China May Let Alibaba and ByteDance Buy New Nvidia Chip
Beijing is reportedly weighing whether to let two of its biggest tech firms purchase Nvidia's newest workstation chip, according to The Information.
The Revision stories tagged nvidia — 72 stories on the wire, decoded into one clear voice.
Beijing is reportedly weighing whether to let two of its biggest tech firms purchase Nvidia's newest workstation chip, according to The Information.
Thieves swiped two Nvidia branded trailers hoping for AI GPUs but found 20,000 pounds of ballast sand in each, 40,000 pounds total.
Nvidia and Third Point are among investors in Nscale's $3.36 billion convertible raise, fueling its AI data center buildout ahead of a US IPO.
London AI infrastructure firm Nscale raised $3.36 billion from Nvidia, Apollo, Citadel and Abu Dhabi's investment arm ahead of a planned NYSE listing.
In a wide-ranging interview, the Nvidia CEO reframed AI's risks as his personal burden and urged people to adopt the technology quickly to benefit from it.
Elon Musk says SpaceXAI will grow its GPU fleet to 1.44 million by December 2026, using its own 1.2 gigawatt power plant to keep the chips running.
OmniDreams generates photorealistic driving video in real time, letting Nvidia test autonomous vehicle policies against scenarios no camera ever captured.
Jensen Huang says frontier labs pushing for AI regulation should just close experiments they can't control, not lobby Washington.
In a New York Times podcast interview, Nvidia's CEO argued that AI labs lacking control over their products have no business shipping them yet.
A Singapore-linked ByteDance subsidiary quietly became Nscale's biggest customer, exposing how neoclouds skirt US chip export controls without breaking laws.
AMD's stock is up 180% this year on AI chip demand, but its trillion-dollar valuation is still a fifth of Nvidia's.
Nvidia-backed Starcloud plans to mine Bitcoin from orbit this year, saying ASICs are 30 times cheaper to power in space than GPUs.
Huawei says it can't make enough Ascend AI chips to meet Chinese demand, so its next-generation Atlas clusters won't get a real international push.
A new investigative report traces how billions in banned Nvidia AI chips reach China through shell companies and Southeast Asian drop shipping.
NVIDIA open-sourced NeMo Data Designer, a config-driven tool for generating multimodal synthetic datasets with a preview-then-scale workflow.
Anonymous sources say the DOJ is probing whether Nvidia's Groq licensing deal was structured to dodge merger review, with no findings yet.
University of Toronto researchers built GPUThor, a Rowhammer attack that beats the ECC protection Nvidia told GPU owners to turn on last year.
University of Toronto researchers found a way to hammer Nvidia Ampere GPUs unevenly, cracking the ECC protection Nvidia has pushed for a year.
The US dangled access to Nvidia's AI chips to help broker peace between Armenia and Azerbaijan, turning silicon into a diplomatic bargaining chip.
Nscale is raising up to $3.5 billion ahead of a US listing, leaning on a $45 billion Anthropic contract that Microsoft and Google both passed on.
The chipmaker is spending $13 billion on the open-model hub it tried to invest in last year, tightening its grip on AI's supply chain.
The AI chip giant is acquiring the platform developers use to share open-source models, pulling a key piece of the AI supply chain in house.
Nvidia is buying into MediaTek's custom chip business, betting that even AI accelerators built by rivals will still need Nvidia's plumbing to work.
Nvidia's $3.5 billion investment in the Taiwanese chipmaker looks less like generosity and more like insurance against Big Tech's custom AI chip push.
Lambda borrowed $1B in private debt to buy Nvidia chips it will lease to Microsoft, the latest sign AI infrastructure runs on borrowed money.
A reported $12.9 billion deal would give Nvidia control of Hugging Face, the open-source hub millions of AI developers rely on.
The chipmaker is reportedly paying $12.9 billion for the AI model hub, deepening its grip on the industry it already powers with silicon.
The chipmaker projects $20 billion in Vera Rubin sales this quarter, calling it the fastest product ramp in its history, even as unit volumes stay modest.
Nvidia's quarterly revenue hit $96.2 billion, up 106% year over year, as it commits $279 billion to secure memory and chip supply for surging AI demand.
The chipmaker's reported $12.9 billion deal would pull the open-source AI hub in-house and put Nvidia back in the cloud business.
Nvidia has reportedly agreed to buy Hugging Face for $13B, putting the internet's biggest open-source AI hub under a chipmaker's roof.
Amazon is buying 2 million more Nvidia GPUs over the next two years, a bet that cloud AI demand still outpaces supply.
A Russian Molniya drone using an Nvidia Jetson Orin module picked its own target and killed three civilians in Zaporizhzhia, forensic investigators say.
Taiwan indicted nine people, including an Nvidia manager, for forging documents to hide illegal AI server exports to China.
Nvidia's third minority stake this month in a land-and-power broker signals it's racing to control data center infrastructure, not just chips.
Prosecutors allege a scheme routed 74 servers with B300 chips to China through Japan and Indonesia, and want five years for the Nvidia employee involved.
Taiwan approved 130 Nvidia B300 servers for a shell buyer, but only 74 were diverted toward China before Japanese customs blocked the rest.
Taiwan reportedly indicted Nvidia employees for allegedly shipping restricted AI servers to China, per a single unconfirmed report.
Nvidia has told its biggest customers that AI server prices will jump more than 15% starting early next year, citing rising memory costs.
Nvidia is reportedly discussing a partnership, investment, or acquisition with Korean AI inference chipmaker Rebellions.
ByteDance and Tencent got their first H200 shipments under tight quotas, a small slice of a huge order as domestic chips quietly take over China's AI market.
Nvidia says reports of a China specific LPU chip are wrong, but its narrow denial leaves room for standard LPU sales to Chinese buyers.
The Financial Times reports ByteDance and Tencent got about 10,000 Nvidia H200 chips each, as Beijing keeps most licensed volume in Hong Kong.
China reportedly approved 10,000 Nvidia H200 chips each for ByteDance and Tencent, a 20,000-chip total that signals a cautious loosening of export limits.
KernelArc, a multi-agent AI system, topped a GPU kernel benchmark leaderboard, suggesting parallel agent search beats single-agent optimization.
A new Nvidia benchmark shows robot foundation models struggle once tasks and scenes stop matching their training data.
Ukrainian intelligence says a Jetson Orin module powers targeting on Russia's new S-71 Monochrome missile, echoing an earlier Shahed drone finding.
In an open letter, Microsoft's Satya Nadella and Nvidia's Jensen Huang push back on a US crackdown on open-weight AI and jab closed rivals.
The Noetra consortium won a government contract worth up to $6.1 billion to build a 140MW AI facility targeting robotics and industrial automation.
A Japanese industrial consortium is deploying 27,500 Rubin GPUs and 13,750 Vera CPUs across 140 megawatts of data center capacity dedicated to physical AI.
Under pressure from Washington, Nvidia has slashed its roster of verified buyers in Asia, sending inspectors to data centers to weed out shell companies.
Nemotron-Labs-3-Puzzle-75B-A9B uses the Iterative Puzzle compression framework to double server throughput and multiply long-context concurrency eightfold.
LoRWeB composes a learned basis of LoRA modules at inference time to match visual transformations shown by example, outperforming single-LoRA approaches.
The CUDA-on-AMD translation layer drops commercial backing for the second time, but version 6 still lands 32-bit PhysX support and PyTorch fixes.
Five years after its debut, the RTX 3060 12GB is back at retail for $339.99 — almost exactly its original MSRP, and barely cheaper than newer cards.
Export controls have pushed black-market prices for Nvidia AI servers to almost three times their US retail cost, according to new reports.
Nvidia's CEO told shareholders national security beats any commercial opportunity, and that data centers built on smuggled chips are a dead end.
OpenAI's Broadcom-built Jalapeño chip targets inference workloads and signals a serious push to reduce the company's dependence on Nvidia.
With smuggling routes shut and legal imports blocked, Chinese buyers are paying triple for aging Nvidia AI hardware with nowhere else to turn.
Nvidia says its Rubin-generation liquid-cooled data center design eliminates almost all water consumption, though cost and construction impacts go unmentioned.
Nvidia's latest open-source world model handles language, image, video, audio, and robot actions in a single architecture, topping benchmarks for physical AI.
Nvidia's ENPIRE framework lets AI coding agents design and run robot training sessions without a human in the loop.
Nvidia showed robots that learned high-precision tasks like PC assembly on their own, raising real questions about automated hardware manufacturing.
Overwhelming demand let Nvidia upsize a $20 billion offering to $25 billion after orders topped $85 billion in a single afternoon.
A photo shows a fried ROG Equalizer connector, the cable Asus sold as the fix for Nvidia's long-running melting problem.
After export controls gutted its GPU sales in China, Nvidia is now pitching its Vera CPU, a product that may sidestep current restrictions.
The agreement gives SK Hynix a formal co-development role in the high-bandwidth memory powering Nvidia's next accelerators, not just a supply contract.
An engineering sample of Nvidia's never-released RTX 3050 Ti desktop GPU surfaced in leaked photos, revealing the Ampere card that quietly never shipped.
At Computex 2026, Microsoft designers explained the engineering choices behind putting a chip built for AI workstations into a laptop.
Jensen Huang confirmed two more RTX Spark generations are already in the pipeline, signaling to OEMs and developers that this platform is not a one-off.
Microsoft's new miniature developer PC runs Nvidia's Arm architecture, a quiet departure from its years-long Qualcomm partnership.
A compact Windows desktop powered by Nvidia's RTX Spark chip lets developers run AI models locally, without the cloud.