Steve Eisman Says AI Giants Have 'No Moats.' Morningstar's $57 Billion Math Says Otherwise

Key takeaways

• AI labs’ cost advantage hinges on bulk GPU deals, but falling market share signals rising competition that could erode margins and impact hyperscaler revenues.
• $57 billion annual cost advantage
• $2.25 per GPU‑hour vs $16.11 list price
• $9 billion annual cost for a gigawatt vs $66 billion

What happened

Steve Eisman warned that AI giants OpenAI and Anthropic may lack sustainable competitive advantages because their high computing costs could be undercut by cheaper rivals, potentially squeezing margins. Morningstar counters that bulk purchasing of GPU capacity gives these labs a $57 billion annual cost edge, citing Nvidia B200 contracts at $2.25 per GPU‑hour versus a $16.11 list price. Vercel data shows Anthropic’s share of AI spending fell from 64% to 44% in September, while overall AI model…

Why it matters

The debate centers on whether AI labs can sustain high spending on compute and still command premium pricing. If cheaper models erode demand, the labs’ ability to pay for expensive GPU capacity could falter, affecting their profitability and the revenue streams of their hyperscaler partners. Morningstar’s $57 billion…

Short-term and longer-term

Short-term. Over the next few days, watch for Amazon and Microsoft earnings releases for any updates on their AI revenue streams and cloud usage.

Longer-term. In the coming months, investors should monitor how AI labs balance compute costs against pricing power.