AI’s Next Winners: Signals From Private Markets

AI’s fastest-growing opportunities are emerging in infrastructure, agents and physical-world applications, but rapid revenue growth does not by itself establish a durable moat.

• More than 100 startups at Morgan Stanley’s Spark conference described exceptionally rapid scaling: one was approaching $4 billion in annual recurring revenue within two years, while another reached $1 billion in under a year.
• Coding and inference tools can scale faster than application software because they are quicker to build and adopt during the current bespoke-development phase; applications require data integration, workflow understanding and operational controls.
• Agentic tools could address $30 trillion to $40 trillion of consumer and enterprise spending, with consumer use cases including shopping and travel and enterprise use cases emerging in legal, healthcare and other verticals.
• Muse is identified as an early consumer-agent leader, while broader consumer competition is expected to increase in 2027; enterprise agents are developing both vertically and across functions.
• The conference reinforced a distinction between an incumbent’s moat and the speed of its AI transition: one legal AI company benchmarks 18 models across 200 criteria, illustrating how value can shift toward orchestration, workflow integration and model selection rather than the model itself.
• AI’s physical applications in defense, warehouses, delivery and agriculture may offer more defensible advantages as software becomes easier to build, with significant adoption expected over the next two, three and five years.

Takeaway: Investors should distinguish fast AI revenue growth from durable competitive advantage and look for companies that own customer workflows, data, integration or physical operations; falling inference costs could also create an upside surprise for software gross margins if vendors retain part of their pricing.