On Thursday, Andreesen Horowitz released its first AI spending report in collaboration with FinTech Farm Mercury. Using transactional data from Mercury, the report analyzes the top 50 AI-Native application layer companies that startups spend, similar to the top 100 Gen AI Consumer Apps previously published.
A16Z partners Olivia Moore and Seema Amble say the data shows that companies employ a variety of AI products for specific tasks. And new apps are rising very rapidly.
“The tools are surged,” Amble said. “We didn’t combine one or two around each category.”
The report also suggests that there are a lot of spending on “human augerment” or “copilots”, which will help employees become more productive, and that startups are not ready to move fully into agent workflows.
“I think there will be a shift as computer usage becomes more modest and the ability to build end-to-end agent flows increases. “Especially because people really want to give it a try.”
Naturally, the top of the list was dominated by major labs, with Openai following up top slots and humanity in number 2. The vibe coding tool was also well represented.
When A16Z created a similar list of consumer habits, many people used it to create projects, so it ranks far higher than replicas with pure traffic alone. However, startups aren’t spending as much money as lovely as replicas due to lack of enterprise features. However, the various companies on the list seem to suggest that there is room for many different companies at once.
TechCrunch Events
San Francisco
|
October 27th-29th, 2025
“This is an upcoming unresolved question with atmospheric coding,” Moore said. “Does spaces begin consolidation and one place become the best platform for the atmosphere in code? Or if there are four or five more atmosphere coding companies that are very big business in different types of applications? There’s no answer yet.”
Moore was also surprised to see startups adopting consumer tools such as Capcut and Midjourney.
“We see that a lot of these (consumer) companies are creating fun consumer tools that people employ and use as individuals and bring to their teams and workplaces, making them faster and faster,” Moore said.
The horizontal application overall accounted for at least 60% of the list’s names, with 40% being made up of vertical applications. The most popular vertical software companies were categorized into three buckets: sales, recruitment and customer service. However, the report also found that AI is advancing in many sectors where previous generations of startups struggled to crack.
“Perhaps before, it was like service companies and consultants were software companies in the age of AI,” Moore said.
Amble gave Crosby Regal as an example. This quickly reviewed the user’s legal agreements and replaced what was at one point in time meetings with the internal general counsel, preventing ideas and research. She said now that most tools are used to help employees (e.g. copilots) make them faster, rather than replacing the suits of employees and talent with autoworkers (end-to-end agents).
“Because the technology gets better and you can actually build coworkers for the agent, you can see that the mix is moving more towards the end-to-end agent, away from the co-pilot,” she said.
There were also many note trustees on the list, including Otter.ai, reading Ai, Happyscribe and more. This is what Amble meant when she said there were no products that still dominate the market. Rather, startups are still choosing their “flavors” to see what tools they like the best. This is also suitable for employees who can choose what application will help them work best, rather than using a single size product “pushing down from above” using a single size product with so many options.
The final big discovery of the report was the increased entanglement between consumers and businesses. People bring personal applications to work from home, and people who start businesses are helping to build their businesses using their favorite personal applications. Previously, there was a portrayal between the two. A set stack of things to use when building a startup.
Amble and Moore cited Canva as an example. This is a popular consumer app with a considerable enterprise audience. It took Canva years to add an enterprise plan. However, as individual and corporate use cases become more difficult to distinguish, companies try to blend the two together.
“Your TAM (Total Addressable Market) is no longer either, but you can sell it to both,” she continued. She also means that companies building these products will also build enterprise teams, such as to market, sales, support, and more, so that they can start selling and collecting enterprise revenues faster than relying on individual consumers.
Moore and Amble hope that this list will change rapidly in the coming years. Older businesses are now launching relevant and accessible AI capabilities, but new entrants are arriving with new ideas.
“Legacy player, Legacy means almost “12 months ago,” Amble said. “If I pull this again in 12 months, is there even an app out there that will make the same notes? Or is there an entirely new set?”
This work has been updated to reflect the note taking app.
