Do We Get Points for Mentioning AI On Our Earnings Call?
Even if we are cutting heads and sales are soft?
This earnings season when CEOs and CFOs tell you they are cutting heads because their companies are enjoying the fruits of “AI” in the form of operating leverage, they are most likely full of the stuff that Bess Truman tried to get Harry Truman to call “fertilizer” for 53 years. They are blowing smoke. They are trying to distract from weakness with a shiny new object.
If you are leveraging language models to help review code, I believe there is an ROI there.
If you are suggesting to the Street that you have replaced core product engineers with AI agents, you’re full of it.
If you are using language models to help automate customer service calls via chat bots, I believe you.
If you are a Consulting firm or I-Bank using language models to perform the initial phase of a research project, I believe you.
I don’t believe you if you tell me you plan to replace Consultants and I-Bankers with language models (LLMs severely hallucinate). If you are not careful and allow models to have the final say in the reports you generate, you will end up like Deloitte and owe your client a mea culpa and lots of money.
If you back out Microsoft and Cursor (two front-end partners to Claude), Revenue from Anthropic’s $5 Billion 2025 Revenue run rate, the $5 Billion is more like $3.5-4.0 Billion. If you back out Microsoft and Cursor Revenue from OpenAI’s numbers, OpenAI’s 2025 Revenue may be closer to $8-10 Billion, rather than its expected $13 Billion (ex-M&A). So, $11-12 Billion of non-partner Revenue between two of the leading LLM companies justifies a combined valuation of $683 Billion? I don’t see it. Feels like a huge short in the making.
Other than companies that have captured ROI on the Software coding side or CSR side, which companies are generating significant ROI from Gen AI? Not from Software automations. Not from Robotic Process Automation. Not from other forms of Advanced Automation, but specifically from Gen AI?
NVIDIA, OpenAI, Oracle and Broadcom can announce all of the sham, non-binding deals they want. At the end of the day, end-users need to use LLMs to support the more than lofty valuations that the LLM companies sport. Partner revenue, datacenter builds and press releases are all great, but the proof is in the pudding. The pudding is end-user usage.
By the way, OpenAI usage is down.