AI Software

Sector thesis

AI Software is the layer of programs and tools that let companies actually *use* AI to solve real problems—think chatbots, recommendation engines, fraud detection, or design tools. It's different from AI compute (the chips and data centers that power it), which we track separately. Right now, the sector is riding a genuine megatrend: every industry from healthcare to finance to manufacturing is trying to embed AI into their workflows to cut costs and move faster. Companies that can build or sell software that makes this easy are in high demand. Within AI Software, there are three main buckets. First: **foundation models and platforms**—companies building the underlying AI systems that others build on top of. Second: **vertical software**—tools designed for specific industries, like AI for legal document review or medical imaging. Third: **infrastructure software**—the plumbing that helps companies deploy and manage AI safely, like monitoring tools or data pipelines. The biggest risks are real. AI software is still early and unproven in many use cases—companies might spend millions on a tool that doesn't actually save them money. Competition is fierce and global; a startup in any country can compete with an established player. Regulation is also a wildcard; governments are still figuring out how to oversee AI, and new rules could change the economics overnight. Finally, the sector is crowded with hype, which means some companies are valued on hope rather than actual revenue. For a retail portfolio, this sector works best as a smaller position—maybe 5-10% of a growth allocation. Watch for companies with *real customers paying real money*, not just pilots or demos. Look for recurring revenue (subscriptions, not one-time sales) and expanding margins over time. The winners will be those solving specific, painful problems, not chasing every trend.