❯ Nvidia board is not considering Huang’s return as CEO; Altman says AI industry will continue to grow
Board focuses on investigationNvidia founder Jensen Huang’s departure as CEO has prompted debate over whether he could return. CNBC cited a person familiar with the matter saying the board is not considering his reinstatement or discussing a return, and is focused on an investigation. Nvidia has hired law firm Wachtell Lipton to lead the inquiry.
Supporters call for returnHuang led Nvidia for 33 years, and his departure has drawn wide attention across the technology industry. The Financial Times previously reported that he had discussed a potential return with allies. CNBC said an online petition had gained support from technology figures including Elon Musk; Altimeter founder Brad Gerstner said Huang had not discussed a comeback with him.
Altman on the industryOpenAI CEO Sam Altman expressed personal empathy for Huang and acknowledged his role in AI’s rise. Bloomberg reported that Altman expects the industry to keep growing because it already has substantial revenue and demand. He did not take a position on whether Huang should return.
Leadership and customer commitmentsThe board is selecting a search firm to help find a successor. Nvidia supplies not only chips but also CUDA software, system designs and long-term product roadmaps, making continuity of supply, customer commitments and strategy important to companies building on its platform.
▮ SIGNALNvidia’s transition tests whether founder-led customer relationships and strategic decisions can remain effective under a new leadership structure.
❯ Altimeter seeks $5 billion of Anthropic IPO shares as allocation remains pending
A large orderAltimeter founder Brad Gerstner says the firm is seeking to invest $5 billion in Anthropic’s initial public offering, more than five times its previous largest IPO order. Vector notes that Altimeter is already an Anthropic investor; its eventual allocation depends on pricing and the distribution of shares.
Scale of the offeringAnthropic, the developer of Claude, and certain existing shareholders plan to sell 825 million shares at a proposed $70–$78 each, implying up to $64.4 billion at the top of the range. Its prospectus also describes a separate $26 billion private placement subject to closing conditions. Together they could total $90.4 billion; the IPO includes shares sold by existing holders.
Orders and pricingReuters cited sources saying orders were about six times the available shares by last Thursday. Demand may change during the roadshow, and final pricing, allocations and proceeds remain outstanding.
Growth expectationsClaude has grown quickly in coding and enterprise use, while competitors including Google continue to release new models. Investors are weighing whether Anthropic can extend its coding position into broader knowledge work and retain pricing power as alternatives improve.
▮ SIGNALThe large order reflects investor appetite for Anthropic’s growth, while sustaining its valuation will depend on extending Claude’s coding strength into broader enterprise work.
❯ OpenAI brings four outside groups inside to assess model safety
Access to internal systemsOpenAI has signed agreements with Apollo Research, METR, Transluce and Goodfire for a first phase of embedded safety assessments. Within agreed scopes, the groups will access selected internal models, training and evaluation data, monitoring systems and researchers to investigate model behavior and safeguards.
Different assignmentsApollo will examine whether models could covertly act against developers’ intentions and whether safeguards would detect it. METR plans to study misalignment and monitoring gaps in OpenAI’s largest reinforcement-learning run; Transluce will examine the quality and coverage of post-training tests, while Goodfire will use interpretability tools to inspect model internals.
Access and disclosureAssessors handling sensitive information will work on site using OpenAI equipment. They retain independence to publish, but OpenAI may redact confidential information and request responsible-disclosure delays. Apollo and Goodfire accept compensation; METR and Transluce do not accept assessment fees, though they may use API credits or computing resources.
Beyond public product testingThe program builds on OpenAI’s earlier third-party work. Deeper access may reveal problems unavailable to evaluators of public products, while requiring agreements on sensitive information, the scope of investigations and publication.
▮ SIGNALAI safety assessment is moving from testing public products toward examining training and internal use; its value depends on how much independent assessors can actually inspect.
❯ Alibaba launches Qwen4-Max with 2.4 trillion parameters and visual input
A new flagshipAlibaba has launched Qwen4-Max, a flagship model with 2.4 trillion total parameters, native vision capabilities and a one-million-token context window. It is designed for coding, agent tasks, workplace use and research.
Parameters used on demandThe model uses a mixture-of-experts architecture, activating only selected parts for each token. Alibaba says it activates about 71 billion parameters per token, 25% fewer than Qwen3.8-Max despite the same total parameter count, reducing the computing required to run it.
Performance and costOn Alibaba’s own Work360 benchmark, Qwen4-Max performed better overall than its predecessor at high reasoning effort while using 35% fewer tokens and reducing average task cost by about 60%. Real-world cost depends on the task and pricing. Independent evaluator Artificial Analysis also found a higher score and lower cost per task on its own index.
Developer accessQwen4-Max is available through QwenCloud’s API and subscription plan, as well as QwenWork and QwenStudio. Alibaba says smaller open-weight Qwen4 models will follow; they are separate products from the current flagship.
▮ SIGNALAs companies use AI for long documents and multi-step work, competition increasingly depends on the computing needed to complete each task.
❯ Baidu posts RMB 30.81 billion in quarterly revenue as AI cloud infrastructure grows 94%
Quarterly resultsBaidu reported third-quarter revenue of RMB 30.81 billion, about $4.6 billion, down roughly 1% year over year. Net income attributable to Baidu was RMB 1.4 billion. Revenue has declined for several consecutive quarters, while individual businesses are moving at different speeds.
Demand for computingAI cloud infrastructure revenue reached RMB 8.2 billion, up 94% year over year; GPU cloud revenue rose 191%. Baidu says its customer base is diversifying, prepayments are increasing and contracts are shifting toward longer terms. External customers’ token-use revenue on its Qianfan platform also grew sharply.
The AI revenue measureBaidu Core AI-powered business generated RMB 13.2 billion, or 53% of Baidu General Business revenue. The measure includes AI cloud infrastructure, applications and AI-native marketing, and its denominator is General Business rather than total group revenue. General Business revenue was RMB 25 billion, up 1%.
Businesses move at different speedsAI cloud growth helped offset pressure on older businesses. AI-native marketing revenue fell 15% to RMB 2.4 billion; Baidu cited competition and its decision to prioritize user experience during its AI search transition. AI application revenue was RMB 2.6 billion, approximately flat year over year.
▮ SIGNALBaidu’s AI growth is currently driven more by enterprise computing purchases, while monetization of search and applications continues to evolve.
❯ Apple and OpenEvidence launch health AI study for up to 200,000 US adults
Studying health AI useApple and clinical AI company OpenEvidence have launched the Apple Health AI Usage Study, seeking up to 200,000 eligible US adults. It will examine how people understand and trust AI health information and whether answers influence everyday health actions.
How the app respondsParticipants can ask health questions in a dedicated iPhone research app, read accessible answers and follow links to peer-reviewed clinical sources. OpenEvidence supplies the medical AI technology, drawing on sources including the New England Journal of Medicine, JAMA and Nature.
Participants control sharingUsers can choose whether to share HealthKit data such as resting heart rate, activity, sleep or laboratory results, allowing researchers to study whether personal information improves relevance. Apple says study data will not train AI models, and the app can suggest contacting a healthcare provider or other resources in certain situations.
Who can joinThe study is open to eligible US participants aged 21 or older with an Apple Intelligence-capable iPhone 15 Pro or later running iOS 26 or newer. Participants download the app and choose to enroll; the research also examines whether the tool helps people prepare for appointments with their doctors.
▮ SIGNALCombining personal health data with medical knowledge requires more than an answer: users also need to understand its basis and decide what to do next.
❯ SK hynix and South Korean government build chip test facility for suppliers
A production-grade test bedSK hynix and South Korea’s central and local governments are jointly investing in Trinity Fab, with a budget of 874.6 billion won, or about $655 million. Korea JoongAng Daily says the Yongin facility will offer suppliers of materials, parts and equipment conditions close to a commercial chip production line.
The validation problemA small supplier with a promising material or tool must show that it can perform reliably in an actual wafer fab. Building a test line is expensive, while using a customer’s production equipment can interrupt output. Trinity Fab plans to install production-type equipment and provide space for suppliers to install their own tools.
Beside a production lineThe facility is inside SK hynix’s Y1 fab in Yongin, adjacent to a future DRAM cleanroom. Similar temperature, cleanliness and utility conditions are intended to help transfer results to production. It remains under construction: reporters saw no equipment or piping installed, and Y1’s first cleanroom is scheduled to open in 2027.
Supporting suppliersReuters says SK hynix also outlined a five-year supplier support program spanning research, testing, production and sales. Growing demand for advanced memory in AI servers requires materials and equipment suppliers to validate and deliver alongside chipmakers.
▮ SIGNALChip expansion depends on materials and equipment suppliers too; production-like testing can help shorten the path into the supply chain.
❯ Duan Yongping’s H&H cuts Nvidia stake about 63% and adds PDD and Tesla
Nvidia position reducedH&H International Investment, associated with investor Duan Yongping, held about 2.319 million Nvidia shares on September 30, down around 63% from 6.281 million at the end of June, according to its third-quarter 13F. Holdings data puts the quarter-end stake at about $530 million, or 3.08% of the disclosed portfolio.
Largest holdingsThe 16 disclosed US holdings were worth about $17.19 billion, versus $19.10 billion at the end of the second quarter. Apple remained the largest at roughly $7.06 billion, followed by Berkshire Hathaway at $4.16 billion. Share counts in Apple and Alphabet fell approximately 21.8% and 41.5% respectively.
Other changesPDD shares rose about 11.9% to roughly 27.99 million, and Tesla shares rose 20.7% to 4.08 million. H&H initiated a position of about 199,900 SpaceX shares and no longer listed Alibaba, Circle or Innodata at quarter-end.
A quarter-end snapshotA 13F reports eligible holdings on a particular date, not the dates, prices or motives for individual trades; it also excludes some assets and short positions. Actual holdings may have changed by the time September’s snapshot was published.
▮ SIGNALDespite the Nvidia reduction, Apple and Berkshire still account for most of H&H’s disclosed portfolio.
❯ Intel and two executives face shareholder suit over PC slowdown disclosures
Shareholders file suitIntel, CEO Lip-Bu Tan and CFO David Zinsner face a proposed shareholder class action. Ars Technica says plaintiffs allege misleading statements or omissions about the impact of the PC slowdown and supply constraints between July 23 and October 29. They seek damages in an unspecified amount.
Shares fell after resultsIntel reported third-quarter results and a weaker-than-expected outlook on October 29. Revenue at its Client Computing and Physical AI Group fell 5% year over year; Tan cited a memory shortage, manufacturing constraints and softer PC demand. Shares fell 28.6% the following day.
Offering under scrutinyThe plaintiffs also point to Intel’s roughly $23 billion stock offering in August, alleging that business risks were not adequately disclosed when it raised funds. Intel’s quarterly filing reports approximately $22.6 billion in net proceeds after underwriting and other expenses.
An early-stage caseThis is a civil shareholder suit; the allegations are plaintiffs’ claims and have not been finally decided by a court. Intel is pursuing a chip-foundry transition while its PC business changes, keeping performance and disclosure in investor focus.
▮ SIGNALIntel is funding foundry expansion while navigating PC market volatility, increasing investor attention to disclosure of business changes.
❯ Apple’s 2027 iPhones may be split between spring and autumn; Air and anniversary plans remain unconfirmed
A possible change in cadenceApple’s future iPhone lineup may be spread across spring and autumn. MacRumors summarized Bloomberg reporter Mark Gurman’s earlier expectation that the standard iPhone 18, iPhone 18e and potentially a second-generation Air could arrive in spring 2027, followed by an anniversary model in autumn.
Trade-offs in a thin phoneThe current iPhone Air emphasizes a thin design, which also constrains room for batteries and cameras. The Information previously said some Apple engineers wanted a second rear camera and a larger battery for the next Air; Gurman considered an ultra-wide camera technically possible but questioned whether Apple would adopt it.
Anniversary and folding modelsGurman previously described a substantially redesigned 20th-anniversary model that could have a screen curved around its edges. Apple has launched the first folding iPhone Duo and is developing subsequent models, but final 2027 names, model count, specifications and dates have not been announced.
Plans can changeSome reporting on the split schedule and Air hardware dates to 2025, and product plans may have changed. The anniversary model’s name, the timing of a Duo 2 and an ultra-wide camera on the Air 2 remain unconfirmed.
▮ SIGNALA split schedule would create more upgrade windows for buyers and require Apple to coordinate sales across price tiers throughout the year.
❯ United resolves technology outage that delayed departures; no cyber link indicated
Departures affectedUnited Airlines experienced a technology issue on November 16 that delayed some departures. The airline later said the problem had been resolved, with no impact on flights already airborne or United Express flights, and said there was no indication of a cybersecurity connection.
Crew applications disruptedABC News cited the pilots’ union saying the outage affected several applications, including crew communications and messaging, as well as the company’s ability to provide final aircraft weights. Weight information is part of pre-flight preparation, so its absence can delay a departure.
Following a separate disruptionThe incident came days after a separate nationwide aviation ground stop caused widespread cancellations and delays. The disruptions involved different systems, and United has not disclosed the detailed technical cause of its outage.
▮ SIGNALAirline operations depend on crew communications, flight data and ground systems working together; missing information can prevent an otherwise ready aircraft from departing.