Discussion of AI Tokens On Latest Earnings Calls

Running AI is getting cheaper. From the latest round of earnings calls and 3,265 recent media mentions we know that:

• 22 of 26 AI hardware companies have discussed tokens or inference on their latest earnings call.

• Executives raised the topic 126 times. Analysts asked about it 7 times.

• Executives were noticeably more upbeat in scripted remarks than under live questioning: +0.77 vs +0.43 on a -1 to +1 scale.

HPE's CEO reportedly told analysts that AI inference will accelerate demand for its servers.

In analyzing 521 recent news stories and social posts about AI costs, the positive sentiment around TSMC, Intel, and NVIDIA has been improving, while the positive sentiment around Micron and Dell worsens.

The pattern that appears to be emerging is that cheaper AI is framed as more business for the companies driving costs down, and as price pressure on memory and standard-server sellers.

Micron is the one to watch. Memory sentiment is rising while Micron's positive sentiment in cost discussions is falling. 

Data Source: MarketPsych MCP for LSEG MarketPsych Analytics, News and Social Media, April to June 2026.