Insights on GPU infrastructure, AI, and data centers.
CoreWeave's H100 GPUs from 2022 contract expirations immediately rebooked at 95% of original pricing.¹ The data point reveals something counterintuitive: AI accelerators retain substantial value even
The handshake between Jensen Huang and Sam Altman represents more than corporate diplomacy. Their companies announced a letter of intent to build **10 gigawatts** of AI infrastructure—enough
Virginia's Senate moves to kill $1.6B data center tax exemption by Jan 2027 while the House fights to keep it. The outcome reshapes Data Center Alley.
US data center capacity under construction fell to 5.99 GW from 6.35 GW. Power, moratoriums, and labor shortages shifted the bottleneck to execution.
Meta's $100B AMD deal deploys 6GW of custom MI450 GPUs, breaking NVIDIA's hyperscaler exclusivity and reshaping the AI chip market permanently.
Seven hyperscalers sign Trump's Ratepayer Protection Pledge March 4, committing to build, bring, or buy power for AI data centers across 56 GW.
PJM capacity prices surged 10x to $329/MW-day, adding an estimated $100B+ in costs for 67M ratepayers. States respond with rate reforms and tax repeal.
Facebook's production AI cluster catastrophically failed 72 hours after deployment when synchronized training jobs triggered thermal runaway across 2,000 H100 GPUs, causing $28 million in hardware
RLHF training spends 80% of compute time on sample generation, making throughput optimization the critical infrastructure challenge for organizations aligning large language models with human
China controls 60% of germanium production, a critical fiber optic dopant. With AI racks demanding 36x more fiber and prices up 200%, the $690B buildout faces a mineral chokepoint.
The gaming industry operates infrastructure at a scale that rivals enterprise computing. With over $200 billion in annual revenue and hundreds of millions of concurrent players globally, modern games
Mistral AI launched directly into Anthropic's territory today. The French AI company released Devstral 2, a 123-billion parameter coding model, alongside Vibe CLI, an open-source command-line coding
Tell us about your project and we'll respond within 72 hours.
Thank you for your inquiry. Our team will review your request and respond within 72 hours.