⚡ Hardware & Infrastructure

AI chips, GPUs, data centers, cloud compute, and the hardware powering the AI revolution.

870 articles

Nvidia effectively shut out of China's AI accelerator market

Nvidia has effectively lost its AI accelerator business in China after U.S. export controls restricted shipments of advanced GPUs to the country. Source: finance.yahoo.com

China floats AI export limits, then lets DeepSeek and others keep selling abroad

Beijing has proposed restricting exports of certain AI technologies. It then quickly moved to let some of its flagship companies, including DeepSeek, continue shipping their products overseas. Source: theregister.com

SpaceX locks in $40B debt financing to buy AI GPUs, Musk says

Elon Musk announced in a post on X that SpaceX has finalized $40 billion in debt financing. The money is earmarked for purchasing GPUs to power AI workloads. Source: cnbc.com

Nvidia's AI financier role faces first major test as neoclouds slump

Nvidia has evolved from a chipmaker into a financier of the AI industry, and some investors see it shouldering outsized risk tied to the AI buildout. The slump hitting neocloud providers now puts that strategy under its first serious stress test. Source: cnbc.com

Fusion shows its AI In-a-Box system to Jensen Huang at GTC Berlin 2026

Fusion demonstrated its compact AI In-a-Box system to Nvidia CEO Jensen Huang at GTC Berlin 2026. The piece looks at what the system does, why it caught Huang's attention, and what it could mean going forward. Source: tech-insider.org

AI critic Zitron: Nvidia manufactures the GPU scarcity it profits from

Ed Zitron, CEO of PR firm EZPR and a prominent skeptic of the AI spending boom, argues that Nvidia is engineering the very scarcity it uses to sell its GPUs. In his view, the only shortage that exists is the one created by Jensen Huang. Source: finance.yahoo.com

OCP 2026: Nvidia, AMD and Intel push open AI rack designs

At OCP 2026 in San Jose, Nvidia released open Vera Rubin rack designs, AMD revealed its Helion rack, and Intel detailed its GPU D-Series. Pricing has not been announced for any of them. Source: shattered.io

Can Amazon's AI chips take market share from Nvidia?

Amazon's AI chips have gained ground in their ability to compete with Nvidia's GPUs. Even so, they're unlikely to have a significant effect on Nvidia's financial results. Source: fool.com

Nvidia debuts shipping-container AI factory as neocloud stocks fall

Nvidia says its DSX POD can shrink AI data center deployment from about six months to as little as 14 days. Deliveries are planned for early 2027. The announcement arrives as shares of neocloud companies slide. Source: qz.com

Nvidia's Huang urges Europe to build out AI data centers

Nvidia's CEO is pushing Europe to catch up in AI infrastructure, noting the continent holds only about 5% of global AI computing capacity. The U.S. accounts for roughly 75%. He encouraged a major data center buildout to close the gap. Source: broadbandbreakfast.com

China's AI compute to double next year but still leans on US chips

A report says China's access to AI computing power will double next year. However, US-designed chips are expected to make up about 65% of that capacity, underscoring continued dependence on American technology. Source: scmp.com

'Jensen's nightmare': Huawei and DeepSeek close the AI chip ecosystem gap with Nvidia

A deepening partnership with AI start-up DeepSeek is helping Huawei Technologies catch up with Nvidia in the AI chip ecosystem. The close tie between the two Chinese players is narrowing the gap that has long favored Nvidia. Source: scmp.com

Nvidia's Berlin GTC: faster AI factories, lower power draw, more secure agents

At GTC in Berlin, Nvidia unveiled DSX POD, a factory-built, modular data center approach meant to speed up the rollout of AI infrastructure. The company also announced further updates aimed at efficiency and agent security. Source: networkworld.com

Scaling Frontier-Class AI Models Across Eight NVIDIA DGX Spark Nodes

NVIDIA's technical blog describes how developers can run frontier-class models on a cluster of eight DGX Spark nodes. The need arises because AI workflows now often orchestrate several models in parallel and chain multiple inference calls rather than relying on a single model. Source: developer.nvidia.com

OpenAI's Richard Ho explains why the Jalapeño AI chip is a breakthrough

An interview with OpenAI's Richard Ho digs into Jalapeño, the new AI chip OpenAI unveiled in late August. The chip is making waves in the industry and is expected to stay influential for years to come. Source: taekim.substack.com

NVIDIA backs six chip suppliers in $279B drive to ease AI capacity crunch

At GTC Berlin on October 21, NVIDIA announced it will fund six supply chain partners to expand their manufacturing capacity. The move aims to break the bottleneck constraining AI infrastructure buildout. Source: app.dealroom.co

NVIDIA DSX POD aims to get AI capacity running faster and more predictably

AI factories convert energy into intelligence, but before a workload can generate its first token, operators must assemble an interconnected system of compute and supporting infrastructure. NVIDIA's technical blog presents DSX POD as a way to make that process faster and more predictable. Source: developer.nvidia.com

Nvidia unveils new AI reference architectures and development tools

Nvidia today introduced a set of technologies designed to make it easier for customers to build data centers and robots on its chips. Source: siliconangle.com

Nvidia wants Europe to build AI, not just use it, pledging nearly $10B

At GTC Berlin, Nvidia announced technology and ecosystem initiatives and committed nearly $10 billion to Europe's technology ecosystem. The focus is on helping European players create AI rather than simply consume it. Source: forbes.com

Nvidia's Huang sees Europe's edge in physical AI, with robot ducks at his side

Nvidia CEO Jensen Huang said on Wednesday that Europe's engineering expertise gives it an advantage in physical AI and robotics. He made the remarks flanked by robot ducks on stage. Source: reuters.com