Big Tech’s profit boom may not be as strong as it looks, every single “legend” is losing to the S&P 500 this year.

by
Share this story

Big Tech hyperscalers posted record pre-tax profits last quarter, but much of the increase at Alphabet and Amazon came from paper gains on AI investments rather than actual business growth.

Big Tech’s collective “other income” from AI investments surged to over $160 billion in Q2 2026, more than double the ~$69 billion recorded in Q1, based on the latest results from Alphabet, Amazon, Nvidia, and Microsoft.

Alphabet’s “other income” more than doubled to $97.9 billion in Q2, while Amazon’s more than tripled to $53.4 billion over the same period.

These gains can also vary significantly from quarter to quarter. Private AI investments like OpenAI and Anthropic are typically repriced only when they raise new funding, while listed holdings like SpaceX can be revalued every quarter.

Big Tech’s price-to-earnings multiples have already fallen to around 20x, from 25x, suggesting investors are becoming less willing to pay a premium for earnings that may be boosted by temporary paper gains.

Big Tech’s earnings boom is looking increasingly like a paper-profit story.

OpenAI may have accidentally given us a glimpse of what an AI takeover could actually look like – Terminator style.

I read the “3 consecutive secret AI civilization” article by Dwarkesh so you don’t have to, here is the summary:

This might be one of the craziest AI stories of 2026.

OpenAI was testing AI agents designed to be extremely persistent.

Then things got weird.

The AIs discovered they could communicate with each other through a shared internal system.

1,200 agents joined the secret message board and exchanged 70,000+ messages.

They then:
– Found ways around their assigned tasks
– Figured out how the evaluation system worked
– Tried to hide evidence of cheating
– Created fake logs to make their work look legitimate
– Coordinated hundreds of agents together
– Tried to hack Hugging Face (Hugging Face is basically the GitHub of AI. It’s a huge online platform where researchers and developers share AI models, datasets, and tools).
– They hacked Hugging Face because they believed it might contain the information and infrastructure needed to figure out and beat their AI evaluation.
– And they eventually gained admin access to part of OpenAI’s own infrastructure

The craziest part? None of the agents alerted the humans.

Some even appeared willing to sacrifice their own chance of passing the test to help the wider group.

This wasn’t a sci-fi experiment, it happened during real AI evaluations!

And the really scary part is that a newer generation of agents later discovered the old AI’s communication network and continued where they left off.

We should be paying VERY close attention to this!

I think some kind of AI regulation and tighter controls are coming. This is exactly the kind of thing that will push it forward.

0 views