Futurum Group CEO Daniel Newman talks demand concerns weighing on the AI trade

AI Chip Market Dynamics - Transition from training to inference may introduce cheaper, less powerful AI chip alternatives [1] - Nvidia aims for frequent chip upgrades, while cloud players prefer longer lifecycles for accounting purposes [2] - Nvidia promotes the economic benefits of its next-generation chips [3] - Older generation Nvidia chips (V series, A series) and TPUs still command meaningful commercial pricing [5] - Current AI chip market lacks supply-demand equilibrium, with high demand for every available chip [6] Investment and Risk Considerations - Over-investment poses a significant risk, especially for companies doing things on spec [1][8] - Depreciation issues could become significant towards the end of the decade when supply and demand reach equilibrium [7] - A six-year depreciation model and GPU collateralization create risk, particularly without value-added services [13] - Potential impact on valuations if AI demand slows down and companies revise constrained numbers [11][12] Company-Specific Insights - AMD is confident in doubling its total addressable market (TAM), including networking and CPU chips [10] - AMD is experiencing margin growth and data center growth due to supply constraints [10]