G'day. It's been one of the biggest weeks in AI history: Nvidia's founder walked out the door, Anthropic kicked off the most anticipated tech IPO in years, OpenAI crossed $10b in quarterly revenue for the first time, and Google wrapped up its entire Gemini 4 model family in four days. If you felt a slight tremor this week, that was the AI industry rearranging itself at speed.
TL;DR
- Jensen Huang Out at Nvidia: The co-founder and 33-year CEO resigned effective immediately on 12 November amid reports of two extramarital affairs, including one with an alleged Chinese intelligence asset. CFO Colette Kress steps in as interim CEO while a permanent search begins.
- Anthropic's $1.4t IPO Roadshow: Anthropic launched its Nasdaq roadshow on 10 November at a $70–$78 per share price band, targeting a fully diluted valuation of up to $1.4t. Final pricing is set for 18 November with trading under the ticker ANT expected to begin 19 November.
- ChatGPT Gets a Wallet: OpenAI launched ChatGPT Wallet on 13 November, letting its AI agents (called 'dots') buy things on your behalf via American Express, Mastercard and Visa without ever handling your real card number. Currently limited to US Pro subscribers.
- Google Completes Gemini 4 Family: Google dropped Gemini 4 Carbon (10 Nov) and Gemini 4 Hydrogen (12 Nov) in quick succession, completing its frontier model tier. Hydrogen is priced at $1.50 per million input tokens and is free on Gemini API, designed for fast agentic and coding workflows.
- DeepMind Proves Signal's Encryption: AlphaProof Code, DeepMind's new AI verification framework, formally proved Signal's cryptographic protocols down to compiled machine code in just four weeks. The same task would take a human expert team years.
- OpenAI Hits $10b Quarterly Revenue: OpenAI reported Q3 2026 revenue of roughly $10b, up nearly 50% from $6.7b in Q2, putting its annualised run rate at around $50b. The flip side: operating losses also widened to $10.9b for the quarter.
Jensen Huang's Sudden Exit Leaves Nvidia's AI Throne Empty at the Worst Possible Moment
On 12 November 2026, Nvidia's board announced that Jensen Huang had resigned as President, Chief Executive Officer, and board member, effective immediately. The company named CFO Colette Kress as Interim CEO and appointed Harvey C. Jones as the first-ever Chair of the Board. A formal search for a permanent CEO covering both internal and external candidates is now underway.
The exit came days after reports emerged that Huang had two extramarital affairs, one of which involved a woman alleged by US intelligence agencies to be a Chinese intelligence asset. Nvidia initially signalled Huang would remain while an internal investigation proceeded. That position collapsed on Wednesday 11 November when further reports surfaced of a second relationship with an Nvidia executive; the resignation followed within hours.
The timing is brutal for the company. Nvidia was already scheduled to report its third-quarter fiscal 2027 results on 17 November, and the firm supplies a majority of the world's AI training and inference chips. Kress, who has served as CFO since September 2013, inherits that earnings call alongside the broader task of steadying investor confidence. NVDA shares fell sharply on the news before partially recovering.
Nvidia's board has approached Arm CEO Rene Haas as a potential successor, according to the Financial Times, though no formal offer has been confirmed. Senate Banking Committee chair Tim Scott has separately called Huang and Commerce Secretary Howard Lutnick to appear before the committee for hearings on national security implications of the espionage allegations.
The human dimension aside, Huang's departure removes the single most influential figure in the AI hardware supply chain. His direct relationships with hyperscalers, government officials, and chip architects were a core part of Nvidia's competitive advantage. Kress is a highly capable operator, but she inherits a company that just had its identity ripped away while AI chip demand — and geopolitical scrutiny — have never been higher.
For Australian businesses, the near-term impact is likely muted: Nvidia's product roadmap and manufacturing partnerships do not change overnight. But leadership uncertainty at the world's dominant AI chip supplier is worth monitoring, particularly if it affects Blackwell supply timelines or pricing negotiations heading into 2027 budget cycles.
Anthropic's $1.4t IPO Roadshow: What the Numbers Actually Mean for AI's Money Era
Anthropic officially launched its IPO roadshow on 10 November 2026, setting a price band of $70 to $78 per share across 825 million Class A common shares. At the top of the range, the offering raises $64.4b and values the company at roughly $1.4t on a fully diluted basis. Final pricing is expected on 18 November, with Nasdaq trading under the ticker ANT beginning 19 November.
Seven strategic partners, including AMD, Broadcom, Google, Micron, Microsoft, Nvidia, and TSMC, have agreed to purchase $26b of Class N common stock in a parallel private placement. The roadshow was quickly reported as oversubscribed. Anthropic's Q3 2026 revenue came in at $17b per its amended S-1, though growth has decelerated from the 140%-plus clip of earlier quarters to around 47% quarter-on-quarter.
Context matters here. No dedicated AI model developer has ever gone public before. The market has no direct comparable. Critics have noted that much of Anthropic's value is likely already priced in; Goldman Sachs models Anthropic reaching $685b in revenue by 2030 and operating profit in 2029. The IPO also excludes orders from China and Hong Kong, a pointed geopolitical signal from CEO Dario Amodei, who has described compute resources as the industry's biggest constraint.
For Australian investors and businesses tracking the sector: Anthropic has opened a retail IPO allocation through CommSec, Stake and Macquarie ahead of its Nasdaq listing, so local participation is on the table. Whether the valuation holds post-listing is a separate question.
DeepMind's AlphaProof Code Proves Signal's Encryption in Four Weeks, Not Years
On 10 November, Google DeepMind announced AlphaProof Code, a multi-agent AI framework that uses Gemini 4 Carbon-powered agents to write formal proofs in the Lean theorem prover. Its debut project: mathematically verifying the security of the Signal Protocol's compiled machine code. The agents produced more than six million lines of machine-checked Lean proof, covering the protocol's post-quantum key exchange mechanism (PQXDH) and the Triple Ratchet scheme that rotates encryption keys with every message.
The significance is that this verification went all the way down to the compiled 64-bit Arm binary, not just the source code. Gaps between source and compiled code are where real-world vulnerabilities often hide. DeepMind checked Signal's compiled library against a model of the processor written in Lean, proving the security guarantees hold even at that level. Signal CTO Ehren Kret described the work as expanding the scope of earlier verification efforts significantly.
The process took four weeks. Equivalent human expert projects typically take years. DeepMind has open-sourced the specifications and proofs on GitHub. The team also stress-tested the approach by reintroducing two previously fixed bugs into an isolated copy of Signal's code: AlphaProof Code correctly identified both as insecure. No new vulnerabilities were found in production, which DeepMind was upfront about: this is a demonstration of a methodology, not a security audit of the full app.
The implications for software security are significant. DeepMind has flagged medical devices and power grids as future verification targets. For anyone building on top of AI-assisted development tools, the promise is eventually being able to prove your code does what you say it does, automatically, before it ships.
Quick Hits
- Claude Fable 5.2 Released: Anthropic dropped Fable 5.2 on 9 November: 20% cheaper at $8/$40 per million tokens, 30% faster than 5.1, and around 55% fewer unnecessary blocks. The restricted Mythos 5.2 variant (same underlying model, fewer safety constraints) is available to vetted cybersecurity and life sciences organisations. A team of 1,000 Fable 5.2 agents also produced a machine-checkable proof of Fermat's Last Theorem in 30 hours. Anthropic / MacRumors
- OpenAI Crosses $10b Quarterly Revenue: OpenAI's Q3 2026 revenue hit $10b for the first time, up nearly 50% from $6.7b in Q2, with the enterprise business doubling quarter-on-quarter and 1.2 billion weekly active users. The operating loss also widened to $10.9b, so the company is spending roughly $1 for every dollar it earns, plus more. TechCrunch / Bloomberg
- Gemini 4 Family Complete: Google released Gemini 4 Carbon on 10 November (frontier reasoning, $2/$10 per million tokens at intro pricing, Ultra subscribers only in Gemini app) and Gemini 4 Hydrogen on 12 November (fast everyday coding, $1.50/$7.50, free API tier). A new Orchestrator mode pairs Carbon as planner and Hydrogen as executor, achieving near-Carbon task success on Terminal-Bench 4.0 with 24% less quota and 19% faster completion. Google / Testing Catalog
- Grok Bot 1.0 Launches: xAI announced Grok Bot 1.0 on 11 November, graduating from beta to a full team product. Bots can now be shared across a whole team, join Slack, Microsoft Teams and video calls, and proactively identify and pick up tasks without being asked. Team and Enterprise plans unlock shared Bots; meeting access also includes Ultra plans. xAI / Windows Report
- Gemini Robotics 2.5 Debuts: Google DeepMind introduced Gemini Robotics 2.5 this week, a vision-language-action model with improved dexterity, task breadth and hardware compatibility. In proof-of-concept testing it achieved a 96% success rate sorting automotive parts on Mercedes-Benz production lines using Apptronik Apollo 2 humanoids. AI Week in Review / The Neuron
- Frontier AI Freeze Lawsuit Filed: Four US nonprofits sued Anthropic, Google, Meta, OpenAI and SpaceX's AI unit under California's Unfair Competition Law this week, seeking a judicial freeze on frontier AI development. The plaintiffs acknowledged that any real freeze would require government inspections and ongoing monitoring. learnrudi.com / AI News Nov 14
- $129m AI-Powered Corporate Fraud Charges: US prosecutors charged three individuals in a $129m corporate fraud scheme that used AI to fabricate financial documents and automate deception at scale. The case is one of the largest AI-assisted fraud prosecutions on record and signals regulators are actively watching for AI-enabled financial crime. learnrudi.com / AI News Nov 14
- Shared AI Training Opt-Out Signal: Google, Microsoft, OpenAI and Anthropic have confirmed they will honour a single robots.txt signal for AI training preferences, giving publishers and website owners one lever to control data use across all four major labs rather than managing separate opt-outs for each. aicodesit.com / AI News November 2026
The ClickedOn Take
Three storylines this week converge into one message for Australian businesses: the AI industry is no longer building hype, it is building infrastructure, and the costs and consequences are becoming very real, very fast.
Anthropic's IPO roadshow is the clearest signal yet that AI is now a capital market story. At a potential $1.4t valuation, the market is pricing in AI becoming critical infrastructure for a generation. That matters if you are deciding whether to commit to Claude via AWS or Azure, because you are now partnering with a publicly accountable entity with shareholder obligations, not just a scrappy research lab.
ChatGPT Wallet is the one to watch most closely from a performance marketing perspective. When AI agents can autonomously complete purchases, the entire conversion funnel shifts. The agent becomes the buyer. That means your product data, your checkout flow and your merchant integrations need to be agent-readable, not just human-readable. OpenAI has published a Commerce Partner Playbook for exactly this reason. If you sell online and are not thinking about agentic commerce readiness, now is the time to start.
Closer to home, the shared robots.txt opt-out agreement from Google, Microsoft, OpenAI and Anthropic is directly relevant to any Australian business with a content strategy. Your SEO and Generative Engine Optimisation (GEO) settings now need to reflect a deliberate choice about whether you want your content used for AI training and how you want to appear in AI-generated answers. We can help you audit both. Check our free tools to get started.
Tool of the Week
Google AI Studio (Gemini 4 Hydrogen) →
Google AI Studio now gives every developer free access to Gemini 4 Hydrogen, the company's fast agentic model with a 1-million-token context window. It is genuinely capable for coding, document analysis and multi-step workflows at $1.50 per million input tokens on paid tiers, or free on the developer tier. For Australian marketers and developers, it is the most accessible entry point into frontier-tier Google AI right now. One caveat: Carbon, the more powerful model for complex reasoning tasks, is restricted to Ultra subscribers and paying API customers, so free-tier users are on the capable but lighter Hydrogen variant.
Sources This Week
Nvidia Newsroom, Washington Post, PBS NewsHour, CNN, Yahoo Finance, Blockchain.News, Google DeepMind Blog, Cyber Insider, Technology.org, Anthropic, MacRumors, TechCrunch, Bloomberg, Dataconomy, Testing Catalog, xAI, Windows Report, AI Week in Review (Patrick McGuinness), The Neuron AI, learnrudi.com, aicodesit.com, Mind and Machine Weekly, headsupai.io, LLM Gateway, NHIMG LLM Release Tracker, opper.ai, cellcog.ai, superpowerdaily.com, founderz.com, openai.com, CNBC, pymnts.com, ppc.land, online-tech-tips.com