Section
Infrastructure
The latest Infrastructure news, analysed daily by the i10X newsroom.
FilterNvidia
-
Mistral AI Raises $830M for NVIDIA Blackwell Data Center
Mistral AI secured $830 million in debt to build a 44MW data center near Paris with 18,000 NVIDIA Blackwell GPUs. Explore Europe’s push for sovereign AI compute capacity.
By
-
DeepSeek TileLang: Open-Source CUDA Alternative for Huawei Ascend
DeepSeek open-sources TileLang and compute libraries to program Huawei Ascend chips, challenging Nvidia CUDA dominance. Learn how this…
-
NVIDIA OpenShell: Secure Runtime for Autonomous AI Agents
NVIDIA OpenShell provides kernel-level sandboxing and YAML policies for safe AI agent execution. Discover how it protects enterprises…
-
NVIDIA OpenAI Deal: Direct 10GW AI Factories Bypassing Cloud
OpenAI now buys NVIDIA hardware directly for utility-scale AI infrastructure. Discover how this disintermediates hyperscalers and cements NVIDIA…
-
LLM Inference Optimization: vLLM, TGI & TensorRT-LLM
Discover how vLLM, Hugging Face TGI, and TensorRT-LLM boost LLM inference with PagedAttention and speculative decoding. Cut costs…
More stories
-
RTX 5090 Laptop GPU Beats M5 Max in Local LLM Inference
Early benchmarks show the RTX 5090 laptop GPU outperforming Apple’s M5 Max by up to 133% in LLM token generation. Discover how NVIDIA’s Blackwell architecture is…
-
Anthropic AMD: 2GW AI Compute Shift From Nvidia
Anthropic taps AMD for gigascale infrastructure with MI300X and future accelerators, validating ROCm over CUDA. Learn how this 2GW buildout signals the end of Nvidia’s hardware…
-
Agentic AI: How Autonomous Agents Reshape Infrastructure
Agentic AI moves beyond simple chat to autonomous loops that use tools and memory. See how this shift affects token use, latency, observability, and enterprise infrastructure…
-
AI Agent Simulation Infrastructure: Training Autonomous Systems
Discover how Meta, NVIDIA, and Google are investing in simulated worlds to train AI agents. Learn about the shift to dynamic environments and the need for…
-
NVIDIA GB300 in Azure: Anthropic’s Secure Enterprise AI Agents
Anthropic deploys frontier models on NVIDIA GB300 in Azure to power secure enterprise AI agents. This enables efficient, data-isolated inference for regulated industries. Explore the architecture…
-
AI Inference Hardware Shift: Custom Chips vs NVIDIA
The AI market is rapidly shifting from centralized training to cost-critical inference compute. Explore custom chips from Broadcom, AWS, Groq, and Google, plus software optimizations like…
-
Nvidia Vera CPU Rumor Highlights Need for Agentic AI CPUs
Reports of Nvidia shipping a Vera CPU to AI labs point to a bigger shift: agentic systems require tight CPU-GPU coordination beyond raw GPU power. Learn…
-
Full-Stack AI Infrastructure: AWS, Google, Databricks, NVIDIA
As AI scales, hyperscalers like AWS SageMaker, Google Vertex, Databricks MosaicML, and NVIDIA push rival full-stack ecosystems for foundation model training and deployment. Tackle distributed training…
-
Anthropic Trains Claude on Google TPUs: Why It Matters
Alphabet’s partnership with Anthropic deepens as Claude models train on Google Cloud’s TPU v5p, validating custom silicon for frontier AI. Explore impacts on Nvidia, cloud wars,…
-
OpenAI MRC: Fixing AI Training Slowdowns Partnership
OpenAI partners with Microsoft, NVIDIA, and AMD on the MRC initiative to combat slowdowns in massive AI training clusters. Standardizing diagnostics for better reliability, throughput, and…
-
OpenAI Rethinks Brute-Force Compute Strategy
OpenAI is shifting from aggressive GPU hoarding to flexible, efficient compute strategies amid rising costs. This analysis explores impacts on Microsoft, NVIDIA, rivals, and the AI…
-
NVIDIA PivotRL: 4x Efficient Training for Agentic AI
Discover NVIDIA’s PivotRL framework, a reinforcement learning breakthrough that cuts AI agent training costs by 4x through fewer simulations. Ideal for complex tasks like software engineering.…