Key figures - 9 months ending Q3:
$33B in revenue; $34.5B operating expenses - $1.25B operating loss
$154B in assets vs. $146B in cash liabilities
Historical net tangible book value per share as of September 30, 2026
-$68.47
Employees have stock at an average of $2.91 per share.
Number of > $100K customers hasn't risen from Q2 - Q3 amendments
Revenue per compute falling from 2.08x -> 1.72x -> this will get worse as their compute commitments go up and their revenue per token goes down.
"In the three months ended September 30, 2026, our underlying compute efficiency improvements were partially offset by the delivery of additional compute, including $5.0 billion of compute under short-term commitments, which currently costs significantly more than comparable compute procured under our multi-year commitments." aka as we buy this extremely expensive compute and can't sell it for more money (e.g. $) our efficiency goes down.
It's a wonderful technology, world-changing. It's been a great business.
It's not obvious any use case will take over the 2026 token-maxxing S Curve.
They also finally released their compute totals - ~$900B in compute spend committed
"As of and subsequent to September 30, 2026, we have entered into purchase commitments and lease agreements, including short-term commitments, as well as leases not yet commenced, with an aggregate committed spend of approximately $895.4 billion, substantially all of which is related to compute and infrastructure. These commitments include approximately $447.5 billion expected to be incurred before December 31, 2030, and substantially all of the remainder is expected to be incurred between 2031 and 2045. "
In Q3, I think they struggled with the government ban on Mythos/Fable, Opus 5 not being so good, OpenAI models being more token efficient, and they decided to dedicate a higher percentage of compute to training.
Revenue per token is down tho. Claude was on a untouchable position 1H2026 that didn't repeat afterwards with competition forcing their hand on price cuts.
"In the three months ended September 30, 2026, our underlying compute efficiency improvements were partially offset by the delivery of additional compute, including $5.0 billion of compute under short-term commitments, which currently costs significantly more than comparable compute procured under our multi-year commitments." aka as we buy this extremely expensive compute and can't sell it for more money (e.g. $) our efficiency goes down.
It's a wonderful technology, world-changing. It's been a great business.
It's not obvious any use case will take over the 2026 token-maxxing S Curve.
Let's hope they make it out alive!
"As of and subsequent to September 30, 2026, we have entered into purchase commitments and lease agreements, including short-term commitments, as well as leases not yet commenced, with an aggregate committed spend of approximately $895.4 billion, substantially all of which is related to compute and infrastructure. These commitments include approximately $447.5 billion expected to be incurred before December 31, 2030, and substantially all of the remainder is expected to be incurred between 2031 and 2045. "
In Q3, I think they struggled with the government ban on Mythos/Fable, Opus 5 not being so good, OpenAI models being more token efficient, and they decided to dedicate a higher percentage of compute to training.