Descriptions:
Arun Mathew, partner at Accel, joins Bloomberg Technology to discuss where AI investment is heading — and where the risks are forming. Mathew argues that while bubble tendencies are emerging in parts of the AI market, the application layer remains fundamentally undervalued because most investors are still thinking about AI as a productivity enhancement to existing software rather than a wholesale reinvention of work. His example: legal software is a $5–10 billion global revenue market, but legal labor spend is roughly $1 trillion — meaning even a 10% shift to agentic work represents a $100 billion opportunity.
Mathew names specific Accel portfolio companies operating in this vertical space — Cursor in coding, Basis and Ligura in legal and finance, Campfire in enterprise — and addresses the direct challenge that foundation model providers like Anthropic (also an Accel investment) could commoditize application-layer companies. His counterargument centers on last-mile complexity: security, compliance, change management, and accuracy requirements in regulated industries create durable defensibility that general-purpose models cannot easily absorb.
The conversation turns to NVIDIA’s NemoClore open-source agent platform, announced at GTC, which Mathew frames as an awakening moment for the industry rather than a threat — evidence that the agent paradigm is becoming mainstream infrastructure. He closes by predicting a market where every individual eventually has a personal set of agents for professional and personal use, with fewer but much larger winners than today’s crowded field suggests.
📺 Source: Bloomberg Technology · Published March 20, 2026
🏷️ Format: Interview
