I have been watching the AI coding market for the past two years, and the funding numbers that landed in the first half of 2026 tell you something that is easy to underestimate if you are not living in the details.

Cursor, the AI-native code editor built by Anysphere, is in talks to raise at a $50 billion valuation. It already crossed $2 billion in annualized revenue and is forecasting $6 billion by the end of the year. Cognition, which builds Devin — an autonomous AI coding agent — just raised $1 billion at a $26 billion valuation, bringing its total funding past $2.5 billion. Anthropic’s Claude Code, which launched 13 months ago, is generating $2.5 billion in annualized revenue and doubled its weekly active users in a matter of weeks.

These are not numbers that belong to an experimental category. This is a market that has found product-market fit at extraordinary scale.

What I want to talk about is not the funding. It is what these numbers mean for businesses that are not software companies but that depend on software getting built, maintained, and improved.

The junior developer is not the future anymore

Here is the shift that the funding numbers are pointing at.

For decades, the way software teams scaled was linear. You hired junior developers, they handled the routine work, they learned, they progressed, and eventually they took on more complex tasks. The economics worked because junior developer time was cheap and the volume of routine work was high.

AI coding tools do not replace senior engineers with complex judgment, the people who understand the architecture, the edge cases, the business context. What they replace is the routine, well-defined, lower-order work that junior developers were hired to handle.

Uber’s engineering team gives you a concrete picture. By the time their 5,000 engineers had been using Claude Code for a few months, 70 percent of committed code was AI-generated. Eleven percent of live backend updates were being written entirely by AI agents with no human review at any point. The company burned through its entire 2026 AI budget in four months because the tools were being used that intensively.

That is not a company experimenting with AI coding. That is a company that has structurally changed how software gets built.

What this means for businesses that hire developers

Most of my conversations with business leaders about this topic start in the wrong place. They ask whether they should be using AI coding tools. That question is already answered. The more important question is what the shift in software development economics means for how they should be building their own technology.

The first implication is that custom software development is becoming significantly more accessible than it was. If AI tools are compressing the time to write, test, and iterate on code by 30 to 70 percent — and that is the range I am seeing across real deployments — then the cost of building custom internal tools, automation systems, and client-facing applications is falling significantly.

Businesses that could not previously justify the cost of custom software are now in a position where the economics work. That changes the build-versus-buy calculation for a lot of internal tools that have been sitting on a wishlist for years.

The second implication is that the quality of software development vendors is diverging faster than it used to. The agencies and development teams that have adopted AI coding tools well are shipping faster and at lower cost than those that have not. If you are procuring software development services and you have not asked your vendors about their AI tooling adoption, you are probably overpaying.

The third implication is about your own team. If you have internal developers or a technical team, the tools they are using right now are shaping how much they can produce. A developer using Claude Code or Cursor is not just faster. They can take on more complex work because the routine implementation work is off their plate. That changes what you can build with the same headcount.

The autonomous agent question is next

Cursor, at its current valuation, is primarily selling AI-assisted development. The developer is still in the loop, still making decisions, still directing the work.

Cognition’s Devin is selling something different: a system where you describe what you want and the agent builds it, tests it, fixes errors, and delivers output without you in the loop for each step.

The gap between those two things is significant. One augments a developer. The other replaces the developer for a defined category of tasks.

Devin’s current reliability on genuinely complex real-world tasks is still building. But the trajectory is clear. Every six months, the class of tasks that can be delegated fully to an autonomous coding agent expands. The $1 billion raised by Cognition is an investment in accelerating that expansion.

For businesses thinking about custom software development over a two-to-three-year horizon, this matters. The question is not just what can AI-assisted development do today. It is what autonomous coding agents will be able to handle in 2027, and whether your organisation is positioned to use it.

What I am telling business owners right now

When I work with businesses on their technology strategy through Omni Advisory, the questions I am pushing them to answer on this are practical.

What internal software development have you been avoiding because the cost-to-value ratio did not work? That calculation has changed. Things that were not worth building 18 months ago might be worth building now.

Who are you hiring for, and what will those roles look like in two years? If you are planning to grow a technical team, the mix of skills and seniority that makes sense is different than it was three years ago.

What does your current development vendor’s AI tooling look like? If they are not using AI coding tools extensively, you are subsidising their slower approach.

Are you investing in the data skills and technical literacy that let your team evaluate and direct AI-built software? The biggest risk is not that AI coding tools produce bad output. It is that teams without the skills to review the output let bad output get deployed.

The compounding advantage is already starting

Businesses that are using AI coding tools well today are not just building faster. They are generating more data about what works, building more internal tools, creating more automation, and accumulating operational advantages that compound over time.

The companies that are going to be hardest to compete with in three years are the ones that are already running fast right now. Not because they are doing something exotic, but because they are deploying the available tools aggressively while their competitors are still debating whether to start.

The $50 billion valuation on Cursor and the $1 billion raise for Cognition are the market’s way of saying it believes AI coding tools are becoming infrastructure. Infrastructure is not something you opt into eventually. It is something you either run on or fall behind by not running on.

The junior developer question is a proxy for a broader question about where your business is in this transition. The organisations I work with that are moving fast have stopped asking whether AI coding tools change the economics. They know they do. They are focused on capturing as much of that change as possible before the window is most open.

---

I work with business leaders through Omni Advisory to build technology strategies that account for where AI is now and where it is going. Book a session if you want to think through what this shift means for your specific situation.