TorchTPU is now open source: Run native PyTorch on TPUs
cloud.google.com·6d ago
Cloud & Infrastructure
The New Stack·14h ago·5 min read
Most of Anthropic’s revenue comes from models that barely existed six months ago
TL;DR
Anthropic's recent S-1 filing reveals that over 80% of its third-quarter revenue came from models released just six months prior. This rapid adoption raises questions about whether customers truly need these new capabilities or are simply drawn to the latest technology.
✦ Why It Matters
Evaluate whether your projects truly benefit from the latest AI models before investing.
Key Takeaways
- Assess if newer models are necessary for your specific tasks.
- Monitor spending patterns on AI models to optimize costs.
- Consider the capabilities of older models before upgrading.
- Stay informed about Anthropic's model release schedule for strategic planning.
Full Summary
Anthropic's S-1 filing indicates that its latest models, released between April and September 2026, accounted for more than 80% of its third-quarter revenue, totaling $16.8 billion. The company launched eight Claude models and three restricted-access Mythos models during this period.
Customer spending appears closely linked to the frequency of new model releases, suggesting that improvements in model capabilities drive increased usage and revenue. However, the filing does not clarify whether customer demand influenced the accelerated release schedule.
Notably, heavy enterprise users of Claude Code spend significantly more than average users, indicating that revenue figures may not reflect the total number of customers switching to newer models. Developers may gravitate towards the latest models without assessing whether older, less expensive options could suffice, contributing to rising costs.