Not All AI Coding Agents and Vibe Coding Platforms Are the Same: A GTM Breakdown of Cursor, Claude Code, OpenAI Codex, Replit, Lovable, Devin, V0 and More

Lovable hit $100M ARR faster than OpenAI. Cognition hit $492M ARR in 12 months through Goldman Sachs and NASA contracts, then raised $1B at a $26B valuation last week. Cursor was featured by Gartner as the leader for AI coding agents and is reportedly being acquired by SpaceX for $60B. GitHub Copilot quietly cleared $1B ARR. 

The term "vibe coding" has generalized a complex market into a single category. But the market, buyers, and ICPs are quite different and can be grouped into 3 main categories:

• Enterprise engineering agents: autonomous or semi-autonomous agents that take on engineering tasks.

• Agentic IDEs for developer teams: agentic IDEs and coding assistants that make developers faster inside their workflow.

• App builders for non-developers: prompt-to-product platforms for non-developers, founders, designers, marketers, PMs, and operators.

When looking at who’s winning GTM, it may seem like a crowded market but they address different jobs-to-be-done in different buyer contexts.

ICP breakdown by category: who actually uses what, and why

Enterprise engineering agents are autonomous or semi-autonomous agents that take on engineering tasks end-to-end.

• JTBD: “Take this engineering task and come back with a working change.”

• Example platforms: Cognition’s Devin, OpenAI Codex, Claude Code in more agentic workflows, and potentially Augment for large enterprise codebases. 

• Buyer/GTM process: Engineering leader. CTO, VP Engineering, Head of Engineering, or platform/DevEx leader. The economic case is tied to throughput.

Agentic IDEs for developer teams make developers faster inside their existing workflow.

• JTBD: “Help me code faster.” 

• Examples: Cursor, Windsurf, GitHub Copilot, Amazon Q Developer, Gemini Code Assist, GitLab Duo, JetBrains AI, and parts of Claude Code and Codex.

• Buyer/GTM process: The buyer starts as an individual developer and converts to a team deal once usage spreads organically.

Cursor may be more loved by AI-native developers. GitHub may be more trusted by enterprise buyers. Amazon Q may win in AWS-heavy environments. Gemini Code Assist may win in Google Cloud-centric organizations. GitLab Duo may win where GitLab is already the development system of record.

AI App builders for non-developers are prompt-to-product platforms

• JTBD: “Turn my idea into a working app or prototype.”

• Examples: Replit, Lovable, Bolt, v0, and similar prompt-to-product platforms.

• Buyer/GTM: Founder, marketer, PM, or designer who wants a working app without writing code. As of 2026, roughly 63% of vibe coding users are non-developers. writers, marketers, investors, and students.

Lovable, Bolt, and v0 each have strong wedges. Lovable is strong for non-technical app creation. Bolt is strong for fast web app generation. v0 is especially strong around UI and frontend generation. These tools can become meaningful businesses if they own repeat workflows for founders, designers, agencies, and semi-technical teams.

But this category has a different economic profile from enterprise engineering agents. The user base may be much larger, but ACVs are lower. For b2b/enterprise, PMs, marketers, ops teams, innovation teams, and agencies inside large companies may use these tools. But that is different from becoming the standard platform for enterprise engineering.

User type, from non-technical to developer-native, mapped against interaction model, from real-time co-creation to delegated agentic execution.

Where the money is, how companies will continue to scale, and who is best positioned

• Enterprise engineering agents have the highest ACV. They sell directly into expensive engineering problems: backlog, modernization, bug fixing, test coverage, and developer productivity. If they work, the budget is obvious. 

• Developer PLG platforms have the strongest seat expansion path. They can spread across teams through developer pull, then convert into enterprise-wide contracts.

• AI app builders have the broadest user expansion. The market of people who want to build software is much larger than the market of professional developers. But currently, the monetization is more likely to look like prosumer, SMB, team, agency, hosting, and usage-based revenue rather than large top-down CTO deals.

The enterprise AI coding agent market is $9.8-11B annualized versus $4.7B for consumer/SMB vibe coding. Enterprise contracts have higher ACV, lower churn, and procurement-driven renewal. Codex's explosive early growth (already at 60% of Cursor's usage within months of launch) shows what bundled distribution does: it converts existing subscribers without requiring a new purchase decision. That's the Microsoft and OpenAI playbook, and it's brutally effective. At large enterprises, IT procurement is overriding developer preference. GitHub Copilot is winning on distribution. Microsoft already owns the enterprise infrastructure layer, so Copilot gets approved before the developer ever makes a choice.

But the consumer PLG counter-case is real. Canva reached $2B ARR largely on individual and small team subscriptions before enterprise motion matured. Figma built to $400M ARR on designer adoption before Adobe acquired it. Notion delayed enterprise for years and survived. The pattern across all three: consumer PLG buys time, brand, and distribution while enterprise contracts buy durability and margin.

GTM opportunities and what could be important moving forward

• The non-developer enterprise buyer: PMs, marketers, designers, and ops professionals at large companies want these tools. The company that can serve these users with a coherent story/capability that addresses enterprise security, compliance, and non-developer ease-of-use simultaneously could create a new enterprise category.

• Security as a brand. As AI generates more code and more non-developers build software, security becomes more than a feature. It becomes a trust signal. Buyers will care about data exposure, code quality, vulnerabilities, permissioning, and who is accountable when something breaks.

• Business outcomes, not just developer metrics. Enterprise coding tools measure productivity: hours saved, PR speed, build cycle time. The platform that builds a business value framework for the economic buyer (time-to-market, engineering cost per shipped feature, revenue attribution from AI-accelerated releases) not just a developer productivity dashboard, can own the C-suite conversation.

So, Who Wins?

There won’t be a single winner here.

GitHub is the safest bet for enterprise SDLC standardization because it already owns the system of record for software collaboration.

OpenAI and Anthropic are likely to capture significant value because frontier model and agent capability will remain strategically important, whether through their own coding products or through the broader platforms that use their models.

Cursor is the strongest independent bet on the AI-native developer environment, but it needs to keep converting developer love into enterprise-grade adoption.

Cognition is the highest-upside pure-play bet on the autonomous software engineer, especially after Windsurf, but it has to prove reliability and workflow trust at scale.

Replit is the strongest bet on non-developer software creation because it has more of the end-to-end build and deployment loop.

Lovable, Bolt, and v0 can become strong segment winners, but prompt-to-prototype alone will get commoditized. The durable value is in what happens after the first generation: iteration, deployment, collaboration, maintenance, and distribution.

The winners will most likely be the companies that own either of these four assets:

• The enterprise system of record

• The frontier model or agent layer

• The developer workflow

• The end-to-end creation and deployment loop