Founder Insights: Weekend Edition – Issue #7

Theme of the week, AI wars. Power is the story, compute power, capital power, national power. Anthropic lines up $50 billion for U.S. capacity. Microsoft commits $10 billion in Portugal and lights up Atlanta. Google puts €5.5 billion into Germany. Cisco books $1.3 billion in AI networking. The buildout is global and accelerating.

Inside the enterprise, agents move from pilot to policy. Amazon unifies ad buying and rolls out planning agents. Cisco reaches for NeuralFabric. TIME turns a century of reporting into an agent with Scale AI. Talent shifts as Yann LeCun prepares a new startup.

Capital moves with intent. SoftBank exits NVIDIA to fund a $30 billion OpenAI push. Cursor jumps to a $29.3 billion valuation. Parallel raises $100 million for live web search APIs for agents. Scribe and Clio raise to automate real work.

Research and policy set the rules of engagement. Pathway explores brain-inspired LLMs. DeepMind maps new singularities. AWS helps digitize the Goodall archive for AI. The U.S. pauses the Affiliates Rule for a year. The World Economic Forum floats a global AI constitution. State AGs partner with OpenAI and Microsoft on safety. EU timing on the AI Act is in flux. Strategy is national, China doubles down with new chips and Ernie 5.0, institutional capital flows into 1.7 GW of data centers, India advances IndiaAI.

The AI wars are here, and the map is changing fast. The question is not who is innovating, it is who is arming up and converting spend into durable advantage. 

Here’s your Saturday guide to the signals shaping the future of AI:

Infrastructure

• Anthropic to invest $50 billion in U.S. AI data centers. Anthropic will roll out a massive infrastructure build-out in Texas and New York to support its Claude model family, create about 800 permanent jobs and 2,400 construction jobs, and keep pace in the escalating AI compute arms race. Click here
• Microsoft to invest $10 billion in an AI data center hub in Sines, Portugal. Microsoft will build out a large-scale facility partnered with Start Campus and Nscale, deploying around 12,600 NVIDIA GPUs as part of its broader move to expand AI cloud capacity across Europe. Click here
• Google commits €5.5 billion to expand AI infrastructure and offices in Germany. Google will invest through 2029 to scale AI compute capacity, including a new data center build-out, as part of its broader push to meet rising demand for AI workloads across Europe. Click here
• Cisco reports $1.3 billion in AI infrastructure orders from hyperscalers in Q1 FY2026. The surge highlights rapid adoption of Cisco’s high-performance networking gear for AI data centers as hyperscalers expand GPU clusters and backbone capacity. Click here
• Microsoft unveils its Atlanta AI data center, part of the Fairwater network. The facility links with a Wisconsin campus to form a “superfactory” powered by hundreds of thousands of NVIDIA GPUs for real-time AI operations. Click here

Enterprise

• Amazon consolidates all ad-buying tools and introduces AI agents for media planning. The unified platform streamlines ad buying across Amazon’s marketplace and the open web while new AI agents help advertisers target audiences and plan campaigns. Click here
• Cisco plans to acquire NeuralFabric to expand its enterprise generative AI stack. The deal brings Cisco a Seattle-based platform focused on secure model deployment, data orchestration and AI-driven automation, strengthening its roadmap for business-grade generative AI. Click here
• Time Magazine launches an AI agent built on its 100-year archive in partnership with Scale AI. The new system enables rich reader interaction with historical content and opens enterprise opportunities for licensing Time’s archival intelligence as an AI-powered platform. Click here
• Meta’s Yann LeCun plans to leave to launch his own AI startup. Meta’s chief AI scientist is preparing to depart after more than a decade, entering fundraising talks as Meta restructures its AI research under a new superintelligence division. Click here

Capital Flows

• SoftBank sells its entire $5.83 billion NVIDIAstake to fund a massive $30 billion OpenAI investment. The firm liquidated its NVIDIA position and additional T-Mobile shares to finance the next phase of its escalating commitment to OpenAI. Click here
• Cursor raises $2.3 billion at a $29.3 billion valuation. The MIT-founded startup behind a popular AI coding tool secured its third funding round of the year, accelerating its growth in the “vibe coding” space and marking one of the fastest paths to decacorn status in AI developer tools. Click here
• Ex-Twitter CEO Parag Agrawal’s AI search startup Parallel Web Systems raises $100 million. Series A funding co-led by Kleiner Perkins and Index Ventures pushes Parallel’s valuation to approx. $740 million and backs its platform delivering live-web search APIs built for AI agents. Click here
• Scribe raises $75 million and hits a $1.3 billion valuation to map and automate high-ROI enterprise workflows. The funding backs its new “Scribe Optimize” platform that mines actual employee task data across apps so companies can identify where AI and automation will genuinely pay off—rather than rolling out bots blind. Click here
• Clio raises $500 million and reaches a $5 billion valuation as it acquires vLex for $1 billion. The Canadian legal-AI company secured a Series G funding round led by NEA, plus a $350 million debt facility, to accelerate its transformation into an “AI-first” platform integrating law-firm operations, research, drafting and analytics. Click here

Research

• “Dragon Hatchling” brain-inspired LLM architecture unveiled by startup Pathway. The new prototype, dubbed BDH, uses a scale-free network of “neuron-particles” and synaptic plasticity to dynamically rewire itself over time, moving beyond fixed-transformer models toward continual learning and reasoning more akin to biological brains. Click here
• AI-accelerated primatology: Jane Goodall Institute partners with Amazon Web Services to digitize 65 years of field research. With a $1 million commitment, AWS will convert decades of handwritten notes, film footage and observational data on chimpanzees and baboons into a searchable, AI-powered research platform—bringing long-cycle, field-based science into the AI age. Click here
• DeepMind cracks a century-old physics puzzle by uncovering new families of unstable singularities in fluid dynamics. The research from DeepMind used physics-informed neural networks to identify previously uncharted “blow-up” solutions across multiple core fluid-dynamics equations, opening fresh pathways for understanding turbulence, aerodynamics, and blood flow. Click here

Policy

• US pauses the “Affiliates Rule” export control for one year in a trade truce with China. As part of a broader agreement with Beijing, the US will suspend the Affiliates Rule—which targets subsidiaries of blacklisted entities—until November 2026, drawing criticism from lawmakers who say the move weakens U.S. national security safeguards. Click here
• WEF proposes a global AI “constitution” to guide international governance. A new World Economic Forum brief outlines a three-pillar constitutional framework for AI and spotlights regional strategies such as the African Union’s Continental AI Strategy as models for building sovereign capability. Click here
• OpenAI and Microsoft partner with state attorneys general to create a national AI safety task force. The effort centers on developing shared guidelines to curb risks such as deepfakes and algorithmic bias while supporting responsible innovation, marking a major step in public–private alignment on AI governance. Click here
• EU signals delay on enforcement of the AI Act amid industry and U.S. pressure. The Artificial Intelligence Act may see grace periods and postponed fines for high-risk and generative-AI systems as the European Commission weighs easing obligations under pressure from Big Tech and the U.S. government. Click here

Global AI Strategy

• “AI Cold War” narrative hardens as U.S. – China rivalry accelerates. The Wall Street Journal reports that China is rapidly scaling its AI ecosystem through DeepSeek R1, national cloud efforts and Alibaba’s $53 billion AGI pledge, while the U.S. maintains an edge in frontier models and chips even as the gap narrows. Click here
• Baidu unveils new AI chips (M100 & M300) and launches Ernie 5.0 as China doubles down on self-sufficiency. The company revealed the M100 inference chip for early 2026, the more powerful training-and-inference M300 for early 2027, and a new multimodal Ernie version, aiming to challenge overseas tech dominance and provide a domestic alternative for AI workloads. Click here
• BlackRock’s GIP and ACS create a 50-50 joint venture for a 1.7 GW global data center portfolio. The partnership accelerates the “assetization of AI compute,” channeling large-scale institutional capital into the data-center layer of the AI economy and positioning compute as a stable, globally essential infrastructure asset. Click here
• India advances its ambitious “AI Sutras” blueprint through the IndiaAI Mission. The strategy outlines democratized GPU access, stronger techno-legal accountability frameworks and nationwide education initiatives to drive equitable innovation and close the AI divide while positioning India as a global AI leader. Click here

📱Social Signals

The most important conversations in AI are unfolding across social media, where top voices are shaping the next wave of signals and strategy. Here are some of the top social signals and their takes from the past week.

• Satya Nadella (Click here) — “Fairwater exemplifies our vision for a fungible fleet, infra that can serve any workload, anywhere, on fit-for-purpose accelerators and network paths, with maximum performance and efficiency.” Nadella positions Azure as a planet-scale AI superfactory, one elastic pool of compute that spans sites over an AI WAN. Fairwater’s two-story, liquid-cooled design and flat network pack coherent clusters of hundreds of thousands of NVIDIA GPUs per site, with more than 100,000 GB300s coming online for inference this quarter.
• Sophie (@netcapgirl) (Click here) — “SoftBank sold its NVIDIA stake to fund companies whose biggest cost is buying from NVIDIA?” Her post went viral on X, highlighting the irony, capital exits NVIDIA through SoftBank’s sale, then cycles into AI startups that spend it on NVIDIA GPUs. The take was amplified by finance magnates and LinkedIn voices, underscoring a circular flow of money that could still benefit NVIDIA.
• Harry Stebbings (Click here) — “AI app companies are meaningfully different than SaaS companies.” Stebbings, summarizing Everett Randle’s first interview as a Benchmark GP, argues that inference lives in COGS, so founders should be judged on gross profit multiples and absolute gross profit per customer, not classic SaaS revenue multiples. He adds that mega funds optimize for capital velocity and that coding assistants are the near term battleground, with distribution tilting toward OpenAI.
• Andrej Karpathy (Click here) — “Self-driving will visibly terraform outdoor spaces and way of life.” Karpathy frames autonomy as a city-scale shift, fewer parked cars and lots, higher safety, less noise, and more human space, with programmable delivery changing logistics costs.

To go deeper, subscribe to my monthly Founder Insights newsletter, where I share lessons from the frontlines of company building, perspectives on AI's future, and our industry's road ahead: https://www.linkedin.com/newsletters/founder-insights-7274531066957217793/

↓ In the comments below, share which areas in AI you want covered next week.