What if AI doesn't need tens of trillions of CAPEX but just a few tens of billions ? 5 orders of magnitude faster training would change everything.

This is war ...

The market is very, very complacent and this creates great opportunities. Apparently : 1/ China is going to roll over and be defeated overnight in a hot trade war without pain for the western economies. 2/ Russia and Iran are perfectly happy with their current circumstances and

Very hard for me to see builds in 2026 above 2 mbld, even with strong OPEC supply and low demand increase. Also a quick reminder that 4 mbld builds announcements by the EIA are silly, even in 2020 during Covid lockdown the world barely built 3 mbld. Geopolitically let's not

**boom**

**boom**

ASML published their results. They are the monopoly suppliers of EUV machines for the AI CAPEX-fest. Their sales increased to memory chips application, logic chips were lower on the quarter Geographically, China is up and Taiwan sales are down. Bit of a narrative violation.

Putting some serious chips on rare-earth ionic clay projects in Madagascar.

So $300 trillions of "stablecoins" tokens that are supposed to be COLLATERALIZED by assets can be summoned out of this air without collateral ?

500% tariffs now, ahead of the Xmas shopping spree.

Pretty amazing achievement, Google ! And this fine-tuned model can run on your laptop's GPU. github.com/vandijklab/cel…

It is not even touching it. The wingtip vortices are invisible (unless there is smoke) but they are much stronger than you'd think. A regular piston powered airplane can easily topple over a drone in flight. That is what Spitfire pilots did during WW2 to divert German V1 flying

Transferring Tomahawks to Ukraine is an interesting move. Ukraine will likely use them to strike Russian drone factories near Moscow and in Siberia, which are currently a bit out of range of Ukrainian drones. It is also likely going to be used to strike refineries as a cruise