Salesforce VP on the leaky AI pipeline: why cheaper tokens won’t fix enterprise AI

The real shift is budget authority moving away from model consumption and toward the control plane of AI. That favors vendors that sit on data quality, permissions, lineage, and identity resolution, because the fastest way to lower unit cost is to prevent bad context from entering the model in the first place. In that framework, INFA is the cleanest beneficiary; CRM is a secondary beneficiary only if its data layer becomes the default context layer, not just another app-layer upsell.

The main loser is the broad universe of usage-based AI vendors and any hyperscaler/model stack that has to defend growth with lower per-token pricing. Price cuts can preserve share, but they also reset customer expectations: once finance teams see token leakage, they will demand evidence of governance before approving more volume. That should compress multiples for names whose AI story depends on ever-rising consumption rather than workflow-level efficiency.

Near term, this is mostly a procurement narrative, not an earnings event. The 1-3 month catalyst path is management commentary on AI budget discipline and attach rates for data-governance products; the 6-18 month path is architectural standardization as AI moves from pilots to production. The contrarian point: cheaper tokens are not automatically bearish for GOOGL — if lower unit prices expand total workload and cloud attach, volume can offset ASP pressure. The thesis is falsified if enterprise AI spend keeps rising despite cost scrutiny, or if CRM shows Data Cloud monetization inflecting faster than expected.