Welcome to Founder Insights: Weekend Edition. Each Saturday, I will distill the essential signals from the AI landscape, surfacing the shifts that truly move the needle for founders.
This week saw Sora 2’s explosive launch, which proved consumer demand for AI creation tools. With 164,000 downloads, it climbed to #1 on the App Store in just 72 hours. Enterprise adoption accelerated as Microsoft added Claude to Copilot, and Google drove efficiency with Gemini 2.5 Flash, cutting token costs by 50%.
Infrastructure momentum hit new heights. CoreWeave secured a $14.2B deal with Meta through 2031. Nebius signed a $17.4B five-year contract with Microsoft. These aren’t one-off bets. They’re decade-long commitments reshaping the compute landscape.
On the governance front, California’s SB 53 sets the first comprehensive standard for AI safety. In research, we’re seeing a shift toward efficiency and agentic reasoning as the new frontier. Globally, the U.S., China, and Europe are doubling down on sovereign compute and AI policy.
Here’s your Saturday guide to what’s shaping the future of AI:
Infrastructure
CoreWeave secures $14.2 billion long-term contract with Meta through 2031, representing a massive commitment to specialized AI cloud infrastructure. The six-year deal reinforces Meta’s aggressive AI scaling strategy. Click here
Nebius signs a $17.4 billion, five-year deal with Microsoft to supply dedicated AI compute capacity, expanding Microsoft’s access to next-generation cloud infrastructure. Click here
Huawei plans to double the output of Ascend 910C AI chips to 600,000 units in 2026, defying U.S. export controls. The move signals China’s accelerating push for semiconductor self-sufficiency and global AI parity. Click here
Enterprise
OpenAI’s invite-only Sora app hit #1 on Apple’s App Store in just 72 hours, logging 164,000 downloads and underscoring explosive demand for AI-driven video creation. Click here
Microsoft 365 Copilot now supports Anthropic’s Claude Opus 4.1 and Sonnet 4, giving enterprises a true multi-model ecosystem. The move signals a shift from monolithic AI platforms to modular, multi-model ecosystems. Click here
Google rolls out Gemini 2.5 Flash and Flash-Lite, reducing token costs by up to 50% while enhancing agent tool use and instruction following. Efficiency gains are now as critical as raw performance. Click here
Capital Flows
Cerebras raises $1.1B Series G at $8.1B valuation, fueled by explosive demand for AI inference and wafer-scale compute. The round, led by Fidelity and Atreides, will fund new data centers and U.S. manufacturing. Click here
Nscale raises $1.1 billion Series B led by Aker ASA, the largest round in European history. The London-based AI infrastructure company will accelerate the deployment of Stargate UK and Norway facilities, with backing from Nvidia, Dell, Nokia, and Fidelity Management for sovereign AI infrastructure across Europe. Click here
Global AI investments surge to unprecedented levels with spending projected to reach $1.5 trillion by 2025 and exceed $2 trillion in 2026, representing nearly 2% of global GDP. The U.S. leads with $470 billion invested from 2013 to 2024, while China follows with $119 billion, as geopolitical competition intensifies infrastructure buildouts worldwide. Click here
Meta acquires RISC-V chip startup Rivos to accelerate internal AI semiconductor development and reduce dependence on external suppliers. The deal, reportedly valued at over $2B, brings expertise in building full-stack AI systems to Meta’s Training and Inference Accelerator program. Click here
Research
Anthropic achieves breakthrough in AI safety with Claude Sonnet 4.5 trained at AI Safety Level 3, reducing sycophancy, deception, and power-seeking behaviors. The model shows 10x improvement in reducing false positives from safety filters while maintaining resistance to prompt injection attacks. Click here
Google DeepMind unveils Gemini Robotics 1.5 and ER 1.5, enabling robots to “think” before acting with multi-step reasoning. The dual-model system separates planning from execution, with ER 1.5 orchestrating tasks and calling digital tools like Google Search, while Robotics 1.5 converts plans into precise motor commands. Click here
The Attribution-Based Control framework proposed to unlock access to 200 zettabytes of unused digital data while preserving ownership. The solution could break the “peak data” bottleneck. Click here
Policy
California enacts SB 53, the nation’s first comprehensive AI safety law, mandating transparency, incident reporting, and whistleblower protections. The law sets a precedent for federal and global regulation. Click here
Meta to use AI chat data for ads and content personalization globally starting December 16, 2025—excluding the UK, EU, and South Korea. The policy shift, with no opt-out, raises fresh privacy and consent concerns. Click here
U.S. rejects international AI oversight at UN General Assembly, clashing with world leaders over global AI governance frameworks. US representatives expressed deep opposition to any global effort to govern AI technology, while heads of state highlighted risks and opportunities from artificial intelligence development. Click here
Global AI Strategy
Italy becomes the first EU member state to pass comprehensive national AI legislation, complementing the EU AI Act, effective October 10, 2025. Law No. 132/2025 establishes fundamental principles for AI deployment across healthcare, public administration, judicial activities, and national security while maintaining consistency with European frameworks. Click here
Samsung & SK Hynix join OpenAI’s Stargate initiative, targeting 900K DRAM wafers/month, including floating maritime data centers. Click here
Alibaba Cloud expands globally with new data centers and launches its AI Catalyst Program, offering up to $120K credits and 2B free model tokens. Click here
