Uploaded August 2026 | Updated September 2026, 3 weeks ago
Michael Burry, the Big Short guy, missed one number and it breaks his
whole argument.
How we're taking advantage of news like this is inside
skool.com/coinpicksgenesis, link in my bio, a dollar a month.
His case is that hyperscalers depreciate AI chips over five or six
years while Nvidia ships a new generation every year. Hopper in 2022,
Blackwell in 2024, Rubin this year. Burry puts the understatement at
$176 billion through 2028.
The math is fine. The target moved.
SemiAnalysis tracks what it costs to rent an H100, the chip from
2023. It bottomed at $1.70 an hour in October and hit $2.35 by March,
nearly 40 percent in six months, while two newer generations were
already shipping. That is what scarcity looks like.
Here is why. In Virginia the main utility has roughly 70 gigawatts of
projects sitting in its queue, about three times its all time peak.
The average wait for a grid connection is seven years. Then you need a
transformer, and those are averaging 128 weeks.
A chip does not lose its value when a better chip comes out. It loses
it the day somebody can fill that same megawatt with something more
profitable. Right now nobody can, so the old stuff keeps earning. The
argument was never about chips. It is about electricity, and more than
$750 billion of capex is sitting on the answer.
Follow for the moves the news skips.
Michael Burry, the Big Short guy, missed one number and it breaks his
whole argument.
How we're taking advantage of news like this is inside
skool.com/coinpicksgenesis, link in my bio, a dollar a month.
His case is that hyperscalers depreciate AI chips over five or six
years while Nvidia ships a new generation every year. Hopper in 2022,
Blackwell in 2024, Rubin this year. Burry puts the understatement at
$176 billion through 2028.
The math is fine. The target moved.
SemiAnalysis tracks what it costs to rent an H100, the chip from
2023. It bottomed at $1.70 an hour in October and hit $2.35 by March,
nearly 40 percent in six months, while two newer generations were
already shipping. That is what scarcity looks like.
Here is why. In Virginia the main utility has roughly 70 gigawatts of
projects sitting in its queue, about three times its all time peak.
The average wait for a grid connection is seven years. Then you need a
transformer, and those are averaging 128 weeks.
A chip does not lose its value when a better chip comes out. It loses
it the day somebody can fill that same megawatt with something more
profitable. Right now nobody can, so the old stuff keeps earning. The
argument was never about chips. It is about electricity, and more than
$750 billion of capex is sitting on the answer.
Follow for the moves the news skips.
