SemiAnalysis analyst Dylan Patel argued substrates, memory, lasers, and optics are more binding constraints than power for AI data centers
SpaceX to Build Turbine Blade Foundry for AI Data Center Power
AI data centers cannot scale without power, and the supply chain for both grid connections and on-site generation hardware has become a hard ceiling on buildout speed. SpaceX entering turbine blade manufacturing could reduce one specific bottleneck, but the broader infrastructure gap spans transformers, substations, substrates, and memory.
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.
The full picture
Elon Musk announced SpaceX will bring turbine blade and vane casting in-house, a move he said could cut natural gas generator procurement delays by up to 18 months. Turbine blade production currently takes 60 to 90 weeks per batch, and surging AI data center demand has created a supply bottleneck for blades and vanes. Howmet Aerospace, which holds roughly half of the global market share for natural gas turbine components, saw its stock fall more than 7% following the announcement before partially recovering. The broader context is a severe power infrastructure shortage: U.S. electricity grid connection delays are pushing AI data center projects toward portable natural gas turbine generators as an alternative. Bloomberg reported that building private gas generation has tripled in cost in recent years, and power-gear backlogs now run past 2030.
How it developed
SemiAnalysis forecast 51.2 GW of new data center capacity in 2027; satellite imagery confirmed 24 GW already under advanced construction
NVIDIA reported DSX platform adoption by GE Vernova, Hitachi Energy, Schneider Electric, and Siemens Energy to address power bottlenecks
Trump posted on Truth Social that large new data centers must generate their own electricity and pay host communities a dividend
JPMorgan projected global AI compute demand of 530 GW by 2031 against supply of 250 GW
Microsoft terminated its UK data center deal with Nscale after grid officials said power would not be available until the 2030s
Amazon CEO Andy Jassy reportedly told an investor AI compute capacity is sold out beyond 2028
Bloomberg analysis detailed costs for a gigawatt-scale Texas data center; private gas generation costs have tripled in recent years
ERCOT's large load queue reported at 474 GW, up from 63 GW in 18 months; Texas paused new data center approvals
Transformer News reported US AI data center power demand projected to grow from 5 GW in 2025 to 50 GW by 2030, with transformer lead times of five or more years
SpaceX confirmed foundry plans; Howmet Aerospace stock fell more than 7% on the announcement before partially recovering
Elon Musk announced on X that SpaceX will bring turbine blade and vane casting in-house, cutting generator procurement delays by up to 18 months
Sources
- SemiAnalysis Forecasts Historic AI Data Center Buildout Doubling in 2027
- JPMorgan Forecasts Massive AI Compute Supply Shortfall Through 2031
- Microsoft just walked away from one of the biggest AI data center projects in the United Kingdom.
- I think investors are MASSIVELY underestimating just how constrained we are when it comes to AI infrastructure right now…
- Texas just exposed the real AI bottleneck and it will make already powered data centers dramatically more valuable.
Related
- Related storyData center power demand versus grid supply
- TrendForce datacenter power demand outlook
- Nvidia AI server prices and memory shortage
- SpaceX Super Intelligence and Tesla merger talks
- Columbia's US AI infrastructure spending research
BackgroundCompute: the chips, clouds and data centers behind AI
Get the daily email
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.