Hyperscaler CapEx fuels hardware over traditional software
The discussion highlighted a stark divergence where hardware, networking, and physical infrastructure are booming due to massive hyperscaler spending, while traditional software remains cold.
The argument
Hyperscalers and model companies are spending an estimated $750 billion to $1 trillion annually, with the majority flowing to compute, power, and networking. The speakers noted that while this creates a bubble-like environment, bureaucratic delays in building data centers may actually prevent oversupply and keep compute prices high.
The thesis, stress-tested
✓ What validates it
- ✓Hyperscalers maintaining or increasing their capital expenditure guidance in upcoming quarters
- ✓Data center construction delays keeping GPU rental and compute pricing elevated
▸ Risks discussed
- ▸A sudden reduction in CapEx by major hyperscalers
- ▸An oversupply of data center capacity if permitting bottlenecks ease rapidly
Hear it yourself
"You're gonna send me an email confirming you're in, and then we'll collect the money. End of conversation. The OpenAI round is, well, Amazon, you're gonna give us 50,000,000,000, but 20,000,000,000 is gonna be upfront. The other 30,000,000,000 is contingent on us going public or AGI. And, Masasat, you're gonna give us 40,000,000,000 from SoftBank, but you gotta borrow 30,000,000,000 to get that 40,000,000,000. So we're gonna give you a little time to pay that money. So we're closing on a 110,000,000,000 of which 20,000,000,000 is clearing now, 30,000,000,000 in six months' time depending on the lending market."
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE