Jensen Huang leaves Nvidia
Nvidia said on Thursday 12 November (US time) that Jensen Huang had stepped down as president and chief executive and left the board, all with immediate effect. Colette Kress, the chief financial officer, is now running the company on an interim basis and keeps her finance job too. She's been at Nvidia since September 2013. Harvey C. Jones, who used to run Synopsys, is the company's first board chair. The board will look at people from inside and outside Nvidia for the permanent role.
A day later the Financial Times said Arm's chief executive Rene Haas was one of several people the board had approached. On 13 November Nvidia shares were down more than 3% and Arm's more than 5%. Reuters, reporting the FT story, said Huang left after extramarital affairs came to light, and that one of them involved a suspected Chinese spy. Nvidia still plans to release its fiscal 2027 third-quarter results on 17 November.
What it means for NZ organisations: Nvidia hasn't announced any change to supply or pricing. If you're negotiating a large GPU or cloud deal, its guidance comes out at about 10:20am NZT on Wednesday 18 November, and the analyst call starts at 11am.
OpenAI turns GPT-6 Cyber into two products
OpenAI president Greg Brockman launched Daybreak Red Team and Daybreak Defense Factory at a keynote on 12 November. Red Team is in early access. It sends many agents through an application's code and context to find routes an attacker could use. Security staff can also use it to check bug reports from scanners or bounty programmes and to retest the fixes. Defense Factory scans repositories with agents running on GPT-6 Cyber. Each fix comes as a merge-ready pull request with a test, and an engineer approves it before it reaches production.
GPT-6 Cyber came out on 5 November and only vetted customers can use it. It's the first OpenAI model rated Critical for cybersecurity under the company's Preparedness Framework. In testing it found 81 serious Firefox flaws that nobody knew about.
What it means for NZ organisations: As tools like this spread, attackers will exploit flaws sooner after they're disclosed. Check whether your patch and change processes could cope with a sudden jump in the number of fixes.
Gemini 4 Hydrogen goes after high-volume agent work
Google started rolling out Gemini 4 Hydrogen on 12 November as the cheaper model in its Gemini 4 range. It costs US$1.50 per million input tokens and US$7.50 per million output tokens, and that rate holds across the full 1 million token context. Google says it scores 57% on Terminal-Bench 4.1, against 18.2% for Gemini 3.8 Flash, and finishes tasks about 2.5 times faster at default settings.
Antigravity also gets a new Orchestrator that switches between Carbon and Hydrogen within a session. On Terminal-Bench 4.0, Google says it came close to Carbon's success rate while using 24% less quota on average and finishing 19% faster. You can get Hydrogen through the Gemini API, AI Studio, Gemini for business, the Gemini app and Search.
What it means for NZ organisations: Teams with heavy coding or agent workloads could have a premium model plan the work and let Hydrogen do it. Google supplied every figure here, so test it on your own workloads first.
Mirendil raises a US$1 billion Series A
On 12 November Mirendil, a San Francisco lab, announced a US$1 billion Series A led by Kleiner Perkins and Clearlake Capital. Google, Battery Ventures and others also invested. The company built its Mirendil-Base coding model from scratch in 3.5 months with less than US$10 million of compute. Fern, trained further from that base, can take up to 10 million tokens of context. Lor, a platform that runs hundreds of agents on coding and experiments lasting several days, is in private beta.
What it means for NZ organisations: Capable models keep coming from smaller labs as well as the big names. Test newcomers on your own tasks before you sign anything.
Inkling 2 Small weights released
Thinking Machines Lab put Inkling 2 Small's weights on Hugging Face on 12 November, a week after it announced the model on 5 November. The lab held the weights back until outside red-teaming and safety testing were done. The model is a mixture-of-experts design with 594 billion parameters in total and 17 billion active, and the lab has said it'll use the Apache 2.0 licence. Fireworks, Together, Nebius, Cloudflare and Nvidia all host it. It scores 31 on the Artificial Analysis Intelligence Index, and the lab says it trails the closed GPT-6.1 Sol on internal misbehaviour tests.
What it means for NZ organisations: Hosting the model yourself keeps data onshore, but the full model needs a lot of GPU capacity. A hosted API is the faster way to try it out.
Nonprofits ask a court to stop frontier development
On Thursday 12 November the Irreplaceable Education Fund, Public Citizen, Heat Initiative and the Alliance for Secure AI filed a lawsuit in San Francisco Superior Court. They're suing Google, Meta, OpenAI, Anthropic and SpaceX's AI division. The claims rest on California's unfair competition law and on public nuisance, and the groups want an injunction that would block new advanced models from being built or released.
What it means for NZ organisations: Nothing changes yet. It's still worth checking the continuity and exit terms in your AI supplier contracts.
Anthropic's IPO filing says AI is speeding up its own research
On 10 November Anthropic filed its second amended S-1 for a planned Nasdaq listing under the ticker ANT. The filing says the company has passed the threshold in its Responsible Scaling Policy for progress sped up by AI. Anthropic bases that on what it expects from future models and its overall pace of work. It also warns that precautions such as delaying releases or limiting access could affect future launches. On 9 November Anthropic released Claude Fable 5.2 at US$8 per million input tokens and US$40 per million output tokens. It scored 65 on the Artificial Analysis Intelligence Index, ahead of Gemini 4 Carbon on 60.
What it means for NZ organisations: If Claude runs important workflows for you, release timing may become harder to predict. Keep a tested backup provider ready.
Sources
- NVIDIA Announces CEO Transition
- Introducing Gemini 4 Hydrogen (Google)
- Inkling 2 Small (Thinking Machines Lab)
- Mirendil raises $1B to democratize frontier AI R&D (PR Newswire)
- OpenAI launches security products (The Register)
- Nvidia approached Arm CEO Haas to succeed Huang: FT (RTÉ)
- Anthropic S-1 accelerated AI progress disclosure (CASRAI)
- Nonprofits sue OpenAI, Google and Meta to halt advanced AI (Dataconomy)
- Anthropic's Claude tops AI rankings as OpenAI limits cyber model (PYMNTS)
Want to know which of these changes matter for your organisation? Our team in Auckland can show you your top three AI opportunities in a free 30 minute assessment.
Get my free assessment How we help