Trump Opens H200 Exports to China with 25% Surcharge
Trump reverses Biden export restrictions, allowing NVIDIA H200 sales to China with 25% surcharge. Blackwell GPUs remain restricted. Tariff-based controls replace bans.
Blog
Insights on GPU infrastructure, AI, and data centers.
Showing:
GPU Hardware & Chips
× Clear filter
The AI Memory Supercycle: How HBM Became AI's Most Critical Bottleneck
The AI memory supercycle has arrived. With HBM sold out through 2026, a $100 billion TAM by 2028, and gaming GPUs caught in the crossfire, memory has become AI's most critical infrastructure constrain...
DeepSeek mHC: The Architecture Fix That Could Unlock Trillion-Parameter AI Models
DeepSeek's new Manifold-Constrained Hyper-Connections framework solves a decade-old scaling problem, enabling stable training of 27B+ parameter models with just 6.7% overhead.
Trump Lets Nvidia Sell H200 Chips to China for 25% Revenue Cut
The Trump administration reversed export controls on Nvidia's H200 chips to China, demanding 25% of revenue. Here's what the policy shift means for AI infrastructure.
Trump Opens H200 Exports to China with 25% Surcharge
Trump reverses Biden export restrictions, allowing NVIDIA H200 sales to China with 25% surcharge. Blackwell GPUs remain restricted. Tariff-based controls replace bans.
First AI Chip Smuggling Conviction: $160M NVIDIA Pipeline Shut Down
DOJ shuts down $160M NVIDIA chip smuggling network to China. First AI diversion conviction. H100/H200 GPUs relabeled as 'SANDKYAN.' Operation Gatekeeper ongoing.
OpenAI-NVIDIA $100B Deal: 10 Gigawatts AI Infrastructure
OpenAI and NVIDIA announce a $100 billion partnership to deploy 10 gigawatts of AI infrastructure, with the Vera Rubin platform delivering eight exaflops starting in 2026.
Local LLM Hardware Guide 2025: Pricing & Specifications
Dual RTX 5090s match H100 performance for 70B models at 25% cost. Complete hardware pricing guide for local LLM deployment from consumer to enterprise GPUs.
NVIDIA's FP4 Inference Delivers 50x Efficiency
FP4 inference delivers 25-50x energy efficiency with 3.5x memory reduction. DeepSeek-R1 hits 250+ tokens/sec. The $0.02/token era arrives.
Why the NVIDIA GB300 NVL72 (Blackwell Ultra) Matters 🤔
NVIDIA's GB300 NVL72 delivers 1.5x more AI performance than GB200 with 72 Blackwell Ultra GPUs, 288 GB memory per GPU, and 130 TB/s NVLink bandwidth. Here's what deployment engineers need to know abou...
H100 vs. H200 vs. B200: Choosing the Right NVIDIA GPUs for Your AI Workload
NVIDIA's H100, H200, and B200 GPUs each serve different AI infrastructure needs—from the proven H100 workhorse to the memory-rich H200 and the groundbreaking B200. We break down real-world performance...
NVIDIA's Computex 2025 Revolution: Transforming Data Centers into AI Factories
NVIDIA CEO Jensen Huang's Computex 2025 keynote wasn't just another product launch—it was the blueprint for computing's third major revolution. As data centers transform into token-producing 'AI facto...