Google Cloud's VP for startups on reading your "check engine light" before it's too late

What It Covers

Darren Morey, Google Cloud's VP of Global Startups, outlines how Google competes for AI startups through credits, engineering resources, and enterprise distribution pipelines, while identifying structural warning signs — LLM wrappers and model aggregators — that predict which startups will fail to generate durable cloud revenue.

Key Questions Answered

• •Startup Credit Programs: Google Cloud for Startups offers tiered credit tranches scaled to funding stage, but Morey identifies engineering access — not credits — as the primary differentiator. Startups receive dedicated technology specialists who monitor both architecture decisions and credit burn rates simultaneously, reducing cost surprises that previously caused founders to exhaust runway before revenue materialized.
• •"Check Engine Light" Framework: Founders building thin intellectual property layers around foundation models (LLM wrappers) or routing users between multiple models without added intelligence (aggregators) show consistently low growth and retention. Morey uses these two patterns as early indicators that a startup lacks the horizontal or vertical differentiation required to survive commoditization of underlying models.
• •Infrastructure Cost Management: As AI workloads shift from GPU/TPU compute toward model APIs and agentic platforms, startup economics change substantially. Gemini API consumption costs significantly less than traditional cloud compute, meaning founders who move architecture decisions up the stack — from chips toward agents and data — can extend runway without switching providers or renegotiating credits.
• •Enterprise Distribution via Gemini Enterprise: Google routes startups directly into its Gemini Enterprise marketplace, giving founders access to large enterprise customers — Walmart, Wells Fargo, Verizon scale — as a distribution channel. Startups build agents on Google Cloud, list them on the platform, and convert enterprise usage into revenue without building independent sales infrastructure or enterprise procurement relationships.
• •High-Growth Vertical Signals: Morey tracks three sectors showing measurable retention and consumption growth: biotech and digital health (using AlphaFold and DeepMind models for previously impossible research), climate tech (blending large datasets in novel configurations), and developer/vibe-coding platforms like Cursor, Lovable, and Replit, which consume cloud resources disproportionate to their employee headcount.