AI Infrastructure
GPUs, silicon, data centres and the data platforms underneath.
How TypeSafe AI's Jev and Classification Routing Help Optimize Inference Costs on GPU Cloud
A comprehensive look at TypeSafe AI's Jev, classification routing, LLM cost breakdowns, and strategies to optimize inference costs on GPU cloud infrastructure.
SAP Bought Data Visibility in 2007. India's Tech Services Giant HCL Is Getting Into the AI Datacenter Business to Own It Today
SAP's €4.8 billion acquisition of Business Objects rewired the enterprise analytics landscape. Eighteen years later, HCL is following the same logic—buying infrastructure rather than software—this time for the AI era.
Efficiency Without Savings: Why Algorithmic Gains Inflate Compute Budgets and Complicate Cost Allocation for Shared GPU Clusters
Algorithmic progress makes AI models dramatically cheaper to train, yet Epoch AI's research suggests this could drive total compute spending higher rather than lower. Explore the economic logic, neural scaling dynamics, and what this means for cost allocation for shared GPU clusters.
Inside CoreWeave’s Marble, North Carolina AI Compute Site
An investigative look at CoreWeave's Marble, North Carolina data center—examining its evolution from a Core Scientific crypto-mining site to a 65 MW high-density GPU facility hosting thousands of Nvidia Blackwell chips.
DeepSeek Enters Silicon Business: plans to develop its own AI chips amid US export controls
DeepSeek, the Chinese startup developing large language models competitive with US companies like OpenAI and Anthropic, is planning to enter the silicon business by developing its own AI chips, according to Reuters.
Meta's 92,000 GPU Farm in the Nebraska Cornfields
Meta Sarpy in Springfield, Nebraska packs an estimated 92k H100-equivalents across 146 acres. Here's what the numbers actually mean, where they came from, and why "operational" is doing a lot of heavy lifting.
The Power Grid Bottleneck: Why Tech Titans and Startups Are Racing to Build Gas Turbines
As AI data centers strain energy grids and gas turbine wait times stretch into years, startups like Parallax and tech moguls like Elon Musk are diving headfirst into power generation.
Two Women, $125 Million, and the Defense Tech Gold Rush Israel Built Without Silicon Valley
Protego Ventures has completed the final close of its $125 million debut fund, backing Israeli and global defense technology startups. Here’s what the fund’s strategy, portfolio and next steps reveal about a changing sector.
Silicon Valley's Ideological Blueprint and the Real AI Impact on Human Psychology
Expanded article exploring Silicon Valley's ideological vision and the profound psychological, cognitive, and cultural liabilities of AI.
China Introduces Xing4.0-29B: A Single‑GPU Agentic Model for Consumer and Enterprise Use
Official launch article for Xing4.0-29B-A4B, covering announcement, architecture, deployment, serving, adoption, and future outlook.
Intel Cancels Falcon Shores, Pivots to Internal Testing and Jaguar Shores System-Level Solution
Intel will not bring its Falcon Shores AI chip to market, repurposing it as an internal test chip while focusing on Jaguar Shores as a system-level rack-scale solution for AI data centers. The decision follows underwhelming Gaudi 3 performance and competitive pressure from AMD and Nvidia.
Nvidia Expands AI Leadership Through Data Center Orchestration Beyond GPUs
Article covers how Nvidia's AI advantage extends beyond GPUs into data center systems architecture, including Vera Rubin architecture with specialized CPUs, inference accelerators, and traffic control systems that improve efficiency through smarter data orchestration rather than just increased compute.