Nvidia's Q2 FY2027 results show surging AI demand driving record revenue and massive memory procurement commitments totaling $279B.

Vera Rubin hardware is on track to account for 20% of Nvidia's data center revenue this quarter, marking the company's fastest platform ramp.

The reported acquisition would strengthen Nvidia's open-model AI strategy and ecosystem position against rivals.

SK hynix's US HBM assembly plant will connect DRAM wafers from South Korea with American AI customers, marking a key onshoring milestone.

OpenAI's purpose-built AI ASIC built from scratch for inference offers strong performance-per-watt and low latency, though raw throughput lags Nvidia's Blackwel

Nvidia's proprietary high-bandwidth memory implementation targets NVLink Fusion partners with a custom base die and PHY for improved performance.

Intel's 14A process defect rates are declining rapidly, with internal products already in development and external customers inquiring about capacity.

YMTC plans to nearly double its market share in 16 months to overtake Samsung and SK hynix in NAND flash production.

Nvidia's VP of hardware detailed the Groq 3 LPX rack architecture with LP30-based hardware already in production.

Cerebras revealed Nexus architecture tripling rack-scale performance and CS-6 wafer plans with stacked DRAM for future generations.

At Hot Chips, Nvidia emphasized power as the hard limit on data center capacity and detailed how Vera Rubin NVL72 systems maximize compute within fixed facility

80% of Micron's Taiwan workforce signaled willingness to strike unless the company replaces its bonus system with a profit-sharing scheme matching rival Korean

A second round of semiconductor tariffs could extend duties beyond chips to finished products, potentially scrapping earlier data center exemptions.

Nvidia pushes back against reports that it paused deals with cloud providers, following partner pushback over customer controls and potential antitrust concerns

Qualcomm is backing Modular's approach to decouple AI software from silicon, aiming to challenge Nvidia's dominance in AI workloads.

Glass-core substrates promised by Intel in 2023 are in final qualification but still not in any commercial product, with continued delays.

Thinner dies, increased TSV area, thermal dissipation challenges, and limited manufacturing capacity complicate HBM scaling as layer counts rise.

The DOJ and FBI announced domain seizures related to Chinese state-sponsored hacking campaigns that infiltrated US government systems.