extends its reach by distributing that…