OpenAI ranked first, Anthropic second, and Replit third in Andreessen Horowitz’s ranking of AI application companies that startups paid through Mercury accounts from June through August 2025. But the more important finding is what sits beneath those leaders: startup spending also extends across creative software, coding tools, meeting assistants, recruiting, customer service, sales, compliance, and other specialized workflows.
The report, published on October 2, 2025, is a useful snapshot of where startup dollars were moving—but it is not a complete ranking of all AI revenue, usage, or market share.
As an Amazon Associate I earn from qualifying purchases.
The short answer
a16z’s AI Application Spending Report analyzed transactions from more than 200,000 Mercury customers. It ranked the top 50 AI-native application companies by startup spending routed through Mercury between June and August 2025.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsThe top three were:
- OpenAI
- Anthropic
- Replit
That does not mean OpenAI has captured all startup AI spending, or that Replit has the third-largest AI business overall. It means those companies received the most spending observed in this particular dataset and period.
#1 Best Overall
What a16z actually measured
The report measured payments made through Mercury, including:
- ACH transactions
- Mercury IO card spending
- Wire transfers
This is different from measuring website traffic, free-trial activity, downloads, signups, model usage, or total vendor revenue. It is also not a full accounting of startup software budgets.
The sample excludes spending made through other banks, non-Mercury cards, procurement platforms, employee cards, and other payment channels. Mercury Personal customers were excluded. The customer base may also lean toward venture-backed, digitally native, and U.S.-oriented startups.
The ranking focuses on application-layer companies. Primarily infrastructure-focused cloud providers, GPU vendors, and similar tools were excluded. Google is a notable qualification: the report combined Google Cloud and Gemini spending because the data could not separate them.
How to read the ranking: treat it as a spending signal among Mercury customers—not as a universal market-share table, product-quality ranking, or return-on-investment study.
Rank #2
The top 10 companies in the June–August 2025 data
| Rank | Company | Category | What the ranking suggests |
|---|---|---|---|
| 1 | OpenAI | General assistants and models | Broad AI capabilities remain a foundational purchase. |
| 2 | Anthropic | General assistants and models | Startup demand is not concentrated in just one frontier-model vendor. |
| 3 | Replit | AI app development | AI coding and application building have become workplace spending categories. |
| 4 | Freepik | Creative tools | Creative production is one of the strongest areas of AI adoption. |
| 5 | ElevenLabs | Voice and audio | Production-ready voice generation is attracting business budgets. |
| 6 | Cursor | AI coding | Developer-focused assistants are being purchased by teams. |
| 7 | Fyxer | Meeting and email assistance | Startups are paying for tools built around specific daily workflows. |
| 8 | Lorikeet | Customer service | Specialized support automation is gaining traction. |
| 9 | Micro1 | Recruiting and HR | AI spending is reaching talent and staffing functions. |
| 10 | Notion | AI workspace and productivity | AI is increasingly embedded in broader work platforms. |
Other companies highlighted in the report include Delve at number 11, Perplexity at number 12, Instantly at number 13, Customer.io at number 14, Clay at number 25, Cluely at number 26, Midjourney at number 28, Manus at number 33, Cognition at number 34, Otter AI at number 41, and Alma at number 42.
The spending is broader than chatbots
The report’s headline leaders are general-purpose AI providers, but the wider list shows startups are building layered software stacks rather than buying a single universal AI product.
General assistants and workspaces
OpenAI, Anthropic, Perplexity, Merlin AI, Notion, and Manus all appear in the general-assistant or workspace discussion. The presence of several strong products suggests the market had not settled on one winner by mid-2025. Teams may use different products for research, writing, coding, internal knowledge, and agentic tasks.
Creative software
Creative tools were the largest individual category identified by a16z. The list includes Freepik, ElevenLabs, Canva, Photoroom, Midjourney, Descript, Opus Clip, CapCut, Arcads, and Tavus.
This category also illustrates the overlap between consumer and workplace software. An individual creator may start with a personal subscription, then bring the tool into a marketing, product, or social-media team. Nearly 70% of the listed companies could be adopted by individuals and then brought into teams, according to a16z.
Coding and application building
Replit’s third-place finish was one of the report’s most notable findings. It ranked ahead of Cursor, Lovable, and Emergent, placing AI application development among the most visible forms of startup software spending.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →a16z attributes Replit’s relative strength against Lovable partly to its broader scope, including application development, agents, databases, authentication, publishing, and enterprise controls. Within the examined Mercury sample, a16z reported that Replit generated approximately 15 times more revenue than Lovable. That comparison applies only to the sample; it should not be generalized to the companies’ total global revenue.
The contrast also shows why popularity and willingness to pay are different signals. TechCrunch noted that Lovable ranked higher than Replit in a consumer-oriented ranking based on web traffic, while Replit ranked higher in this spending-based startup list.
Meeting and communication tools
Fyxer, HappyScribe, Plaude, Otter AI, Read AI, and Cluely represent a fragmented meeting and communication market. No single meeting-notetaking product dominated the list, suggesting that startups were trying multiple approaches rather than standardizing around one tool.
Business-function software
The report also includes specialized tools for:
- Customer service: Lorikeet, Customer.io, Ada, and Crisp
- Sales and go-to-market: Instantly, Clay, and 11x
- Recruiting and HR: Micro1, Metaview, and Applaud
- Operations: Delve and Combinely
- Legal and compliance: Crosby Legal and Delve
- Engineering and IT: Cognition and Serval
Replit versus Lovable: why payment data matters
The Replit-Lovable comparison is a useful case study in how different datasets answer different questions.
Free tools Windows power users keep installed
One-click scans. No signup required.
Web traffic can indicate attention, discovery, or consumer interest. Payment data indicates that money moved from a particular group of customers to a vendor. Neither measure alone proves product quality, retention, profitability, or long-term market leadership.
Lovable may have had stronger consumer visibility, while Replit generated more spending among the Mercury startups analyzed by a16z. A buyer deciding which product to use should therefore examine the actual workflow, deployment requirements, technical control, security needs, and cost—not simply traffic rankings or a venture report’s position.
Are startups replacing workers with AI agents?
Not at scale according to this dataset. a16z classified 17 companies as vertical applications. Twelve primarily augmented human workers, while five were designed to complete workflows end to end.
The more autonomous group included Crosby Legal, Cognition, 11x, Serval, and Alma. The broader pattern, however, was still one of human augmentation: helping people draft, search, summarize, qualify, code, respond, or make decisions faster.
That distinction matters. The report supports the conclusion that startups are buying copilots and workflow tools. It does not prove that companies are broadly replacing employees with autonomous digital workers. Greater autonomy may become more common as computer-use capabilities improve, but that is a forecast rather than a finding established by this three-month spending snapshot.
Best Value
What this means for startup buyers
The ranking is most useful as a map of categories worth evaluating—not as a shopping list. Before adding an AI subscription, buyers should ask:
- Does it solve a recurring workflow? A tool used weekly by a team is more valuable than one that produces impressive but occasional demos.
- How is it billed? Compare per-seat, usage-based, API, credit, and enterprise pricing. A low subscription price may conceal substantial consumption costs.
- Who controls the account? Confirm company ownership, administrator access, SSO, permissions, audit logs, and offboarding procedures.
- What happens to company data? Review retention, training use, data residency, recording policies, and deletion controls.
- Does it integrate with the existing stack? Overlapping meeting, creative, sales, or assistant tools can create duplicate costs and fragmented data.
- Can the company export its work? Check portability of documents, prompts, code, customer records, transcripts, and generated assets.
- Where is human review required? Legal, recruiting, customer service, sales outreach, and compliance workflows need clear approval controls.
- Can savings be measured? Track time saved, response speed, conversion, output volume, error rates, or revenue—not just the number of active users.
Products such as OpenAI and Anthropic may serve as foundational tools, while Replit, Cursor, Lovable, ElevenLabs, Canva, Clay, or Otter address narrower jobs. They are not interchangeable, and stacking every highly ranked product can create unnecessary software spend.
What the report cannot prove
- It is not a ranking of all startups. It covers Mercury customers and transactions routed through Mercury.
- It is not total vendor revenue. Companies receive payments through many channels.
- It is not usage data. A payment does not show how often a product was used or whether it delivered value.
- It is not product validation. Rank does not establish quality, security, retention, profitability, or return on investment.
- It is not a current leaderboard. The data covers June through August 2025 and was published October 2, 2025. Pricing, products, and adoption can change quickly.
- It may contain overlapping layers of spending. A startup can pay OpenAI or Anthropic for models while also paying Replit, Cursor, Notion, or another application that uses AI.
- It does not always identify a pure AI vendor. The list includes established products that have incorporated AI capabilities.
The larger takeaway
The report does not show that one AI company has won the startup market. It shows that startup AI spending was already spreading across multiple layers: general models, creative production, coding, communication, sales, recruiting, customer service, compliance, and operations.
Recommended Free Tools
OpenAI and Anthropic led the measured application spending, but the commercial story is broader. Startups were paying for tools that fit specific work, often through bottoms-up adoption that began with individuals and moved into teams. For buyers, the practical lesson is to evaluate AI as part of a layered software stack—and to measure the business result of each layer before adding another subscription.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
