Index Company
Nvidia
39 CÈ pieces since 3 June 2026: 3 deals, 30 intel items, 6 essays. Newest first. All companies →
-
Washington banned the chips. Chinese labs rent them in Thailand instead.
Washington will license the wire rather than the chip, and Commerce will do it by rule before the Senate ever votes.
-
The base is free, the moat is not: Harvey post-trains Kimi K3 for law
Harvey and Fireworks AI ran asynchronous reinforcement learning on Moonshot's open Kimi K3 across roughly 150 NVIDIA B300 GPUs for two months. Tenet nearly doubles completed tasks on Harvey's Legal Agent Benchmark. No weights, no API.
-
China's chips handle cheap inference; coding still needs Nvidia, and tokens hit 140 trillion a day
Chinese inference firms run low-value workloads on domestic accelerators and reserve scarce Nvidia chips for coding and agent tasks, where quality pays. Daily token calls passed 140 trillion in March, up more than a thousandfold since early 2024.
-
Vietnam runs 15,000 GPUs while Seoul plans 250,000: NVIDIA offers a national LLM blueprint
NVIDIA's ASEAN vice-president proposed four cooperation tracks to Vietnam's science ministry, including a Vietnamese large language model built on the open Nemotron family and a national computing-capacity plan.
-
Sovereignty costs 50% more engineering time; domestic chips still take 56% of China's AI servers
Domestic vendors are forecast to take 56 percent of China's AI server chip market in 2026, up from 46 percent, while foreign share falls to 21 percent from 34. Meituan trained LongCat-2.0 on a 50,000-chip domestic cluster.
-
Python was the deployment tax: NVIDIA's TensorRT Model Connect ships 76 model families to C++ under Apache-2.0
NVIDIA released TensorRT Model Connect under Apache-2.0, taking a Hugging Face checkpoint to native C++ inference in two commands with no ONNX step and no PyTorch at runtime. It covers 105 profiles across 76 model families.
-
Ooredoo bought 49 percent of Zankore for US$800 million. Indonesia supplies the power.
Ooredoo, Nokia and NVIDIA put a gigawatt of AI compute into Indonesia, and the terms of the deal say plainly who collects on it.
-
Who owns the GPU hours behind sovereign AI? Indonesia opens UGM's centre with Indosat and NVIDIA
Indonesia inaugurated its first university-based AI research centre at Universitas Gadjah Mada, running on NVIDIA's stack and GPU Merdeka, Indosat's sovereign GPU-as-a-service platform. First projects cover tuberculosis screening, tropical agriculture, and disaster geospatial work.
-
Routing beats buying the biggest model: NVIDIA opens Switchyard at a third of Opus 4.8's cost
NVIDIA released Nemotron 3.5 Lightning, a 30B mixture-of-experts model with published training data, alongside NeMo Switchyard, an open routing library that sends each agent step to the cheapest capable model. Ramp cut costs 58 percent and runtime 33 percent.
-
Singapore opened 200MW to new data centres. One Batam campus contracted 450.
On apatheia, and the compute money landing across the strait from a business it is never going to touch.
-
Indonesia stops renting compute and starts owning it: Indosat launches Zankore at 1GW
Indosat Ooredoo Hutchison, Ooredoo Group, NVIDIA and Nokia launched Zankore, targeting up to 1GW of AI-factory capacity. Phase one delivers roughly 200MW on GB300 NVL72 systems in the first half of 2027.
-
$2B at a $10.5B valuation, and some of it is earmarked for Indonesia: Firmus raises
Blackstone Tactical Opportunities, Coatue, NVIDIA and Jane Street backed the Australian AI-factory builder, which keeps a Singapore office. The capital funds Project Southgate and early development in Indonesia.
-
A fully open MoE trained without a single Nvidia GPU: AMD ships Instella-MoE-16B-A3B
AMD released Instella-MoE-16B-A3B, a 16-billion-parameter mixture-of-experts model with 2.8 billion active per token, pretrained on 7.1 trillion tokens across Instinct MI300X and MI325X GPUs, with the training code under MIT.
-
A breach turned AI security into shared plumbing: NVIDIA, Microsoft and IBM open-source their guardrails
The Open Secure AI Alliance launched with more than 60 partners. NVIDIA put its NOOA agent-safety framework on GitHub, Hugging Face offered Safetensors to the PyTorch Foundation, and Microsoft contributed MDASH, its agentic bug-hunting harness.
-
Can a frontier lab run without Nvidia? Z.AI fires up a 1GW all-Chinese data centre
The Chinese AI company completed a gigawatt facility built entirely on domestic semiconductors, drawing power equivalent to 750,000 homes, and reports $1 billion in annual recurring revenue after hitting its 2026 sales target in July.
-
AI factories become a sovereign asset class: Naver lands $10B from Nvidia and Brookfield
Naver founder Lee Hae-jin said Nvidia and Brookfield will invest $10 billion, some 14.6 trillion won, into the company's AI factory buildout, with Nvidia's $1 billion arriving alongside Brookfield's larger infrastructure commitment.
-
Beijing Will Not Wall Off Its AI (Even Now That It Leads)
Days after an open-weight Chinese model took the top of the coding charts, Beijing began consulting its champions on export-controlling AI, the same weapon it called economic bullying when Washington aimed it the other way.
-
China launched a 29-nation AI governance body in Shanghai; it does not own the frontier it wants to govern
China launched WAICO in Shanghai this week with twenty-nine mostly Global-South members and a mandate to govern artificial intelligence. The 1974 New International Economic Order is the rhyme: a numerically large coalition can pass a declaration and still fail to bind an order, because standards follow whoever holds the frontier, and this coalition does not yet hold it.
-
The memory shortage just minted a war chest: SK Hynix raises $26.5 billion on Nasdaq to lock in its HBM lead
SK Hynix debuted on Nasdaq under SKHY on July 10, raising about $26.5 billion in the second-largest US listing after SpaceX. The memory maker holds 58% of the high-bandwidth memory market feeding Nvidia's next-generation AI accelerators.
-
A frontier open model without a single Nvidia chip: Meituan releases the 1.6T LongCat-2.0 under MIT
Meituan open-sourced LongCat-2.0 under an MIT license: a 1.6-trillion-parameter mixture-of-experts model activating roughly 48 billion parameters per token, with a native one-million-token context, scoring 59.5 on SWE-bench Pro and trained end-to-end on a 50,000-card cluster of domestic Chinese AI chips.
-
With Nvidia's China share collapsing and Huawei near 50%, DeepSeek designs its own AI chip
DeepSeek is designing an in-house AI chip to cut its reliance on Nvidia and Huawei silicon, joining Alibaba and Baidu as US export controls erode Nvidia's China position and lift Huawei toward half the market.
-
Tensordyne exits stealth with a logarithmic-math inference chip in TSMC production
The startup rearchitects the math inside the silicon. Its 72-chip inference pod fits in 13 rack units and draws 30 kilowatts against 150 for a comparable Nvidia system, with the first chip already in production at TSMC.
-
NVIDIA and Hugging Face bring open humanoid-robot models to LeRobot
The two released Isaac GR00T 1.7, an open vision-language-action model for humanoids, plus the Isaac Teleop data-collection framework, into LeRobot, connecting NVIDIA's three million robotics developers with Hugging Face's sixteen million AI builders.
-
Southeast Asia now holds 2,000 data centres. It captures almost none of the margin.
The land, power and water are local; the models, the margin and the IP are almost entirely foreign, which makes the operator question simple: own a customer, not a GPU.
-
NAVER scales sovereign AI to gigawatts with NVIDIA's DSX platform
NAVER will expand its GAK Sejong data center from 55 megawatts toward gigawatt scale on NVIDIA's DSX blueprint, powering next-generation HyperCLOVA X models and agentic services across Asia.
-
Upscale AI raises $190M with Temasek aboard to wire AI clusters
Premji Invest led the Series A-1; Nvidia, Salesforce Ventures and Temasek joined, lifting the cluster-networking startup to a $2B valuation and $500M raised in 18 months.
-
NVIDIA RTX Spark Superchip delivers one petaflop to consumer workstations
NVIDIA's RTX Spark Superchip combines CPU and GPU with up to 128 GB unified memory, delivering one petaflop of AI compute capable of running local models up to 120 billion parameters with 1-million-token context windows.
-
NVIDIA ships RTX Spark Superchip with 128 GB unified memory for local agents
Unveiled at Computex 2026, the RTX Spark combines Arm CPU and Blackwell GPU in a Windows-on-Arm platform capable of running local models up to 120 billion parameters with 1-million-token context windows. Jensen Huang positioned it as meter-free agent infrastructure.
-
China drafts $295 billion AI data-center plan mandating 80% domestic chip sourcing
National Development and Reform Commission blueprint calls for networked computing hubs by 2028, with state telecoms China Mobile and China Telecom operating facilities powered predominantly by Huawei, Alibaba, and Biren chips instead of Nvidia or AMD.
-
Zyphra ships ZAYA1-8B sparse reasoning model trained end-to-end on AMD Instinct hardware
Released under Apache 2.0 license, the 8-billion-parameter model routes only 760 million active parameters per token and was built entirely on AMD GPUs, proving a viable non-Nvidia path for high-efficiency training and breaking the assumption that frontier open models require Nvidia infrastructure.
-
NVIDIA RTX Spark Superchip ships 128GB unified memory, one petaflop local AI compute
Announced at Computex June 2026, the RTX Spark Superchip pairs a 20-core Grace CPU with a Blackwell RTX GPU delivering up to 128GB unified memory and one petaflop of AI compute, enabling local execution of 120-billion-parameter models with million-token context windows.
-
Zyphra ZAYA1-8B sparse model trained on AMD hardware, ships under Apache 2.0
Zyphra released ZAYA1-8B in June 2026 under Apache 2.0, an 8-billion-parameter sparse-routing model with only 760 million active parameters per token, trained from scratch on AMD Instinct MI300 GPUs to demonstrate inference efficiency outside Nvidia-dependent pipelines.
-
NVIDIA RTX Spark Superchip delivers one-petaflop AI compute to consumer laptops
The RTX Spark Superchip combines CPU and GPU with up to 128 GB unified memory, running local models up to 120 billion parameters with million-token context windows on workstation laptops.
-
Zyphra ships ZAYA1-8B reasoning model trained entirely on AMD Instinct MI300X GPUs
Released May 6 under Apache 2.0 license, the 8-billion-parameter mixture-of-experts model was pretrained, midtrained, and fine-tuned on a 1,024-GPU AMD cluster with zero Nvidia hardware, matching far larger proprietary models on math and coding benchmarks.
-
China announces $295 billion five-year plan to build nationwide AI data centers
Beijing is preparing to spend 2 trillion yuan over five years on interconnected computing hubs operated by China Mobile and China Telecom, with 80 percent of AI chips sourced from domestic suppliers including Huawei. The network is designed to fuel domestic AI sector growth and reduce dependence on Nvidia and AMD.
-
Microsoft releases Coreutils for Windows, NVIDIA RTX Spark Superchip, and OpenClaw native support
At Build 2026, Microsoft shipped Coreutils for Windows (a Rust-based reimplementation of GNU Coreutils from the uutils open-source project), announced the Surface RTX Spark Dev Box with NVIDIA RTX Spark superchip delivering one petaflop of AI compute and 128 GB unified memory, and confirmed OpenClaw runs natively on Windows with MXC containment.
-
Zyphra releases ZAYA1-8B sparse model trained from scratch on AMD Instinct hardware under Apache 2.0
Zyphra's ZAYA1-8B model uses a sparse routing architecture with 8 billion total parameters but only 760 million active per token. Trained entirely on AMD Instinct hardware without NVIDIA dependency, it ships under Apache 2.0 license, demonstrating decentralised training infrastructure.
-
Washington closes one-year loophole letting Chinese firms route banned Nvidia chips through Singapore and Malaysia
The U.S. Bureau of Industry and Security issued guidance May 31 closing a loophole that let Chinese-owned subsidiaries purchase high-end Nvidia GPUs via Singapore and Malaysian shell units. Investigations confirmed nearly US$2 billion in chips flowed to Indonesia and Malaysia data centers remotely serving Chinese clients.
-
Nvidia RTX Spark superchip puts 120B-parameter models on Windows laptops this fall
At Computex, Nvidia unveiled RTX Spark: 20 Arm cores, Blackwell GPU, 128GB unified memory, running 120B models with million-token context on-device. Over 30 laptops and 10 desktops ship Q4 2026, enabling local inference without cloud API calls.