
| Token spending growth (Aug-Sept) | Flat (Ramp data) |
| Total AI spend increase (Oct) | +1.65% (Ramp data) |
| Frontier models share of spend | 69.5% (Ramp data) |
| Customer Success AI spend | 13.3% of total (author's claim) |
| Google Orchestrator efficiency | 24% less quota used (Google claim) |
Who will this analysis save an hour?We’ll send the link with the “actually” takeaway — they won’t have to watch the whole video.
Businesses are becoming 'smarter' by using smaller, specialized models and automated routing tools to avoid paying for expensive frontier intelligence for every task.
The video suggests a risk of oversupply if the massive capacity NVIDIA built doesn't match the slower-than-expected growth in actual customer spending.
These are Google's new specialized models: Carbon is for complex reasoning, while Hydrogen is optimized for speed and lower cost.
By using an orchestrator or manual routing to ensure only the most difficult problems use the most expensive models like GPT-4 or Gemini Ultra.
The author argues that a declining engineering culture, loss of senior talent, and over-reliance on AI-generated fixes may be hindering system reliability.
★ 6/10
★ 5
★ 9