CarGurus (CARG)

The sleeper story is the buyback: CarGurus went from $175M deployed in Q1 to just $29M in Q2 with cash building to $122M, a far bigger behavioral shift than the headline beat-and-raise suggests, and management offered zero explanation for it. The obvious headline, a revenue beat and raised EBITDA margin guide, is real but is being built on decelerating top-line growth (13% vs 15%) and accelerating opex (16% vs 13% YoY), with the margin improvement coming from unquantified AI-driven cost cuts rather than demand strength. The binding constraint right now isn't product or AI capability, it's dealer willingness to spend incrementally, which management insists is 'temporary, not structural' for the second straight quarter without new evidence to support that claim. The most revealing moment was the flat denial on AI monetization: 'it's not about monetizing an AI search result separately from how we monetize other user activity on our site', a retreat from last quarter's promise to share more detail on agentic use cases. My view changes if next quarter's buyback reaccelerates meaningfully above $29M, dealer net adds snap back toward the 900+ range seen in Q1, or management puts a real dollar figure on the AI-driven opex savings they're using to justify the margin guide raise.