Bitka za prevlado v UI ni v čipih, ampak v teravatih kapacitet in teravatnih urah energije na voljo za uporabo umetne inteligence. In v ceni te energije. Zmagala bo država, ki bo zagotovila dovolj velike kapacitete za stabilno in stanovitno proizvodnjo elektrike ter hkrati najnižje cene elektrike. Podatkovnih centrov se ne da poganjati na solarne panele in baterije, ker je proizvodnja elektrike nestanovitna, njena cena pa bistveno previsoka. Zato se podatkovni centri selijo v bližino jedrskih elektrarn, trije največji ameriški ponudniki (Google, Microsoft, Amazon) pa investirajo v jedrske elektrarne.
Vendar bodo to bitko zmagali kitajski ponudniki. KItajska je v zadnjih 15 letih izgradila več kapacitet za proizvodnjo elektrike kot ves preostali svet skupaj. Kitajska gradi 30 novih jedrskih elektrarn, ZDA nobene. Podatkovni centri na Kitajskem plačujejo 3 cente za kilovatno uro, v Ameriki pa 3-krat več.
In seveda, več kot očitno je, da je Evropa že zdavnaj odpadla iz te igre za umetno inteligenco. Na solarne panele in vetrrnice in po ceni 30 centov na kilovatno uro ne moreš poganjati podatkovnih centrov.
Zelena norost je v Evropi ubila ne samo tradicionalno industrijo, pač pa preprečuje tudi razvoj in razmah industrij prihodnosti. Pritožbe pošljite političnim elitam, ki vodijo Evropo v propad.
_________
Everyone is watching the chip war.
Nobody is watching the power war.
That’s the trap.
China’s power grid: 3.9 terawatts
America’s power grid: 1.3 terawatts
Read that again.
China has 3x the electrical capacity of the United States.
They added more power generation between 2010-2024 than the rest of the world combined.
Right now:
→ China: 30 nuclear reactors under construction
→ United States: Zero
China’s data centers pay 3¢ per kilowatt-hour.
Virginia—America’s AI heartland—pays 9¢.
That’s not a gap.
That’s a structural execution.
Here’s what nobody will say publicly:
AI doesn’t run on chips.
AI runs on electricity.
The chip is the brain.
The grid is the blood supply.
You can design the most advanced GPU on Earth.
If you can’t power a million of them cheaply, you lose.
Goldman Sachs projects China will have 400 gigawatts of spare capacity by 2030.
That’s triple the entire projected global AI power demand.
They’re not building for today.
They’re building for the decade when AI scales 1000x.
My prediction (bookmark this):
By July 2026, the first major US tech company will announce an AI facility in a foreign country citing “power constraints” as the primary driver.
That headline will be the moment America realizes:
We didn’t lose the AI race in silicon.
We lost it in kilowatts.
The century belongs to whoever can keep the lights on.
DAMIJAN blog 