Anthropic Ties Itself to Nearly $901 Billion in Compute Contracts
Key points
- AI advances intensify the infrastructure bet: Anthropic’s long-term compute contracts show how costly it can be to plan ahead for rapidly growing models.
- The model race remains open: Google’s Gemini 4 Carbon tops a math benchmark, while Anthropic’s new prompting guidance shows how working with models is changing.
- AI is moving further into the physical world: OpenAI is reorganizing its robotics teams, while Runway has released a robotics model with open weights and source code.
- Governance and leadership are taking center stage: The debate around Nvidia and calls for Jensen Huang’s return show how closely corporate oversight and AI infrastructure are intertwined.
🔬 Models & Research
Google beats OpenAI: Gemini 4 Carbon tops the MathArena AI math benchmark
14.11.2026, 17:37
Google’s Gemini 4 Carbon has overtaken GPT-6.1 Sol on MathArena, placing it ahead of the tested OpenAI and Claude models. For teams choosing models for demanding math tasks, the reported weakness under time limits matters alongside the rankings.
Anthropic explains which prompting habits users should drop with Claude Opus 5.5
14.11.2026, 17:14
Anthropic recommends dropping older prompts such as “think hard” or “think carefully” when using Claude Opus 5.5, because the model handles such instructions differently. Prompt templates should therefore be reviewed for new model generations, rather than automatically reusing existing wording.
Claude Mythos was no outlier but part of an accelerating trend, says former researcher
14.11.2026, 11:48
Former OpenAI and Anthropic researcher Jacob Coxon sees Claude Mythos as a sign of accelerating development and points to a chart from Anthropic’s IPO roadshow. His assessment draws attention to whether safety evaluations can keep pace with increasingly capable models.
💼 Business & Markets
“Great uncertainty”: Anthropic CEO wrestles with the AI industry’s biggest bet shortly before record IPO
14.11.2026, 18:51
Anthropic CEO Dario Amodei describes long-term purchases of compute capacity as a key business risk; according to the report, the commitments total nearly $901 billion. These large upfront commitments highlight how difficult it is to plan capacity one or two years in advance when model performance and demand are changing rapidly.
Where was Nvidia’s security team?
15.11.2026, 01:08
A report examines the departure of Nvidia’s longtime head of security in August and raises questions about internal oversight amid current allegations involving Jensen Huang and possible chip diversions. For companies with critical AI infrastructure, the case shows how important clear responsibilities are across product security, corporate leadership, and compliance.
Elon Musk and other tech executives petition for Jensen Huang’s return to Nvidia
14.11.2026, 22:41
According to the report, a petition calling for the reinstatement of ousted Nvidia chief Jensen Huang gathered more than 115,000 signatures, including those of Elon Musk and other tech executives. The public support puts more pressure on the board and highlights the importance of transparent decision-making in leadership changes.
Jensen Huang and the decline of the visionary showman founder-CEO
14.11.2026, 13:00
The article places Huang’s departure in the context of changing public perceptions of charismatic founder-CEOs and refers to the allegations against him. For a company at the center of the AI market, the debate shows how closely trust in a leader is tied to perceptions of governance and corporate risk.
⚖️ Politics & Regulation
Asian governments react to Nvidia CEO Jensen Huang’s departure
14.11.2026, 17:34
Governments in Asia are responding differently to Huang’s departure: some are reaffirming chip supply commitments, while others are monitoring local projects or ongoing investigations. These responses underscore how closely regional AI infrastructure is connected to government interests in reliable supply chains.
The former OpenAI board member who ousted Sam Altman says AI CEOs still can’t be trusted
14.11.2026, 16:00
Helen Toner, a former OpenAI board member and current head of the Center for Security and Emerging Technology, discusses corporate oversight and AI safety. Her perspective highlights the question of what independent oversight mechanisms can prevent commercial interests from displacing an AI company’s safety mission.
🤖 Robotics
OpenAI reassigns robotics researchers to other teams because general-purpose models make specialized robotics AI unnecessary
14.11.2026, 11:35
OpenAI is reorganizing its robotics researchers, integrating research into its central research team and applied robotics and hardware into its Applied organization. The restructuring is intended to connect robotics more closely with general models and product development, rather than abandon the field.
Open-source robotics model: Runway releases Praxis-1 with open weights
14.11.2026, 10:34
Runway has released Praxis-1 with open weights and source code under Apache 2.0. The model processes camera images, text instructions, and position data to derive movements for robots. The open license makes it easier to adapt the model to different systems, although it still makes mistakes with similar-looking objects and complex action planning.