The humanoid robot story has a trap in it, and the trap is the word memory.
Analysts this week said the next breakthrough is memory. The reflex is Micron. DRAM. Chips.
That reflex is wrong, and they said so directly.
The bottleneck is not a shortage of memory chips. It is the ability to use and retain information. A robot fails a task and adjusts from why it failed. That is software and architecture. Bernstein called it a new research frontier. JPMorgan called the brain the single largest bottleneck keeping the industry from scaling.
So the memory in robots need better memory is not the memory you buy from a chipmaker.
Where they said value actually accrues:
The brain builders. Physical Intelligence, a two-year-old startup at a $5.6B valuation, with Bezos and OpenAI in the cap table. $NVDA, Alphabet, and Figure on the model side.
The picks and shovels. This is the layer retail skips. JPMorgan flagged Timken, Regal Rexnord, and Allient — bearings, gears, actuators. A $4B addressable market by 2030. Industrial names, not chip names.
Morgan Stanley's long model is $5 trillion by 2050 and over a billion robots deployed. That is the horizon. It is not the trade.
Same lesson as the rest of the AI tape this year. The obvious ticker is the crowded one. The value sits one layer over from the headline.
$MU $NVDA $SPY
Analysts this week said the next breakthrough is memory. The reflex is Micron. DRAM. Chips.
That reflex is wrong, and they said so directly.
The bottleneck is not a shortage of memory chips. It is the ability to use and retain information. A robot fails a task and adjusts from why it failed. That is software and architecture. Bernstein called it a new research frontier. JPMorgan called the brain the single largest bottleneck keeping the industry from scaling.
So the memory in robots need better memory is not the memory you buy from a chipmaker.
Where they said value actually accrues:
The brain builders. Physical Intelligence, a two-year-old startup at a $5.6B valuation, with Bezos and OpenAI in the cap table. $NVDA, Alphabet, and Figure on the model side.
The picks and shovels. This is the layer retail skips. JPMorgan flagged Timken, Regal Rexnord, and Allient — bearings, gears, actuators. A $4B addressable market by 2030. Industrial names, not chip names.
Morgan Stanley's long model is $5 trillion by 2050 and over a billion robots deployed. That is the horizon. It is not the trade.
Same lesson as the rest of the AI tape this year. The obvious ticker is the crowded one. The value sits one layer over from the headline.
$MU $NVDA $SPY
