How does TAHO Labs engage with partners? Here's what we don't do.
We don’t ask for your model. We don’t ask you to move a workload into our environment. We don’t ask for a rewrite, and we don’t ask you to stand up a second stack alongside your own.
One real workload, on your infrastructure. Then the number. DM if you're curious to learn more.
GPU lead times are still 36 to 52 weeks. So you commit capital today for hardware that lands next fall, into a building that may not see power until 2029, to run workloads you’ll mostly leave idle... Walk me through the part of the spreadsheet where this is a good IRR.
Everyone models this as a supply problem: buy more chips, pour more concrete, sign more PPAs.
Almost nobody models it as an efficiency problem. Double the utilization on the installed base and you’ve effectively built a second installed base without a single new interconnection request.
My mental model for AI infrastructure right now is one line: revenue equals tokens per watt times available gigawatts.
Everything else seems like a rounding error.
Most companies are optimizing neither variable. They’re optimizing the press release about how many GPUs they bought
The fastest workload is the one that never runs. Content-addressed execution means work computed once stays globally referenceable, so repeated jobs quietly stop being jobs
Arguably the most detailed compute disclosure produced landed Monday, and it still can't tell you whether we got more efficient.
Anthropic S1's compute line: 201% of revenue in H1 2025, 55% in H1 2026. Pricing, amortisation and efficiency all land inside one line.
Huh.
AWS is out of compute until beyond 2028.
Andy Jassy told z16z co-founder, Ben Horowitz directly.
Not "supply is tight"
Not "we're scaling."
Out. Until beyond 2028.
This is where TAHO comes in. Get more capacity out of the compute you already have.
https://t.co/ejUl1K67GC
so i actually starting watching and wow this one is insane. nothing like it exists on the internet (and its free!)
@neuralkian literally holds your hand all the way through. no excuses. this is not another hand wavy gpu economics video. everything is explained starting from a simple physics formula.
also lots of lore about how the industry actually works and how the math connects. well fucking done!
Gotta say - I love Rick Rubin, and really looking forward to his new book 'The Way of Code'
Rubin is an interesting persona, because he's recognized as an incredible producer, but he doesn't play a single instrument. He also states that he has no idea how to run a session or a board.
So what exactly is his superpower?
He listens deeper than the people who do, and trusts his instincts deeply.
Founders tend to believe the edge is technical, but usually the edge is noticing what everyone else stopped noticing.