Edge AI Daily Morning Brief: Jensen Huang Resigns as NVIDIA CEO, Apple Rebuilds Siri with App Intents
NVIDIA announced that Jensen Huang has immediately resigned from his roles as President, CEO, and Director; CFO Colette Kress will serve as interim CEO, former Synopsys CEO Harvey Jones will serve as the first Chairman of the Board, and a search for a permanent CEO has been launched. The key to Apple's rebuild of Siri is using App Intents to turn third-party app capabilities into invocable, authorizable structured actions, in order to adapt to screenless smart glasses. OpenAI has disbanded its standalone robotics team and merged it into its core research organization, saying that general frontier models can already be used directly for robot control.
Topics
Body · AI translation
Silicon Valley Frontier:
1. [Background and Arrangement of the Abrupt Personnel Change] On November 12, 2026, Nvidia announced that Jensen Huang would immediately resign as President, Chief Executive Officer, and Director; Chief Financial Officer Colette Kress would serve as Interim CEO, Harvey C. Jones would serve as the first Chairman of the Board, and a search for a permanent CEO would be launched. This transfer of power occurred rapidly after the board investigation was initiated; it was not a long-anticipated smooth handover, and the market shock came mainly from "expectations" rather than from the production lines.
2. [Key Data and Scope of Impact] Nvidia's 2026 annual report shows that data center business revenue was approximately $194 billion, up 68% year over year; networking business revenue exceeded $31 billion; and major cloud providers and hyperscale data centers had deployed close to 9 gigawatts of Blackwell AI factory capacity. Precisely because of this enormous scale, the CEO transition will affect not only one chip company, but also judgments about cloud providers' capacity expansion, model training schedules, and data center construction.
3. [Core Risks and Industry Direction] The most critical risk is whether the "circular economy" will slow down: Nvidia both sells infrastructure and acts as an organizer of AI capital and ecosystem. If the new management, under governance pressure, shifts toward more cautious capital allocation, the pace of investment and cooperation may slow. The industry may move from "unconditional expansion" to "case-by-case verification." In the short term, Nvidia faces a trust discount; in the long term, it may accelerate toward a phase with multiple suppliers, multiple architectures, and greater emphasis on cash flow.
II. Apple's Siri Wants Smart Glasses to Truly Get Things Done
1. Core judgment: The key to Apple's remake of Siri is not speech recognition accuracy, but using App Intents to restructure the relationship between Siri and third-party apps—turning app capabilities into structured actions that the system can discover, pass parameters to, invoke, and authorize, making Siri an "operating system transaction layer" between user intent and app execution, with the interaction unit upgraded from a single question-and-answer to cross-app task chains.
2. Architectural focus: This design is particularly suited to screenless smart glasses. The microphone and speaker are responsible only for input and feedback; what truly determines product value is whether the middle layer can reliably translate "what I want to do" into "which app did what." However, glasses will amplify latency, misrecognition, and privacy issues, so low-risk actions should be executed directly, high-risk actions should require confirmation before execution, and if confirmation is impossible, they should be refused.
3. Implementation constraints: According to reports, Apple had delayed Siri features originally scheduled for launch in spring 2025 that could take actions in other apps, and had previously suspended some Apple Intelligence news summary features. Judging the success or failure of Siri AI should depend on metrics such as the number of top apps integrated with App Intents, the first-attempt success rate of cross-app tasks, and the misoperation rate. Developers' willingness to integrate, together with privacy governance combining on-device processing and Private Cloud Compute, jointly constitute the adoption threshold.
III. OpenAI Breaking Up Its Robotics Team Does Not Mean Abandoning Robotics
1. OpenAI has disbanded its independent robotics team: According to a Fortune report on November 13, the robotics research team was merged into the core research organization and reports to Chief Research Officer Mark Chen, while the applied robotics and hardware teams moved into the Applied department under President Greg Brockman. The company emphasized that the reorganization is meant to accelerate robotics progress rather than exit the field, and that robotics remains part of its long-term vision.
2. This is a correction after repeated shifts in direction: OpenAI set up a robotics team early on, consolidated it in 2020, and then canceled the independent unit again about five months after rebuilding it this year. According to Fortune, the company judged that general frontier models, including GPT-6 Astra, can already be used directly for robot control; the boundary between robotics research and general model research is thinning, and the need for an independent team has declined.
3. Impact and boundaries: General models may become robots' "digital brains," and model companies are moving upstream (perception, reasoning, planning), which may compress the specialized model layer; however, touch, friction, real-time control, real-world data collection, and safety validation remain hard thresholds. The company said that since February its hardware team has used Codex and ChatGPT to assist with CAD and simulation, and model usage measured in tokens has grown 23-fold, but this reflects changes in engineering design usage, not an improvement in robot performance.
IV. After Jensen Huang, Nvidia Faces an Institutional Power Test
1. Event and governance arrangement: On November 12, 2026, Nvidia announced that Jensen Huang would resign as the company's President, CEO, and Director, effective immediately; the board appointed Executive Vice President and CFO Colette Kress, who joined in 2013, as Interim CEO (while remaining CFO), and appointed former Synopsys CEO Harvey Jones as the first Chairman of the Board, while launching a search for a permanent CEO covering both internal and external candidates, with business continuity as the overall priority.
2. Financial and growth background: Nvidia's fiscal 2026 revenue was $215.938 billion, up 65% year over year, with data center revenue of approximately $194 billion, up 68% year over year; in the second quarter of fiscal 2027 (ended July 26, 2026), revenue was $96.221 billion, up 106% year over year, with data center revenue of $89 billion, up 117% year over year. Strong growth provides a buffer for the transition period, but it is also why the market views the succession issue as an operating variable rather than a purely personnel matter.
3. Key points of analysis and observation: The article argues that the core of the market repricing is not a single day's rise or fall, but whether Jensen Huang's personally borne technical judgment, customer trust and ecosystem coordination capabilities can be converted from personal assets into organizational assets; a successor must simultaneously take on the technology line, the customer line and the governance line. Investors should watch the pace of CEO selection, the board's substantive involvement in strategy and risk, and whether the product roadmap can be delivered on schedule after management changes. (Market data in the text such as "first up 10.05%, then down 4.16%" and "$3.925 trillion market capitalization" lack consistent authoritative sources and should not be treated as confirmed market moves.).
V. Meta fined €805 million for default photo album access crossing the privacy red line
1. On November 14 local time, Italy's data protection authority fined Meta Platforms Ireland Limited €805 million (about $887 million): since 2023, the Facebook Android app has enabled by default the feature "get recommendations from your photo album while browsing Facebook," accessing users' photo albums based on a single blanket authorization covering the entire app, without obtaining separate consent for the specific purpose and without adequate notice, affecting about 20 million Android users in Italy.
2. The regulatory focus is the "granularity of consent": Italy treats the phone photo album as private living space and holds that "the user did not refuse" does not equal "the user has consented," requiring Meta to obtain consent again through a separate, explicit request, change the feature to off by default, complete remediation within 60 days, and report progress within 30 and 90 days; Meta has said it strongly disagrees and plans to appeal, and the Italian side also rejected its jurisdictional objection.
3. Impact and trends: The fine is about 0.4% of Meta Group's global revenue and comes on top of an existing 2019 ruling and a $25 billion fine from New Mexico in October, showing that the growth model of "on by default plus blanket authorization" is facing regulatory pressure; Italy said it will separately assess Meta's new 2026 feature for cloud processing of photo album content, and this proceeding does not concern whether the data is used for AI training.
VI. Where did Nvidia's security team go
1. Nvidia Chief Security Officer and head of product security David Reber left in August 2026; he had been at Nvidia for seven years and CSO for four, and the stated reason was that he had been diagnosed with early-onset Parkinson's disease more than two years earlier and needed to put family and health first; his successor is Saša Zdjelar (former senior IT manager at ExxonMobil, with a background at Salesforce and ReversingLabs). Public information does not establish a link between his departure and the Megaspeed investigation; the report merely places the two time points side by side.
2. Singapore data center company Megaspeed is under investigation by U.S. and Singapore authorities; it denies wrongdoing and says it is cooperating; public reports indicate the U.S. investigation was later referred to the Department of Justice. Nvidia GPUs move through distributors, cloud service providers, data centers and cross-border logistics, and customer due diligence, export controls, channel monitoring and technical support records are scattered across multiple departments, with risk signals often only becoming apparent once pieced together.
3. The article's core point is the governance architecture rather than conviction in an individual case: when a core executive has a personal relationship with the management of a customer under scrutiny, the company needs clear conflict-of-interest disclosure, information isolation, independent review and board escalation mechanisms. The view is that a mature security system should continue to function after key personnel depart and ensure that independence is maintained even when risk reaches the CEO, understanding security as a set of organizational capabilities running through business relationships, personnel relationships and national regulation.
VII. The OpenAI turmoil shows AI safety cannot rely on the CEO
1. Event and review: On November 17, 2023, OpenAI announced Sam Altman's departure, officially saying his communication with the board was "not consistently candid"; after management turmoil and board restructuring, Altman returned in late November. In 2024, the WilmerHale review examined more than 30,000 documents and dozens of interviews, concluding that the dismissal stemmed from a breakdown of trust between the board and Altman rather than product safety, development speed or financial concerns, while criticizing the board for making the decision hastily and without notifying key stakeholders.
2. Core view: Helen Toner warned that if an AI safety plan depends on one person's integrity and self-restraint, institutionally it amounts to no safety plan at all. Although the board had the power to replace the CEO, because the information was held by the person being overseen, it had only "the right to explain afterward"; claims that the board knew nothing about the ChatGPT launch and was not informed of Altman's fund interests are her personal statements, and OpenAI did not respond to them item by item.
3. Governance recommendations: Key powers should be separated, achieving a structural separation between risk assessors and business decision-makers, and safety leaders should have blocking power and written records. Regulation should focus on verifiable behavior (high-risk model evaluation, incident reporting, internal audit, external oversight and measures triggered by risk thresholds) rather than looking for an "unreliable CEO"; the goal is for the organization to still be able to detect, block and correct when any CEO makes a mistake.
1. AI has not replaced hackers but has entered several links in the attack chain: writing more convincing phishing emails, checking code, finding vulnerabilities, analyzing data and assisting in negotiations, compressing the most time-consuming and easily replicated parts; Allan Liska of Recorded Future and Cynthia Kaiser of Halcyon judge that successful groups are still mainly operated by humans, and there is not yet a fully AI-driven mature attack.
2. Microsoft's "Digital Defense Report 2025" says that activity targeting Azure customer environments for destructive purposes such as ransomware and mass deletion increased by 87%; more than 40% of ransomware attacks now include hybrid-environment components (below 5% two years ago). Attackers use compromised identities and remote management tools to move laterally, delete virtual machines and backups, exfiltrate and encrypt cloud resources, so a single identity theft can become a pass across cloud and endpoints.
3. The defensive focus should shift from "identifying a particular virus" to "limiting the scope of impact after a breach": multi-factor authentication, least privilege, network segmentation and timely patching need to cover cloud accounts, remote management tools and backup environments; CISA recommends maintaining offline encrypted backups and regularly testing recovery, because only when attackers cannot delete them at the same time and the enterprise can restore operations under pressure do backups truly constitute a line of defense.