todd

Nvidia's Jensen Huang is incredible and the company is loved, but the economic life of GPUs is uncertain due to coming memory bottlenecks and architecture shifts, making long-term asset class durability risky.

Traditional SaaS seat license models face structural risk as AI-driven token usage and consumption-based pricing squeeze per-seat revenue.

AMD's upcoming MI450 ramp and strategy of providing compute without competing with customers positions it well in the AI infrastructure buildout.

Traditional SaaS seat license models face structural risk as AI-driven token usage and consumption-based pricing squeeze per-seat revenue.

Traditional SaaS seat license models face structural risk as AI-driven token usage and consumption-based pricing squeeze per-seat revenue.

Traditional seat-license software models (Salesforce, Workday, ServiceNow) face structural risk as AI token-based pricing squeezes per-seat revenue.

AMD's 450 ramp looks promising with a whole range of compute offerings and less customer competition than Nvidia, making it very well positioned in compute.

Traditional seat-license software models (Salesforce, Workday, ServiceNow) face structural risk as AI token-based pricing squeezes per-seat revenue.

Traditional seat-license software models (Salesforce, Workday, ServiceNow) face structural risk as AI token-based pricing squeezes per-seat revenue.

GPUs may become an asset class but face durability risk from memory bottlenecks and architecture changes in 5-10 years; circular financing in Nvidia's ecosystem creates risk for longer-term investors.