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.
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.
AMD said the XDNA 2 NPU in Ryzen AI processors is the first neural processor to support Unsloth Desktop’s local interface. Compatible applications can send model requests to the Unsloth Desktop API, while Lemonade and FastFlowLM handle model serving and inference on the NPU and return responses to the applications. AMD also said more than 100 agents, including OpenClaw and Hermes, can run locally on Ryzen AI systems.
Microsoft announced a developer preview of USB4-based RDMA that lets two or more compatible Windows 11 PCs pool memory and GPU capacity to run a shared AI model through supported inference frameworks, without dedicated RDMA adapters or switches. Microsoft reported round-trip latency of about 28–50 microseconds for small transfers. In separate vendor tests, Microsoft used three AMD mini PCs to run a four-bit-quantized 235-billion-parameter Qwen3 model at about 18.5 tokens per second; AMD reported about 22 tokens per second for Qwen3-235B-A22B, while Intel reported 54.2 tokens per second for GPT-OSS-120B on four Core Ultra Series 3 PCs.
TUXEDO released the InfinityBook Pro 15 - Gen11 - AMD, a 15.3-inch laptop built around the AMD Ryzen AI 9 HX 470 processor. IT之家 reports a 2560×1600 300Hz display, 99Whr battery, expandable DDR5 and PCIe storage, and a starting price of about €1,500 before VAT and import duties; TUXEDO claims the cooling system can sustain 90W of performance.
TechNews reported that Taiwan's three major institutional investors net sold more than NT$46.1 billion as the Philadelphia Semiconductor Index fell 2.41% and TSMC's ADR dropped 2.90%. The decline was accompanied by an unconfirmed market rumor that Nvidia CEO Jensen Huang had been removed, along with concerns involving Supermicro's China litigation and AMD firmware exports. No official Nvidia clarification on the leadership rumor was available.
Hunterbrook reported that AMD TurinPI 1.0.0.B startup firmware was found in a BIOS update posted by IEIT Systems, a Chinese server maker added to the U.S. Entity List in March 2025. The investigation also identified nearly identical files from Maginfra, Kaytus and Aivres, which the report said were not on the Entity List. Hunterbrook did not establish how IEIT obtained the firmware or whether any license covered the transfers. The hedge fund disclosed a short position in AMD, while Wccftech presented its report as a summary of Hunterbrook's investigation.
AMD shares fell more than 7% after Hunterbrook Capital disclosed a short position and released a report alleging circular deals, excessive customer commitments and business ties to Chinese companies on a U.S. blacklist. Hunterbrook estimated $28 billion in credit support for partners and $48.8 billion in supplier commitments; it also alleged that AMD issued stock warrants to OpenAI and Meta. AMD disputed the report's characterization of its commercial practices and partnerships and said it complies with U.S. export-control law.
llm-d 1.0 adds multi-tenant resource management, request-priority scheduling, high availability, rolling upgrades and context-aware routing for agent workflows. An IBM Research benchmark deployed an approximately 753-billion-parameter model through llm-d, supporting up to 3,000 concurrent coding agents at peak output throughput above 6.6 million tokens per minute. Red Hat AI 3.5 also added governance capabilities, while IBM and Red Hat jointly disclosed that Lightwell Clearinghouse had reached general availability.
Anthropic priced its upsized IPO at $83 per share, above the marketed range of $70 to $78. The IPO and a private placement raised a combined record $107.3 billion. Seven companies, including Google, Nvidia and Microsoft, agreed to buy $26 billion of stock in the private placement at the IPO price. Shares were scheduled to begin trading on Nasdaq under the ticker ANT. Gray-market indications pointed to a roughly 20% to 27% opening gain, with estimates varying by venue and instrument.
Microsoft executive Judson Althoff said Nvidia’s leadership turmoil strengthened customers’ interest in AMD chips, cloud providers’ in-house processors and other alternatives, though most customers are still waiting for the situation to settle. Microsoft plans to launch Maia 300 in 2027 and, according to The Information as cited by TechNews, discussed reserving production capacity for more than 300,000 chips with TSMC; no completed reservation or measured customer adoption was confirmed.
World Labs’ Atlas world model can generate new views from one or more reference images after users specify a virtual camera position and angle, according to 36氪’s account of the vendor demonstration. The report says Atlas is available through a web application and developer interface, with film and visual effects identified as an application area; it also explores how world models could assist autonomous photography hardware.
AMD released Software: Adrenalin Edition 26.11.2 for Radeon desktop GPUs and platforms with Ryzen integrated graphics. The driver adds ROCm 10.2 support, support for Phantom Blade Zero and the Red Desert: Enhanced Edition - Unknown Journey DLC, and fixes for several stability and crash issues. AMD says the release improves overall driver stability and runtime reliability. Known issues include possible AI Bundle installation failures in regions where Hugging Face and GitHub are inaccessible.
OpenAI reported that its AI models accelerated chip-design work on its in-house Jalapeño processor and AMD flagship GPU designs. The company said some optimizations were incorporated into production-ready designs, while its research reported up to 14% improvements in power, performance and area for AMD’s GPU design.
Anthropic began trading on Nasdaq under ticker ANT on November 19, 2026, pricing 980 million shares at $83 in an IPO worth about $81.34 billion. The shares opened at $111.50 and closed at $97.17, up about 17%. Alongside the IPO, seven technology companies agreed to buy $26 billion of non-voting Class N shares: Nvidia, Google, AMD, Broadcom, Micron, Microsoft and TSMC. Reports put the combined transaction size at roughly $107.3 billion, while company net proceeds were reported at about $104.8 billion.
AMD Japan’s X account said the company “loves gamers” and values interaction with them above all else, drawing criticism and an X Community Note. Tom’s Hardware, IT之家 and TechNews reported that users contrasted the message with CEO Lisa Su’s “data-center-first” positioning and AMD’s Q3 fiscal 2026 results, which showed data-center revenue rising 97% to $8.546 billion while gaming revenue fell 40% to $773 million.
The Verge AI reports that Nvidia generated about $8,300 in profit per second, with quarterly profit reaching $65.3 billion as data-center and AI demand drove a sharp rise beyond earlier crypto and GPU booms. Following Jensen Huang’s resignation, interim CEO and CFO Colette Kress described operations as “business as usual,” said demand was accelerating and supply was growing, and projected $130 billion in quarterly revenue. Nvidia also plans to return all excess free cash flow net of strategic uses to shareholders, while its gross margin is expected to ease to 71–72 percent.
AMD and Dell executives said agentic AI workloads require substantial CPU capacity because they need rapid execution and high concurrency. AMD's Scott Tease also said attaching too many GPUs to one CPU can leave accelerators waiting for data, while Dell's David Schmidt emphasized rack power and cooling constraints. Dell's PowerEdge R9825, built on AMD's sixth-generation EPYC 9006 processors, was presented as an efficiency-focused response to those limits.
Nvidia data center CPU head Ian Finder told Tom’s Hardware that it was “absurd” to assume Nvidia foresaw the agentic wave when Vera was being developed. He said the processor instead grew from co-design with Nvidia accelerators, design intuition, and the goal of keeping accelerators supplied with compute, while early signs of tool calling informed Vera’s head-node requirements.
MINISFORUM has begun selling the 5-liter T5 Max-388 compact desktop through e-commerce channels in China. The Ryzen AI Max+ 388-based system includes 48GB of LPDDR5-6400 memory and a 1TB SSD, with a listed price of 9,699 yuan.
The Texas A&M University System is deploying VISION, an Nvidia DGX H200 SuperPOD, and planning IGNITE, an AMD-GPU platform that Dell Technologies will build at an Equinix data center in Dallas. VISION ranked No. 66 on the June 2026 Top500 list, while CIO Vince Kellen said the clusters could reduce research programs that normally take five months to one week. Texas A&M is also addressing employee vetting, credentials and security as it expands access for researchers across 12 universities and eight state agencies.
AMD announced the Instinct MI450X as an extension of its Instinct MI400 accelerator lineup, according to IT之家 AI. AMD says it retains the MI455X's core specifications and compute performance, with unchanged peak memory bandwidth, but reduces HBM4 capacity from 432GB to 288GB. AMD positions the accelerator as a lower-capacity option for workloads that fit within 288GB while meeting throughput and response-time goals.
Microsoft commercial CEO Judson Althoff said Nvidia’s leadership turmoil may encourage customers to seek AI chip alternatives, noting that customers wanted more choice even before Jensen Huang’s resignation. Althoff said Microsoft will continue working with Nvidia, while Microsoft, Amazon and Google are developing in-house chips and Anthropic and OpenAI are exploring alternatives; the report cited no specific deployments or orders.
Muddy Waters disclosed a short position in Nvidia, focusing on whether former Nvidia executive Ajay K. Puri and former AMD Chief Growth Officer Cenly Chen were connected to a GPU resale case. U.S. prosecutors allege that three AMD-related people used Southeast Asian shell companies to export Nvidia-GPU servers in violation of export controls, while a class-action complaint alleges Chen controlled relevant procurement and allocation decisions. Muddy Waters did not allege financial fraud or establish misconduct, and the report cautioned that neither person has been criminally charged and the allegations have not been judicially confirmed.
AMD CEO Lisa Su praised Nvidia as an excellent company and Jensen Huang as an outstanding industry leader at the TomorrowX summit in Austin. According to IT之家, cnBeta and other reports, she said AMD should distance itself from the surrounding “noise” and focus on its own business, while describing AI infrastructure and customer service as a “huge opportunity.”
AMD has completed its acquisition of Taalas, an AI inference chip startup that specializes in hardwiring models into workload-specific ASICs. IT之家 AI reports that AMD plans to integrate Taalas technology into its accelerator roadmap and develop system-level solutions combining the technology with Instinct GPUs. AMD says the deal strengthens its full-stack AI portfolio and long-term inference roadmap.
OpenAI's technical report "Advancing chip design with OpenAI models" says its models improved power, performance and area (PPA) by up to 14% on AMD's heavily optimized flagship GPU IP, a result it equates to roughly a year of engineering work done in under five months. AMD has already folded some model-generated SystemVerilog RTL changes into designs sent for manufacturing and into production design files, and confirmed single-digit PPA gains in key modules of its next flagship CPU. On OpenAI's own Jalapeño inference chip, model-suggested changes cut the size of two key blocks by about 13% and 10% and were validated in finished silicon, while RISC-V core evaluations placed GPT-6.2 Astra ahead of four other model generations. OpenAI also said it is partnering with Synopsys to train a new model, GPT-Synopsys, to operate Synopsys EDA tools the way expert chip designers do.
At SC26, AMD promoted the Instinct MI430X accelerator for supercomputing customers that require native FP64, contrasting it with Nvidia’s use of Ozaki-style FP64 emulation. AMD says fewer than 10% of its profiled HPC workloads use dense GEMM, while workloads such as sparse iterative linear algebra, hydrodynamics, Monte Carlo and particle transport still need strong native FP64 performance; these are vendor claims. Tom’s Hardware also reports planned MI430X deployments in the Discovery, Alice Recoque, LUMI-AI and Herder supercomputers.
The Information reports that AMD’s specialized AI chips were already attracting customers before Jensen Huang’s resignation as Nvidia CEO, with smaller chip rivals beginning to win conversions. Cirrascale CEO David Driggers said the cloud provider received significant customer interest last year in running AI inference workloads on AMD hardware, and that AMD chips are particularly suited to inference. The Information also reported that Huang’s departure is likely to reinforce the shift in customer sentiment, while Google was described in the headline as another rival chiping away at Nvidia.
Anthropic’s filing with the US Securities and Exchange Commission says Amazon will hold 14.7% and Google 13.7% of the company’s outstanding capital stock after its IPO and private placement, a combined 28.4%. Amazon could reach 23.8% if it converted all remaining convertible notes, but its disclosed post-IPO stake is 14.7%. AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC Partners have agreed to buy $26 billion of non-voting stock in the private placement. Anthropic has committed to spend more than $100 billion on cloud services, including AWS, over the next decade. Founder LLC, whose sole members are Anthropic’s seven co-founders, will hold the F-class share carrying 50.1% of voting power on ordinary shareholder matters.
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
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
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
Users need a local service that can expose Google AI Studio Playground and Build conversations through familiar API formats. The project addresses protocol differences across OpenAI, Anthropic, Gemini, agent clients, media generation, and other AI Studio capabilities while providing account and service management.
Key features
Four compatible API protocols — The service expo…
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
Managing remote systems often requires separate tools for shell access, file transfer, remote desktops, databases, containers, and monitoring. uniTerm brings these workflows together and adds an AI Agent that can plan and execute multi-turn shell commands in the terminal.
Key features
Terminal and remote access — Use remote, local, and serial terminals with password or key a…