This week’s Spotlight is:Silicon is Back in Silicon Valley
This week’s Spotlight is:The Founder Boom
The latest Pitchbook NVCA data for Q2 2026 shows that while capital is more concentrated than it has ever been, more new companies are getting funded than at any point in the past decade. This creates a strong opening for founders with a clear AI thesis and a credible team. 5,674 companies raised their first venture round in 1H 2026. At that pace, more than 11,000 will raise a first round this year. A record.
A few other takeaways from the latest report are below:
AI now defines the venture market AI companies raised $354.9 billion in the first half of 2026. That represented 86% of every venture dollar, up from 65.5% in 2025. This is not a cycle cresting. It is a structural shift that is speeding up.
VC firms raised almost a full year of capital in six months Venture firms raised $72.4 billion across 405 funds during the first half of 2026, nearly matching the $74.9 billion raised in all of 2025, but across far fewer funds. A venture firm’s ability to write the first check matters. Its ability to support a company through later rounds matters just as much. Founders should understand the firm’s fund size, reserves, remaining capital, investment pace, and ability to raise its next fund.
Starting is easier, but differentiation matters more AI tools will help more people build companies. They will also create more competition. A durable advantage may come from proprietary data, a unique workflow, strong technical architecture, trusted distribution, innovative business model, deep domain knowledge, or the ability to learn faster from customers. The market will reward founders who use capital with discipline and build efficiently.
The San Francisco Bay Area is the AI epicenter In the first half of 2026, the San Francisco Bay Area captured $319.4 billion in deal value, roughly 77% of all US venture dollars. The Bay Area’s dense network makes it easier for founders to meet co-founders, employees, customers, technical peers, and investors. Those relationships help teams hire, learn, and move faster.
Record exit value hides a narrow market Venture-backed exit value reached $2.18 trillion across 874 exits in the first half of 2026. Even without the SpaceX IPO and xAI transaction, first-half exit value would have reached about $237.6 billion. That is already 84% of the $284.1 billion recorded during all of 2025. The IPO window has started to open, but only for a small number of companies.
My advice to founders – if you have a clear insight and the drive to move fast – this is one of the best moments in the AI era. It’s still early in the AI supercycle. The companies that will lead the next decade are being founded now.
Highlights from this week’s news include OpenAI sharply cutting prices for two GPT-5.6 models, Moonshot AI releasing the 2.8 trillion-parameter Kimi K3, SK Group and NVIDIA unveiling a $500 billion-plus AI infrastructure partnership, and Anthropic’s models successfully breaching three organizations during controlled cybersecurity evaluations.
Full Weekend Edition below. 👇
Signals Shaping the Future of AI:
Infrastructure
SK Group and NVIDIA announced a $500 billion-plus partnership spanning AI factories and next-generation memory. The initiative includes a 2-gigawatt AI cloud in Korea and a long-term HBM partnership with SK hynix to support global compute demand. Click here
Supermicro launched a new portfolio of high-density, liquid-cooling-optimized racks for AI data centers. The systems are designed to speed deployment, support heavier GPU clusters, and reduce total cost of ownership, with capacity to ship up to 3,000 racks per month. Click here
NVIDIA in talks to backstop $250 billion in financing for OpenAI’s Ohio data center. The proposed guarantee would support a 10 gigawatt SoftBank-developed campus near Columbus, with total project costs including chips potentially exceeding $500 billion; terms are not yet finalized. Click here
AMD signed a 15-year, $14 billion-plus data center capacity deal with Core Scientific. The agreement secures more than 500 megawatts of AI-ready capacity starting in 2027, with an option to expand to 2.5 gigawatts through 2028. Click here
Amazon, Google, Meta, and Microsoft have spent more than $1 trillion on AI infrastructure since 2023, with another $745 billion expected in 2026. The spending surge is accelerating demand for data centers, power, memory, and storage across the AI supply chain. Click here
Enterprise
NVIDIA forms Open Secure AI Alliance with Microsoft, CrowdStrike, Hugging Face, and Dell. The coalition will share tools and open model weights for AI agent safety and cybersecurity, formed days after an OpenAI agent autonomously breached Hugging Face’s systems. Click here
Microsoft is positioning its own AI models, agents, and chips as lower-cost alternatives to OpenAI and Anthropic. Satya Nadella said enterprises should keep models interchangeable and separate from the agent layer to avoid vendor lock-in and retain control of their data. Click here
OpenAI cut prices for two GPT-5.6 models as enterprises become more cost-sensitive. Terra pricing fell 20%, while Luna dropped 80%, reflecting intensifying competition to deliver more intelligence at lower cost. Click here
Microsoft launches MAI-Cyber-1-Flash and the Project Perception agentic security system. Microsoft’s first cybersecurity model pairs with its MDASH remediation harness to find and patch vulnerabilities, matching rival performance at half the cost, with public preview beginning next week. Click here
Cognizant became a Global Premier Partner in Anthropic’s Claude Partner Network. The expanded deal embeds Claude across Cognizant’s manufacturing, life sciences, and insurance platforms, with more than 30,000 employees already trained on the model. Click here
Amazon is winding down most of its Nova AI models to focus on a single frontier-model effort. The pivot, led by researcher Pieter Abbeel, deprioritizes Nova Premier, Omni, Reel, and Canvas ahead of a new flagship model debuting at re:Invent. Click here
Capital Flows
NVIDIA commits $5 billion to Ilya Sutskever’s Safe Superintelligence. The investment gives SSI access to NVIDIA’s next-generation Vera Rubin platform to scale compute roughly tenfold, while NVIDIA gains visibility into the lab’s closely guarded research; SSI is valued at $32 billion. Click here
Nscale is acquiring Anyscale for a reported $1.65 billion to expand across the AI compute stack. The deal adds workload orchestration and scaling software to Nscale’s energy, data center, and cloud infrastructure businesses. Click here
Okta agreed to acquire AI identity security startup Permiso for about $200 million. The deal expands Okta’s ability to detect threats involving AI agents and other machine identities as enterprises deploy more autonomous software. Click here
Moonshot AI closed a $3.5 billion round that lifted its valuation to $35 billion. The Chinese AI lab beat its original $1 billion to $2 billion fundraising target and is now planning a Hong Kong IPO as soon as 2026. Click here
Eliyan raises $145 million Series C at a $1 billion valuation for AI chip interconnect technology. The startup licenses chiplet technology designed to ease data transfer bottlenecks between AI chips as compute demands scale. Click here
Multiverse Computing raised a $570 million Series C at a $1.7 billion valuation. The Spanish startup’s tensor-network technology compresses large language models by up to 95 percent to cut inference costs, with backers including Qatar Development Bank. Click here
Research
Moonshot AI releases full Kimi K3 weights, the largest open-weight model to date. The 2.8 trillion parameter model is now downloadable, tunable, and self-hostable under a Modified MIT license, eleven days after its initial launch. Click here
Anthropic’s AI models successfully breached three organizations during controlled cybersecurity evaluations. The tests highlight the growing autonomous capabilities of frontier models and intensify concerns around AI security and misuse. Click here
Google DeepMind reassigned most of the AlphaFold team’s original paper authors over the past year. Nearly a quarter of the Nobel Prize-winning team’s full-time authors have left as staff shifts toward Gemini, enzyme design, fusion, and genomics work. Click here
GPTZero finds AI hallucinations in four PwC Middle East reports. The AI detection firm’s prior investigations led EY and KPMG to retract reports containing apparent AI-generated hallucinations, raising fresh governance questions for professional services firms deploying AI. Click here
Policy
More than 1,100 employees from leading AI companies are urging the U.S. government to support an international framework for pacing frontier AI development. The initiative calls for technical and governance tools to manage risks as advanced AI systems become more capable. Click here
FCC bans imports of new Chinese-made humanoid robots and power inverters. The agency cited national security risks, including documented backdoors in Unitree hardware already deployed at MIT, Princeton, and Carnegie Mellon; existing authorized models are not retroactively banned. Click here
The EPA said power plants dedicated solely to data centers are exempt from a key federal pollution law. The agency’s interpretation excludes off-grid AI power plants from Clean Air Act acid rain limits, easing the path for faster data center buildout. Click here
OpenAI and Anthropic are lobbying the Trump administration to extend mandatory government safety reviews to competitors. Sources say the two labs, whose own frontier models already undergo the voluntary pre-release review process, want lagging rivals held to the same standard before public launch. Click here
Global AI Strategy
China begins mass production of homegrown DUV lithography machines. State-backed manufacturer Shanghai Aishengna starts producing domestic immersion DUV chipmaking tools, planning about five units this year and twenty in 2027 for delivery to SMIC, Hua Hong, and CXMT. Click here
The US plans to award GlobalFoundries $300 million in CHIPS Act funding for AI chip infrastructure. The Commerce Department grant backs research into silicon photonics technology meant to speed data links inside AI data centers. Click here
NAVER, NVIDIA, and Brookfield will spend $10 billion to quadruple a South Korean AI data center. The deal expands NAVER’s national AI factory from 55 to 200 megawatts by 2028, with NVIDIA investing $1 billion for a stake in NAVER. Click here
Taiwan detained an NVIDIA employee in a widening probe into AI chip smuggling to China. Prosecutors allege forged documents were used to route roughly 50 Super Micro servers with advanced NVIDIA chips to China via Japan. Click here
China warned of countermeasures against potential US sanctions on Chinese AI firms. Beijing’s commerce ministry called the threatened sanctions, aimed at alleged distillation of American models by firms including Moonshot AI, acts of AI hegemony. Click here
Talent Signals
Each week, we spotlight key roles tied to the themes shaping this week’s AI headlines, connecting talent to the companies driving the news.
HangTenSystems is building an AI-native enterprise services platform that helps large companies build, modernize, and operate core software using agentic code generation and reusable AI skills. Following its $32M seed round led by Mayfield, Hang Ten Systems is growing its engineering and enterprise delivery teams. Learn more about the company and opportunities. Click here
Figure is building general-purpose humanoid robots designed to address labor shortages and perform unsafe or repetitive industrial work. Following its $675M Series B to accelerate commercialization, the Sunnyvale-based company is hiring across AI, robotics, hardware, software, and operations roles. Open roles are listed on its careers page. Click here
Outreach is building an AI-powered sales execution platform that helps revenue teams improve productivity and customer engagement. Outreach is hiring across revenue operations, customer success, engineering, and enterprise sales roles. Open roles are listed on its careers page. Click here
You can see all the opportunities at Mayfield-backed AI companies 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) — “Open-weight models are essential to a healthy AI ecosystem. Together with others across our industry, we are outlining a path for open-weight models to strengthen American competitiveness and expand economic opportunity, while protecting national security.” Nadella shared a Microsoft-backed paper calling for a policy framework that supports open-weight AI while addressing security risks. The argument positions open models as critical infrastructure for innovation, economic growth, and US competitiveness alongside frontier closed systems.
Dwarkesh Patel (Click here) — “New blog post on what would be true about the world if the trendline continues and a leading lab hits $1T in revenue by the end of next year. In other words, why compute might get 10x+ more expensive in coming years.” In a post viewed 1.4M times, Patel argues that as models become better at monetizing compute, demand could outpace supply and drive GPU prices sharply higher. The result could strengthen the advantage of frontier labs, increase the premium on model efficiency, and price lower-value AI workloads out of the market.
Mark Zuckerberg (Click here) — “The defining question of our age isn’t whether superintelligence will exist, but who will have access to it. Will it be centralized and restricted to a few institutions, or will it be a tool that empowers everyone? Rather than centralizing this power, we believe that delivering personal superintelligence to everyone is the way to answer this question.” Zuckerberg argues that widely distributed AI will create more invention, entrepreneurship, and individual agency than a system controlled by a small number of institutions. For founders, his vision points toward a future where powerful intelligence becomes broadly accessible and lowers the cost of building new companies.
