SERV Expands to 44 Cities, But Scale Profitability Remains Elusive

What happened

Serve Robotics announced expansion into 44 U.S. cities and is exploring Canadian markets, signaling aggressive geographic scaling. The DeepValue report notes that such expansion risks amplifying losses, as Q3'25 gross loss was $(4.38)M on only $0.69M revenue. While fleet deployment headlines boost the narrative, filings show that scaling has so far increased cost intensity rather than improving unit economics. The company continues to rely on equity financing, having raised ~$100M in October 2025, and any new markets will add launch costs before generating revenue. Until reported metrics demonstrate gross loss compression and multi-partner utilization, the stock remains a speculative wait-and-see.

Implication

The news of expansion into 44 cities and potential Canadian entry reinforces the aggressive scaling narrative, but the DeepValue report highlights that scale to date has been loss-amplifying, with Q3'25 revenue of $0.69M against cost of revenues of $5.07M. Investors should remain cautious: the company's cash burn and reliance on equity dilution (recent $100M offering at $16) mean that each new market adds upfront costs without guaranteed paid utilization. The key catalyst remains the upcoming Diligent acquisition pro forma and sequential Q4 2025/Q1 2026 results showing whether gross margins improve. Without a clear path to gross profit, the stock's current valuation near $9.39 offers no margin of safety, and further dilution is likely if losses persist. Patience is warranted until the company can demonstrate that scaling reduces unit costs rather than expanding losses.

Thesis delta

The news does not change the core thesis; it amplifies the existing tension between fleet growth and negative unit economics. The DeepValue report's WAIT rating remains appropriate as expansion increases both potential future revenue and near-term cash burn. Investors should continue to monitor for gross loss compression and multi-partner utilization rather than react to deployment counts.