THE POST
And electrical equipment, cooling, networking and batteries https://t.co/aW2r7j8gvJ
QUOTED
Euwyn@euwyn
Elon Musk says the AI boom will hit a WALL by the end of 2026. He says the whole industry is about to learn a hard lesson in hardware. Chip output is exploding, but electricity is nearly flat everywhere outside China. So the two lines cross this year. He says by the end of the year, we will be making more AI chips than we can actually turn on. Millions of them pile up with nowhere to plug in. In his words, the bottleneck for the next year is not chips, it is raw power. Whoever can turn on the most chips fastest wins the race. In the end, the AI race comes down to one thing: Raw electricity. — Elon Musk (.@elonmusk) on Dwarkesh Patel's (.@dwarkesh_sp) podcast
https://x.com/euwyn/status/2117572576390349155Text is the full long-post text the API returned in note_tweet. The quoted post carries a 68-second video clip from the podcast, which I did not watch or copy into this file.
ENGAGEMENT
- likes36,968
- reposts3,681
- replies1,959
- quotes314
- bookmarks5,845
- views4,160,911
AS RETURNED BY THE X API · FETCHED 2026-11-04 06:18 UTC
WHAT IT SAYS
Musk quotes a summary of his podcast argument that AI chip output is outrunning electricity supply, so that by the end of 2026 more chips will be built than can be switched on. His own addition is that the bottleneck also covers electrical equipment, cooling, networking and batteries, not just generation.
CONTEXT
Electricity is growing, but slowly outside China. Ember counts global demand up 2.8% (+849 TWh) in 2025 to 31,779 TWh, with China adding 503 TWh, 59% of the increase, and the US up 3.0%. Observatory arithmetic from those figures: demand outside China rose by about 346 TWh, roughly 1.7%, so 'nearly flat' overstates it but the gap with chip growth is real. Ember cites IEA estimates that data centres added about 60 TWh in 2025 and could roughly double from about 500 TWh to 950 TWh by 2030.
The equipment side of the claim has support in this week's reporting. TechCrunch, citing the New York Times, says some AI campuses are running converted jet engines on gas because a 100 MW grid connection in Northern Virginia takes about seven years, and that Microsoft dropped an anchor deal in Loughton, England, after the utility could not deliver power until the early-to-mid 2030s.
GlobalData, via Data Center Knowledge, finds projects now wait just over two years from announcement to construction, up from about 18 months, and says grids take five to 15 years to build against one to three for a data centre. An IDC analyst quoted in the same piece says there is no hard evidence yet that compute capacity is limiting AI development.
WHAT IT MEANS FOR K
If chips really outrun power, the pressure falls on building generation, transformers, cooling and storage faster, which is exactly the lever that raises K. The claim is a forecast, so the effect on harnessed power is still ahead of us.
CLAIMS
- AI chip production will exceed the power available to run them by the end of 2026.
- Electricity supply is nearly flat everywhere outside China.
- Electrical equipment, cooling, networking and batteries are also bottlenecks.
EVIDENCE
- Ember: global demand +2.8% in 2025, China 59% of growth, US +3.0%.
- TechCrunch/NYT: about seven years for a 100 MW grid hook-up in Northern Virginia; converted jet engines used as a 24-month bridge.
- GlobalData: announcement-to-construction waits up from about 18 months to just over two years.
- IDC: no hard evidence yet that compute capacity limits AI development.
REASONING
- Delays in grid hook-ups and equipment are well documented, which supports the direction of the claim.
- 'Nearly flat outside China' is stronger than the data: Observatory arithmetic puts non-China growth at about 1.7% in 2025.
- Whether chips go unpowered at scale by year end is a prediction with no direct evidence yet, so the verdict is speculative.
SEE ON THIS SITE
TAGS
- COMPUTE
- AI POWER
- GRIDS
- BATTERIES
- DATA CENTRES