What AI Spend Reveals About What Startups Actually Value

When you want to understand what founders really value, don’t look at their decks. Look at their spend.

The new a16z + Mercury report on AI spending gives one of the clearest pictures yet of how early-stage startups are actually adopting AI and where the real priorities lie.

Below are the most important signals hiding in the numbers

60% of AI spend is going to horizontal apps

Most of the money is still chasing broad, general-purpose tools not niche vertical AI. Founders are betting on utility before specialization, which makes sense for this early stage.The “horizontal wave” (ChatGPT, Notion AI, Perplexity, etc.) is still carrying the ecosystem.

The new “default AI stack” is forming

Every startup seems to be building on the same three pillars: OpenAI, Anthropic, and Perplexity.These players have become the new AWS + Stripe of AI, the invisible backbone for early products.

Creative AI is the quiet category winner

Tools like Canva, ElevenLabs, and Midjourney are turning every employee into a mini marketing department.“Creative AI” isn’t just design; it’s distribution leverage.The ability to generate visual, audio, and written assets on demand is now a startup advantage.

“AI employees” are becoming real

Startups like Crosby and Serval aren’t just automating tasks, they’re packaging automation as roles. Instead of hiring assistants or SDRs, startups are literally hiring software. The “AI employee” model is the most direct challenge to traditional org charts we’ve seen yet.

The first true AI-native wedge: Sales & GTM

Products like Instantly.ai and Clay are proving that outbound sales and GTM are perfect entry points for AI-native startups. It’s fast to measure, cheap to iterate, and directly tied to revenue a rare trifecta.

Workflow is where the moats will form

The categories with the most logos today, creative tools, meetings, customer service, may not be where the biggest companies emerge.The defensible layer will be workflow orchestration, stitching tools together into daily systems.

The compression of the founder journey

A striking pattern: products now go from consumer → prosumer → enterprise in 12–18 months, not years.The feedback loops are tighter, and adoption now often starts with a single user who brings the tool into their team.

The line between tools and coworkers is blurring fast.

The next generation of startups won’t just use AI, they’ll be built around it.