/Infrastructure
Nvidia Updates Data‑Center Roadmap After GTC 2026 LPU Launch
Following its Language Processing Unit reveal at GTC 2026, Nvidia expanded its data‑center product roadmap to include new GPU and LPU generations through 2028.
← Momentum 
Company momentum
NVIDIA Corporation
30 Active
Latest coverage
/AI
NVIDIA Releases Nemotron‑Labs‑TwoTower: an Open‑Weight Diffusion Language Model
NVIDIA unveiled a new diffusion language model that retains almost all autoregressive quality while boosting throughput, marking a significant step in efficient large‑language‑model inference.
/Technology
Claude models go live on Microsoft Foundry via Azure
Anthropic’s Claude models are now generally available in Microsoft Foundry, running on NVIDIA GB300 Blackwell Ultra GPUs hosted on Microsoft Azure, expanding infrastructure options for organisations building AI agents.
/AI
Recursive Self‑Improvement Now Has a Co‑Evolving Evaluator: Cambridge‑NVIDIA Paper Raises the Stakes
A new preprint shows how an AI system’s evaluator can evolve alongside the agent it grades, potentially accelerating self‑improving models and raising fresh safety concerns.
/Infrastructure
SpaceX inks compute deal with Reflection AI, an open‑source AI lab
SpaceX will supply Reflection AI with $150 million/month of Nvidia GB300 chips, a $6.3 billion commitment that underlines the growing importance of open‑weight AI infrastructure.
/AI
NVIDIA Announces BioNeMo Agent Toolkit to Accelerate Scientific Discovery
NVIDIA unveiled the BioNeMo Agent Toolkit, a suite of agent‑ready tools that let AI agents run biology, chemistry, genomics and drug‑discovery workflows, promising faster, more accurate scientific research.
/AI
AI chipmaker Groq confirms $650M raise, re‑staffs after Nvidia’s $20B not‑acqui‑hire deal
Groq secured a $650 million funding round and is hiring to pivot its inference cloud after Nvidia’s licensing deal stole its key talent.
/Infrastructure
Dell launches AI server based on Nvidia Vera Rubin GPUs
Dell unveiled the PowerEdge XE8812, a high‑end AI server powered by Nvidia’s new Vera Rubin NVL4 architecture, aimed at enterprise AI factories.
/AI
AI coding agents can autonomously direct robot training
Nvidia’s GEAR lab unveiled a new harness that lets AI coding agents teach robots to install GPUs and cut zip ties, achieving a 99% success rate on complex manipulation tasks.
/AI
Recursive Superintelligence emerges from stealth with $650M raise
London‑based AI startup Recursive Superintelligence raises $650 million, valuing it at $4.65 billion, with backing from GV, Nvidia and AMD.
/AI
The $14 Trillion Warning: AI’s Riches Will Be a “Few and the Rest” Story
Larry Fink, CEO of BlackRock, warns the AI boom risks worsening inequality, with rewards flowing only to a handful of firms and investors. In his annual letter, he noted that AI threatens to repeat—and amplify—a pattern where wealth accrues to those who already own assets. His solution: more people should invest in capital markets rather than rely on homeownership. The warning comes as BlackRock manages $14tn and Fink’s own $30.8m pay package has drawn shareholder scrutiny.
/AI
Iran Is Targeting Data Centers. Silicon Valley Should Be Worried
Three drone strikes. Two data centers offline. One very expensive warning to Silicon Valley: the Gulf is no longer safe ground for the infrastructure of the future.
/Business
Nvidia and UK Wealth Fund Throw Weight Behind British Self-Driving Pioneer Oxa
Nvidia has teamed up with the UK's National Wealth Fund to invest $103 million in British autonomous driving startup Oxa. The Oxford-based company, which develops software for self-driving industrial vehicles in ports and factories, has pivoted away from the complex passenger car market to focus on automating repetitive goods movement—a shift that co-founder Paul Newman says makes the "economics...cracking."
/Business
Meta Bets Big on AMD in $60bn AI Chip Deal
Meta has agreed to a sweeping $60bn, five-year deal with Advanced Micro Devices, acquiring chips and a 10% stake in the company as it doubles down on AI — the latest sign that the industry's hunger for processing power is only growing stronger.
/AI
NVIDIA Blackwell Ultra Cuts AI Inference Costs Up to 35×, Boosts Agentic AI Performance by 50×
New performance data shows that NVIDIA’s next-generation Blackwell Ultra platform delivers up to 50× higher throughput and up to 35× lower cost per token for agentic AI workloads compared with the previous Hopper generation — enabling significantly faster, more economically scalable real-time coding assistants and autonomous AI agents.