AI Inference

NVIDIA’s $7B Poolside Bet: Why Open-Weight AI Is a Hardware Strategy

by
NVIDIA CEO Jensen Huang on stage pointing to three open rack-scale AI server hardware modules and GPU compute trays.

NVIDIA is committing approximately $7 billion around AI startup Poolside, channeling the lion’s share of that capital into open-weight model initiatives. At first glance, this looks like an economic contradiction. Why would the world’s undisputed sovereign of accelerated computing spend billions developing architectures and weights that developers, enterprises, and sovereign nations can download, modify, fine-tune, … Read more

Mistral and HUMAIN: How Saudi Arabia Is Building a Sovereign AI Powerhouse

by
Mistral and HUMAIN partnership for sovereign AI infrastructure in Saudi Arabia

Saudi Arabia’s AI ambitions are moving beyond buying access to foreign proprietary models. The kingdom increasingly wants the compute, infrastructure, localized models, and data sovereignty required to operate artificial intelligence entirely on its own terms. Paris-based Mistral AI and Saudi state-backed AI powerhouse HUMAIN have formed a major strategic collaboration Mistral and Humain worth hundreds … Read more

NVIDIA Rubin GPU the powerhouse: Architecture, HBM4, Performance and the Future of AI Inference

by
NVIDIA Rubin AI platform rack-scale system with multiple GPU and networking modules in a dark data center environment.

NVIDIA Rubin GPU, the economics of artificial intelligence have reached an inflection point. As frontier reasoning models execute extended test-time compute loops, multi-agent frameworks run continuous tool-use pipelines, and video models generate massive contextual volumes, inference has surpassed training as the primary operational cost for AI enterprises. Running trillion-parameter architectures at production scale under older … Read more

Apple M6 and M5 Ultra Push AI Beyond the Cloud With Powerful On-Device Computing

by
Square graphic showing the Apple M6 chip with a green and blue ambient glow alongside the Apple M5 Ultra chip with a purple and blue glow against an off-white background.

For the past several years, the explosive growth of artificial intelligence has been inextricably tethered to the cloud. Whenever a user drafted a complex prompt, synthesized code, or generated high-resolution media, that request was almost certainly shipped across fiber-optic cables to a massive, power-hungry data center packed with server-grade accelerator racks. That paradigm is undergoing … Read more