Fun fact: I wrote a memo for my friends at Inflection more than a year ago arguing that their best play was to specialize in models you can fit on an iPhone.
Now - I obviously see the value of AI on the edge, but to me the small models are likely to be limited to UI-for-unstructured-data problems. I expect agents to become broadly available within a year (see Devin from Cognition, Minion AI for early glimpses). They will require inference in the data centers. Memory seems like the best play targeting both.
What about $ENVX as a play on needing battery life for AI on smartphones?
If you use "aiPhone" it's probably like 57% liss cringe
Fun fact: I wrote a memo for my friends at Inflection more than a year ago arguing that their best play was to specialize in models you can fit on an iPhone.
Now - I obviously see the value of AI on the edge, but to me the small models are likely to be limited to UI-for-unstructured-data problems. I expect agents to become broadly available within a year (see Devin from Cognition, Minion AI for early glimpses). They will require inference in the data centers. Memory seems like the best play targeting both.
Fantastic write up.
Why AMKR when BESI NA is the sole provider for these next gen Apple phones? They dominate the smallest nanometer category, incredible company imo