SpaceX paid $60 billion to acquire Anysphere, the company behind the AI coding tool Cursor, in June 2026. This transaction is the largest venture-backed startup acquisition on record. Vertical AI companies, such as those in legal, healthcare, construction, and insurance, compete for the 13% of US GDP spent on business labor. Horizontal AI companies, such as those building coding tools, agent infrastructure, and voice automation, compete to become the default infrastructure every vertical product depends on. This distinction remains clear because vertical AI targets expensive professional work, while horizontal AI targets the software layer. A legal AI copilot does not compete against a competitor’s software line item; instead, it competes against the cost of an associate’s billable hour.
The coding and legal dominance
Software development is the largest area of concentration for agentic capital. Cognition, the company behind the coding agent Devin, closed a Series D above $1 billion at a $26 billion valuation on the back of revenue that grew from $37 million to $492 million in just twelve months. Another player, Anysphere, maintains a $29.3 billion valuation. In the legal sector, companies like Harvey and Legora each closed rounds of at least $150 million in 2026. Legal AI captured $1.0 billion, or 33.5% of total vertical AI dollars this year.
| Company | Category | Total Funding | Valuation |
|---|---|---|---|
| Cognition | Software Engineering | $3.9 billion | $48 billion |
| Sierra | Customer Service | $635 million | $10 billion |
| Harvey | Legal | $600 million | $5 billion |
| Replit | Software Engineering | $650 million | $9 billion |
Customer service is the third major cluster. Sierra leads with a $10 billion valuation and $100 million in annualized revenue. The companies in this space provide task execution for industries like finance and healthcare. In property technology, EliseAI reached a $2.2 billion valuation for its AI leasing services. Bedrock Robotics also reached a $1.75 billion valuation for autonomous construction vehicles. Healthcare AI investment moved into drug discovery with Atlas Discovery from the YC Summer 2026 batch. In the healthcare sector, the YC Winter 2026 batch included Patientdesk for scheduling and Zatanna for administrative processes. Overdrive works on medical coding and claims automation. In construction, Field AI raised $405 million for construction robotics.
The production gap and orchestration frameworks
The shift from 2025 to 2026 moved AI from the chat window to the delegation of tasks. Users define objectives, and agents figure out the execution. Frameworks like LangGraph, CrewAI, and the OpenAI Agents SDK handle the plumbing, such as prompt orchestration and tool integration. While LangGraph provides stateful multi-agent orchestration, CrewAI uses a role-based model for its multi-agent tasks. However, building an agent is easy, but shipping one reliably into production is hard. Most multi-agent AI systems fail in production because they lack the necessary governance, testing, and oversight. Getting from 80% accuracy to 99% accuracy requires 100x more work than the initial development phase, which remains a barrier for many startups. I see companies struggle with agent drift, where model providers change configurations and cause output behavior to shift unexpectedly. Research from METR shows the task horizon for frontier agents doubles about every seven months. Will the move to autonomous software engineering completely erase the need for traditional junior developers?
