Wall Street still doesn’t understand AI

Can't argue with the title of this SEMAFOR article by Reed Albergotti. Yes, this week (or the past month) was Anthropic's DeepSeek moment on Wall Street. But this time it was felt broadly by the "software ecosystem", not just the hyperscalers planning unprecedented levels of AI-related Capex in 2026. The Semafor article correctly points to the paradox of Anthropic's success and Amazon's stock dive despite the latter building the data centers for the former.

Yes, Wall Street is confused. In its confusion, it is throwing away the baby with the bathwater, lacking a nuanced understanding of software business models. Software is not going away, but the way it is priced and delivered is changing radically. Perhaps even more fundamental is the nature of the software winners: typical SaaS companies need to reinvent themselves or accept being commoditized, but to do so they need to understand where the control plane resides in an ecosystem that is becoming agentified. Raphaelle d'Ornano has been predicting this shift longer than anyone I know: orchestrators of critical workflows residing near the expression of user intent have a chance to win, others not so much. And that is true not just of technology companies but also of pretty much any incumbent in any industry: they risk being disintermediated by faster, cheaper, smarter players that dominate the control plane.

In the age of LLM chatbots, it is important to know which questions to ask.