NVIDIA
GPU giant powering most AI training/inference. NIM microservices and Nemotron-3 30B open model expand the software stack.
GPU giant powering most AI training/inference. NIM microservices and Nemotron-3 30B open model expand the software stack.
All NVIDIA Plans(1)
NVIDIA History Timeline(30)
Nvidia quarterly lobbying spend hits record $2.2M
For July through September, centred on chip export rules.
SourceNVIDIA and Mistral establish industrial engineering AI partnership
Mistral models will plug into NVIDIA's CUDA-X, Warp and Omniverse stacks.
SourceTorchTPU and TorchNeuron erode Nvidia CUDA default ecosystem lock-in
TechCrunch reported Oct 20 that Google open-sourced TorchTPU letting PyTorch apps move to TPUs without rewriting, while AWS released TorchNeuron for Trainium a day earlier and contributed hardware-decoupling code to PyTorch; the report cites that about 85% of AI apps use PyTorch but offers no measured post-migration performance or cost data.
SourceNemotron 3.5 Super VL and Bolt 74B-A9B 24GB open-weight models
Oct 19 Nvidia VP Kari Briski released three open-weight agent models: 120B Super VL for enterprise data centers, 74B-A9B Bolt 24GB for RTX Spark/DGX Spark AI PCs, and Bolt E4B coming later this month; Skydance, EY, Ramp, CrowdStrike, Perplexity, Salesforce announced adoption
SourceNVIDIA open-sources post-trained multimodal Nemotron 3.5 Super VL
NVIDIA on October 15 open-sourced the post-trained multimodal Nemotron 3.5 Super VL, beating Claude Opus 5 and GPT-6.1 Sol on specialised benchmarks across medical imaging, radiology and industrial inspection, with a focus on enterprise inference and on-prem deployment.
SourceNemotron 3.5 Bolt 74B-A9B and Bolt E4B edge models open-sourced
Oct 15 NVIDIA simultaneously open-sourced two edge models alongside Nemotron 3.5 Super VL: Nemotron 3.5 Bolt 74B-A9B (24GB VRAM, balance of speed and quality) and Bolt E4B (4GB VRAM, for conversational agents)
SourceNVIDIA CEO Huang says he would cut chip supply to customers developing unsafe AI
NVIDIA CEO Jensen Huang said on October 14 that he would cut off any developer who uses NVIDIA hardware to build unsafe AI software, while stressing he would talk to customers before acting. Huang emphasised NVIDIA's technology is general-purpose and the company has limited visibility into how customers use it; he also said AI infrastructure shortages will last for years.
SourceNVIDIA releases cuPotential CUDA-X library for atomic-scale semiconductor simulation
NVIDIA released on October 14 the new CUDA-X library cuPotential, letting developers build AI agents that simulate how atoms and molecules interact; early-access testing accelerated already GPU-optimised workflows by an additional 2-3x. It is targeted at semiconductor process simulation.
SourceNVIDIA at OCP Global Summit releases Vera Rubin AI factory designs and USFP 400Gb/s optical standard
NVIDIA at OCP Global Summit on October 13 released Vera Rubin AI factory designs, grid resilience frameworks, and open-source DSX components, and with Cisco introduced the USFP 400Gb/s pluggable optical standard.
SourceNVIDIA invests in Doctolib and three other French AI firms, meets Macron
NVIDIA CEO Jensen Huang announced on October 13 during his Paris visit investments in Doctolib, Genesis AI, Pigment and Entalpic, four French AI companies spanning medical, robotics, enterprise software and materials; NVIDIA has not disclosed individual investment sizes. The same day he met French President Macron to discuss French AI investment, open models and AI safety.
SourceNemotron-3 achieves IOI 2026 and IMO 2026 gold
IOI 535.4/600 above gold threshold and top human; IMO 30/42 above 29 gold threshold
SourceNemotron-3-Ultra-CC open-sourced
550B total / 55B activated, SFT only, specializing in informatics olympiad
SourceNemotron-IMO-Bench released
200-problem new evaluation, with SFT/RL checkpoints and training datasets
SourceLambda raises up to $4B at a $14.5B pre-money, led by Blackstone and Coatue
Nvidia-backed Lambda is raising its final private round before a planned 2027 IPO; the figure climbed from $15B in June to $50B in September, largely on Anthropic's $35B commitment.
SourceNvidia launches Open Agent Safety Platform (OpenShell + Sentry)
Nvidia launched the Open Agent Safety Platform on September 28 with 100+ organizations. Core components include Apache 2.0 open source runtime OpenShell (on Vera CPUs) and hardware watchdog Sentry (on BlueField-4 DPUs, millisecond-level isolation of out-of-bounds agents). Ecosystem partners include Anthropic, Microsoft, Salesforce, SAP, SpaceXAI, Hugging Face, JPMorgan, Citi.
SourceJensen Huang publicly opposes blanket ban on US AI products going to China
NVIDIA CEO Jensen Huang on September 20 publicly opposed a blanket ban on US AI product sales to China, noting Vera Rubin is in volume production and years ahead of China while calling for enforcement of existing cybersecurity and product liability laws.
SourceNVIDIA, Google and Emerald AI launch AEMA
18 founding members, AEMA targets unlocking up to 100GW of grid capacity through flexible AI data centers.
SourceVera Rubin NVL72 debuts on MLPerf Inference v6.1
Vera Rubin NVL72 hits up to 3.7x on Qwen3-VL and 2.5x on DeepSeek-R1 vs GB300; MLPerf v6.1 adds end-to-end RAG and edge agent inference tests.
SourceNVIDIA CEO Huang rejects AI doom, says no new laws needed
NVIDIA CEO Jensen Huang on September 15 rejected AI doom predictions and said market forces are sufficient, calling for no new AI laws.
SourcePartnered with eight Australian providers for 2GW DSX AI factory buildout
NVIDIA announced in Melbourne partnerships with Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC, and AirTrunk for an AI factory buildout of up to 2GW by 2027. Sharon AI will deploy up to 68,000 GPUs.
SourceAnnounces $12.93B acquisition of Hugging Face
On Sept 3 NVIDIA announced an agreement to acquire open-model platform Hugging Face for about $12.93B, including roughly $11.9B in cash and up to $1B in employee equity incentives, pledging to keep the platform open.
SourceAnnounces $12.93B acquisition of Hugging Face
On Sept 3 NVIDIA announced an agreement to acquire open-model platform Hugging Face for about $12.93B, including roughly $11.9B in cash and up to $1B in employee equity incentives, pledging to keep the platform open.
SourceNVIDIA announces $12.93B acquisition of Hugging Face
On Sep 3 NVIDIA blog and Hugging Face jointly announced a $12.9303B cash acquisition, closing expected H1 2027. NVIDIA pledged Hugging Face remains an open neutral platform; three founders enter a 6-year retention agreement.
Source$3.5B investment in MediaTek for NVLink Fusion
Aug 31 joint statement, covers AI data center/PC/auto; part of $3.9B overseas convertible bond
SourceGroq 3 LPX inference accelerator enters mass production
Announced at Hot Chips 2026. Groq 3 LPX delivers 3,400 tok/s on Gemma 4 31B at 100K context. Nebius Token Factory among first adopters.
SourceSpectrum-X Multiplane and ScaleIn released
Spectrum-X Multiplane scales to 512,000 GPUs per layer; ScaleIn accelerates infrastructure services on BlueField-4.
SourceSpaceXAI adopts Vera CPU and extends Vera Rubin to Starmind satellites
SpaceXAI becomes the first customer to deploy an optimized Vera Rubin NVL72 in orbit.
SourceIn talks to invest in Perplexity at $30B+ valuation
In talks to invest in Perplexity at $30B+ valuation
SourceGroq 3 LPX inference accelerator mass production
Groq 3 LPX inference accelerator mass production
SourceNvidia pays $6B for Poolside AI license, adds $1B investment, 109 job offers
Nvidia signed a $6B non-exclusive license with Poolside plus $1B strategic investment at a $12B pre-money valuation; 109 Poolside staff received job offers. Poolside stays independent. The $6B license proceeds are expected to be distributed to existing investors by late next year.
Source