The “Magnificent Seven” framework is getting tired — a new lens is more useful. Five large AI CapEx spenders (, , , , and ) and roughly 40 AI supplier stocks that are the direct beneficiaries of that spending.
The average software business faces a tougher road ahead. AI data centres are increasingly functioning as “token factories,” and tokens — the outputs of large language models — are now an input cost that software companies must purchase from the likes of OpenAI, Anthropic, and Google. That spells higher costs and new competitive pressure from AI-native companies for the average software business.
Not all software stocks are the same — the three-bucket framework matters. It is worth to sort software companies into three groups: those that need to sell tokens (more cautious); those that likely don’t (mission critical or less sophisticated end markets, and currently being unfairly pulled down); and those that need to sell tokens but are providing the scaffolding for the agent infrastructure expected to proliferate in 2027 and 2028 – the group owned most today.
Mid-cap growth is a compelling segment of the U.S. equity market right now. With enterprise values roughly between US$5 billion and US$60 billion, these companies are well-positioned to benefit from ongoing hyperscaler CapEx, harness AI to expand margins and accelerate revenue, and take market share from smaller competitors who lack the culture and resources to implement this technology as effectively.
