Amazon is transforming the site of a former 2-GW coal-fired power plant near Homer City, Pennsylvania, into a proposed 4.4-GW data-center campus with 36 buildings. The Information reports that the gas-fired site could power AI workloads for Anthropic or other customers and become one of the country’s largest AI power parks. The report contrasts Amazon’s ability to marshal capital and large-scale infrastructure with smaller cloud providers that have received financial support from Nvidia.
Tom’s Hardware reviewed the Lenovo Pro 9n, a premium Windows-on-Arm laptop powered by the 18-core Nvidia RTX Spark N1X configuration. The review covered productivity, gaming, its 15.3-inch 165 Hz OLED display, battery life and compatibility, finding strong ports and a vivid screen but also noting more than 10 hours of battery life, imperfect Arm compatibility and limited distinctive AI features. The reviewed configuration was priced at $3,249.99.
China's main A-share indexes fell on November 20, with the Shanghai Composite down 1.23%, the Shenzhen Component down 3.09%, and the ChiNext Index down 3.54%. Turnover reached 142.22 billion yuan, while banks and selected Huawei-compute, advanced-packaging, and semiconductor stocks bucked the broader decline, according to TMTPost. The analysis also highlighted valuation and performance risks across six limit-up stocks.
In an opinion article, Yang Jian frames AI-era leadership as four successive models: practical executors, individualists, strategists and alchemists. He argues that these models differ in decision boundaries rather than intelligence, using Apple, NVIDIA, Microsoft and other companies as illustrative cases. The article also proposes a path from standardizing work around teams and AI to converting personal insight into organizational capability and eventually defining cross-cycle values.
TechRadar highlights a paradox in AI infrastructure: memory is critically scarce while proposed data-center demand exceeds what can realistically be delivered. TrendForce reported that conventional DRAM contract prices rose roughly 93%–98% in the first quarter of 2026, while SK Hynix warned that high-bandwidth memory shortages could persist until 2030. The analysis says planners should distinguish announcements from binding contracts, secured power, financing, procured equipment and construction.
Business Insider, as reported by cnBeta, outlined five possible routes for Jensen Huang to challenge Nvidia’s board and seek another term as CEO. Nvidia’s board has said it has no intention of reinstating him, while Semafor reported that three directors are pushing for his return and a reinstatement petition has received more than 200,000 signatures. The report said his 3.58% stake and status as founder could support other approaches, but noted that his holding alone would not determine a shareholder vote and that the feasibility and outcome of each route remain highly uncertain.
新智元 cited an SEC filing showing that Jensen Huang and his wife transferred nearly 60 million NVIDIA shares to an irrevocable family trust in March, with the transfer value reported at more than $10.5 billion. The report also said Huang had resigned as NVIDIA CEO, but provided no evidence that he planned or founded a new chip company; Semafor reporter Rohan Goswami presented that scenario and a $50 billion startup fund as speculation.
SanDisk's Tokenomics 2030 outlook estimates that AI systems will produce about 150 quadrillion tokens in 2026 and potentially 10 quintillion annually by 2030, equivalent to roughly 317 billion tokens per second. TechRadar reports that the figures come from SanDisk's unpublished internal model and are intended only for illustrative purposes, making them a capacity estimate rather than a demand forecast; the company also promotes enterprise SSDs as infrastructure for supporting token growth.
The Financial Times reported that Alice Huang had ties to gaming firms, alleged Prince Group figures and hundreds of shell companies in a network moving Nvidia chips to China. It also described Huang as a Chinese executive in a romantic relationship with former Nvidia CEO Jensen Huang.
Amazon is nearing a $7.5 billion deal to buy Decart, an Israeli AI startup developing video models, the Financial Times reported. The talks are at an advanced stage, and a deal could come together as soon as next week. Decart is three years old.
California is investigating whether Nvidia illegally sold advanced chips to China and misled investors about that conduct. The probe is also examining potential criminal violations of state law.
Nvidia researchers argue that language models often do not need to process their entire context windows, which the team describes as slow, expensive, and prone to errors. According to The Decoder, their tests found that periodic summaries and targeted keyword search usually matched full-context results at a fraction of the cost, and the team proposed a new model class that works with short notes.
Anthropic ended its first week of trading at a reported $1.6 trillion fully diluted valuation, citing CNBC. Its shares closed Friday at $88.91, down 8.5% for the day and about half the initial 17% debut gain, but still above the $83 IPO price. The company closed an offering that raised $93.5 billion, with several major technology companies buying another $26 billion at the same price.
Anthropic’s record IPO did not reverse deteriorating Wall Street sentiment as investors weighed Nvidia turmoil, AI-spending concerns and selling by systematic funds, Bloomberg reported. The Goldman Sachs Panic Index reached 9 out of 10, while heavy positioning, weak liquidity and potential new stock issuance were cited as possible market pressures. The report provides a notable AI-market signal but relies on one Bloomberg reporter’s assessment.
Former Intel CEO Pat Gelsinger said at COSM Technology Summit 2026 that AI needs a 10,000-fold increase in computing efficiency, beyond routine generational chip improvements. He argued that reviving Moore’s Law requires breakthroughs in light, power, cooling, materials and packaging, while also identifying power infrastructure as a major constraint. The figures and technical path are Gelsinger’s predictions, not demonstrated results.
IT之家, citing Business Insider, relayed an account from Jamal, 43, a former NVIDIA director of AI application software product management who worked with Jensen Huang for about three and a half years, from 2022 until leaving in late 2025. He described Huang's frequent informal contact with employees, preference for group meetings that included seven or more people so they received the same information, and use of “Top 5 Things” emails to reduce reliance on management layers. The report says employees sent these emails weekly or monthly and that Huang read thousands of them, but it does not specify that all were weekly. Jamal also recalled Huang inspecting GTC demonstration booths for nearly 30 minutes on one occasion and sending a card and gift box after the birth of his daughter in early 2024. He said he never heard an employee complain about Huang and personally believed Huang would eventually return, though that was his intuition rather than an announcement.
TechRadar reviews the Surface Laptop Ultra, Microsoft's first laptop powered by Nvidia's RTX Spark chip. It is a premium work machine aimed at creative professionals and heavy AI workloads rather than general use. The base model has an Nvidia RTX Spark N1X chip, 24GB unified memory and a 512GB SSD, and starts at $2,599.99, £2,599 or AU$4,399. The reviewer considers 512GB limited for a costly device intended for large creative files. Against the 16-inch MacBook Pro with PROLED XDR Display (2026, M5 Max), the Surface Laptop Ultra scores lower in the reported Geekbench AI, PugetBench Adobe CC and 3DMark Steel Nomad comparisons. Its bright 4K mini LED display supports adaptive refresh up to 120Hz, and it includes a broad port selection and a relatively removable SSD.
Concerns over leading technology companies’ business models, uncertainty about the future of data centers and Nvidia’s troubles wiped out billions of dollars in market value, sending US stocks through one of their worst weeks since the Covid crisis.
36氪, citing unnamed U.S. media sources, and cnBeta, citing Reuters and CNBC, report that the SEC is investigating possible accounting issues at Nvidia. CNBC figures cited by cnBeta show Nvidia shares falling 14% on Friday, extending declines to seven consecutive trading days and seven consecutive weeks, with a reported cumulative drop of 46% and a market-cap loss of about $2.6 trillion.
NVIDIA Chair Harvey Jones said Jensen Huang voluntarily resigned after disclosing personal conduct inconsistent with company policy, and that the board unanimously accepted his resignation without plans to reinstate him. Huang separately acknowledged serious mistakes in his private life, apologized, and said he would be honored to return as CEO if the board deemed that best for shareholders, employees, and NVIDIA. The dispute emerged as NVIDIA reported third-quarter revenue of $115.8 billion, up 103% year over year.
Oracle says it has raised all the funding required to fulfill its infrastructure contracts with OpenAI, completing the final financing round earlier this month. The company also says the Abilene campus is nearing completion, the gigawatt-scale Shackelford campus has received all permits, and Nvidia Vera Rubin racks are expected within weeks. Oracle forecasts positive free cash flow in the second half of fiscal 2028 and more than $30 billion in fiscal 2029.
US technology stocks suffered a broad sell-off on November 20, with the Nasdaq Composite down 5.28%, the S&P 500 down 3.70%, and Nvidia falling about 14%, according to 36氪 and other reports. Anthropic closed down 8.5% at $88.91 on its second trading day, giving back most of its first-day gain despite a reported $119.5 billion raised through the IPO and concurrent private placement. Reports tied the pressure to questions about AI investment returns, Nvidia’s accounting review and higher interest rates, while noting that strong demand and financial results prevent the move from establishing an industrywide collapse.
Nvidia's board primarily accepted Jensen Huang's resignation because of his repeated lack of candor and alleged lies, according to people familiar with private board discussions cited by WIRED. Huang acknowledged an extramarital affair on November 8, when the board also launched an independent investigation, but sources said trust concerns—not the affair alone—drove the decision.
Reuters sources, as reported by Wccftech, say the SEC is investigating NVIDIA’s accounting practices. The report comes after NVIDIA agreed in 2022 to pay a $5.5 million SEC fine without admitting or denying findings related to undisclosed cryptocurrency revenue. Reuters noted that an investigation is not evidence of wrongdoing and may end without enforcement.
Reuters, citing two people briefed on the matter, reported that the US Securities and Exchange Commission is probing Nvidia's accounting practices. Nvidia changed how it reports revenue the prior year, removing Singapore from its geographical breakdowns. The supplied report does not explain the reason for the change, assess its validity or specify whether the probe concerns that change or other accounting practices. It also does not include confirmation of the probe from the SEC or Nvidia.
Sources cited by Wired said Nvidia's board primarily cited Jensen Huang's alleged "lying" and lack of candor when removing him over the Alice Huang affair. The report said the affair involved allegations of multiple relationships, including with a direct report and an alleged Chinese intelligence asset, but these more serious claims remain unconfirmed.
According to Kif Leswing of CNBC, Nvidia shares closed down 14% on Friday, marking their seventh consecutive decline. The stock fell 44% over seven weeks, reducing the chipmaker’s market capitalization by $2.5 trillion.
Nvidia’s stock plunged 14%, extending its weekly losing streak to the longest run since the 2008 financial crisis. Once the world’s most valuable company, Nvidia now ranks third behind Apple and Microsoft.
At SC26, DataDirect Networks introduced software-defined EXAScaler deployment, intended to give enterprise AI customers greater hardware flexibility when assembling AI infrastructure. The company also highlighted joint offerings with Super Micro Computer and said DDN’s Managed Lustre customers on Google Cloud achieved 70% better training performance and 40% higher GPU efficiency. SiliconANGLE AI reported that AI infrastructure sales cycles have shortened to as little as 30 days, although the performance figures and sales-cycle examples were presented by DDN executives.
An author tested 65 AI systems by asking each one the same questions ten times without a system prompt. The author reports that 73 of 100 book recommendations were Man's Search for Meaning, 76 of 100 business ideas were subscription boxes, 90% of study recommendations were computer science, and 98% of investment answers favored an index fund. Answers showed less convergence for movies, dinners, baby names, testimonial names and brand colors. The report also found that harmless questions could reveal model-family similarities, with 59% of models most closely matching another model from the same lab versus 4% by chance.
What problem it solves
AI agents need to choose among models while balancing task success, cost, and latency for their workloads. Applications also need flexibility in where routing runs and who controls model clients, credentials, retries, and service operation.
Key features
Multiple routing algorithms — Routing can be configured with Auto, Task, Execution, or Composite strategies, including mode…
What problem it solves
Serving large language models requires efficient memory management, flexible model execution, and practical integration with existing models and APIs. vLLM addresses these needs with inference and serving capabilities spanning batching, caching, quantization, distributed execution, and multiple hardware platforms.
Key features
Efficient memory and serving — vLLM manages atte…
What problem it solves
AI agents need product-specific instructions to use NVIDIA software reliably across physical AI, robotics, simulation, CUDA-X libraries, RAG, and platform workflows. The catalog centralizes these skills while keeping them maintained in their respective product repositories and synchronized through an automated pipeline.
Key features
Agent-discoverable skill catalog — The NVI…
What problem it solves
AI agent skills need to be checked for quality, security, duplication, and their actual effect on agent behavior. Static checks alone cannot establish whether guidance improves task results, so evaluation must also support live comparison with and without a skill.
Key features
Three-tier evaluation pipeline — Independent tiers validate skill safety and structure, detect over…
What problem it solves
Running large language and vision-language models can require substantial setup and may depend on the available hardware. The project aims to make inference accessible across a wide range of local and cloud hardware with minimal setup.
Key features
Minimal-setup inference — The project provides installation options for common platforms, including scripts, Docker, pre-built b…
What problem it solves
Serving language models requires reading checkpoints, tokenizing requests, and running accelerated kernels while managing a local model server. TensorFold addresses these requirements with a native binary that can run on Apple Silicon and NVIDIA GPUs without requiring Python or MLX for serving.
Key features
Native Metal and CUDA inference — The Zig-based engine executes mode…
What problem it solves
Interactive multimodal and speech systems often require generation to span heterogeneous stages with different compute patterns, dependency structures, and resource needs. Serving these pipelines requires coordinated stage lifecycles, inter-stage transport, model integrations, and a unified API surface.
Key features
Coordinated multi-stage runtime — Models generation as coor…
What problem it solves
Data analysis workflows often require expressive query composition, flexible execution models, and interoperability with columnar data. Polars addresses these needs by allowing the DataFrame API and SQL to share the same optimizer and engine, with support for large workloads and custom extensions.
Key features
Unified DataFrame and SQL Queries — Queries can be written with e…
What problem it solves
Running frontier-scale open-weight models locally requires an inference platform that can make efficient use of the heterogeneous hardware available on a personal device. The project addresses this by treating GPUs, CPUs, host memory, and interconnects as a unified, elastic inference platform.
Key features
Edge-native MoE runtime — The runtime provides MoE serving with bandw…
What problem it solves
Autonomous AI agents often need to read files, install packages, call APIs, and use credentials, but unrestricted access can expose host data, secrets, and networks. OpenShell provides these capabilities within an isolated runtime where declared policies constrain agent access.
Key features
Kernel-level runtime enforcement — Agents run in isolated sandboxes where kernel cont…
What problem it solves
AI agent skills execute with implicit trust and often receive minimal vetting before installation. SkillSpector helps users assess whether a skill contains vulnerabilities, malicious patterns, or other security risks before installing it.
Key features
Multi-format scanning — Scan agent skills supplied as Git repositories, URLs, zip files, directories, or individual files.
So…
What problem it solves
Mosaics obscure portions of videos and images, making it difficult to view the affected content without changing the rest of the picture. Users need a restoration workflow that can identify selected mosaic regions and process them with GPU-accelerated video handling.
Key features
Targeted mosaic restoration — Detects mosaics in videos and restores them while leaving the rest…
What problem it solves
Deep learning models often need to be adapted for efficient deployment across different inference frameworks. Model Optimizer addresses this by providing a unified set of optimization techniques and export workflows for quantized checkpoints.
Key features
Post-training quantization — Compress models with quantization techniques available for Hugging Face, Megatron-Bridge, Di…
What problem it solves
Platform teams need to create and operate isolated Kubernetes clusters for customers on infrastructure they supply and control. They need a management plane that can provision GPU machines, manage cluster lifecycle operations, and establish node identity before issuing join credentials.
Key features
Independent workload clusters — Create workload clusters through the NVKE AP…
What problem it solves
Running AI models locally can require integrating multiple inference engines, model formats, hardware accelerators, and API surfaces. LocalAI addresses this with a small core, on-demand backends, and compatible APIs that work across supported hardware without requiring a GPU.
Key features
Composable inference backends — Backends are packaged separately and pulled only when a…
What problem it solves
LLM and Visual Gen inference requires specialized kernels, efficient runtime orchestration, and flexible deployment across NVIDIA GPU configurations. Developers also need a modular framework that supports customization while integrating with established inference tooling.
Key features
Optimized inference kernels — Custom kernels accelerate common inference operations, includ…
What problem it solves
Rust developers need host-side access to the CUDA runtime and tools for expressing GPU kernels while preserving Rust's ownership and borrowing rules. The platform provides separate programming models for workloads that require direct thread-level control and workloads naturally expressed through tensor operations.
Key features
Shared CUDA runtime — Host-side crates provide d…
What problem it solves
Agents can retrieve documents and query databases, but unreliable results undermine agentic analytics. Teams need a context layer that records where data lives, what its definitions mean, whether it is usable, how it should be queried, its provenance, and related concepts.
Key features
Enterprise ontology generation — Generate semantic ontologies from database metadata, infe…