Infrastructure
SpaceXAI Puts Grok's Compute Beyond Any Grid
NVIDIA's new Vera CPU now runs Grok's AI orchestration for SpaceXAI, and the same reveal unveils Starmind, the first AI satellite built for orbit.
Infrastructure
Infrastructure
AWS's Next 2 Million GPUs Skip Naming Europe
AWS and NVIDIA's new 2 million GPU deal names a 100,000-GPU US government carve-out. It names no equivalent commitment for Europe.
Infrastructure
Why OpenAI's New Chip Won't Replace Nvidia
OpenAI's Jalapeno chip beat Nvidia's Blackwell on performance per watt in new benchmarks, but OpenAI's own CFO says it complements Nvidia rather than replacing it. Here is what that means for anyone buying AI compute.
Infrastructure
Cisco Just Named Sovereign Clouds As A Customer
Cisco and Nvidia added Supermicro rack-scale systems to their Secure AI Factory, and named sovereign cloud operators as a target buyer, not just hyperscalers. Available October 2026.
Infrastructure
OpenAI's Efficiency Claim Has No Independent Check Yet
OpenAI published first benchmarks for its Jalapeno inference chip, claiming better power efficiency than Nvidia's GB300. The numbers are self-reported and unaudited.
Infrastructure
775 Microns Now Caps Every AI Memory Chip Nvidia Ships
SK Hynix and Samsung showed rival memory-packaging fixes at Hot Chips 2026, and SK Hynix is running the safe one through Nvidia's whole Rubin generation.
Infrastructure
The Chip Nvidia Licensed From Groq Now Ships
Nvidia's Groq 3 LPX inference chip entered full production on August 24, 2026, built on technology and talent Nvidia took from Groq in a December 2025 licensing deal.
Infrastructure
AMD Crosses 30 Percent of Every x86 Chip Segment
AMD hit 30.3 percent of x86 client CPU shipments in Q2 2026, per Mercury Research, its first time above 30 percent in every segment at once.
Infrastructure
Nvidia Backs Starcloud's $2.3B Orbital Data Bet
Starcloud raised $250 million at a $2.3 billion valuation with Nvidia as a new investor, betting solar-powered orbital data centers can outrun grid limits.
Infrastructure
IBM Fuses Arm and Mainframe Code on One Chip
IBM's new 2nm processor runs Arm and IBM Z instructions on the same core - the first chip from its April 2026 Arm partnership, aimed at mainframe's staffing gap.
Infrastructure
SK Hynix Names Interconnect as AI's Next Bottleneck
SK hynix's peer-reviewed CPO roadmap sets targets of over 100 Tb/s, under 1 pJ/bit, and sub-10ns latency - and signals that interconnect, not compute, is AI infrastructure's next constraint.
Infrastructure
15 Percent More: Bloomberg Says Nvidia Warned Its Biggest Cloud Customers
Bloomberg reports Nvidia told Microsoft, Google and Oracle that AI servers using its Vera Rubin and Grace Blackwell chips will cost over 15% more from 2027, driven by memory prices.
Page 3 / 11
The Operator Brief
One considered note on infrastructure, governance, and measurement, most mornings. No theory.