July 10th Newsletter

Author

Da Yu
July 10, 2026

As of close on July 9th, 2026

This week’s news shows how AI is driving bigger capital commitments, sharper market reactions and rising infrastructure and legal risk across tech. SK Hynix raised $26.5B in a landmark debut, making it the largest U.S. market debut by a foreign company, while Samsung’s chip unit is reportedly headed for a record 2026, with Q2 profit at $58.4B exceeding Nvidia’s $53.54B as DRAM, NAND and HBM demand lifts profit pools. Dell shares rose 7%+ after President Trump’s Oval Office mention, spotlighting Dell’s $6B+ pledge and political endorsement risk. Microsoft is cutting 4,800 jobs, including up to 3,200 Xbox roles, with shares down 19% YTD. Elsewhere in Tech, Apple lost a court challenge over whether iOS and the App Store fall under the EU's Digital Markets Act. Meanwhile, Google’s data center power use rose 37% and emissions 18%; Meta faces a $1.4T state penalty demand and contractor scrutiny over teen chatbot tests; and Apple’s first foldable iPhone may be scarce, limited to 0.5M–1M units in Q3’26 and priced at a rumored $2,300–$2,500.

We’ve published our Monthly AI Deck for June 2026. Check it out here.

Top 5 AI Highlights

[1] 💸 OpenAI broadly released GPT 5.6 and ChatGPT Work, an AI agent that can gather context across apps and files to create documents, on Mac and Windows. Separately, OpenAI's deployment arm agreed to acquire Northslope, its second enterprise AI deployment acquisition after Tomoro, expanding its forward-deployed engineering bench as labs compete to turn models into production workflows. Former Instacart CEO Fidji Simo has announced to leave her full-time role as CEO of AGI deployment due to medical conditions. More from Axios, OpenAI, and WSJ.

[2] 🏗️ AI infrastructure dealmaking is changing at a very fast pace as Anthropic locked in a 20-year, 401 MW TeraWulf data center lease worth roughly $19B, Blackstone’s QTS abandoned a massive 2,100-acre Virginia campus after legal and community pushback, and Nscale secured a $900M revolver to fund global AI compute buildouts, highlighting both the capital intensity and execution bottlenecks shaping the data center race. More from Reuters, Bloomberg, and TNW.

[3] 🧠 Perplexity plans to use Nvidia's new Vera CPU for AI agent workloads, with executives saying it is a strong fit for autonomous coding tasks, while Nvidia's roughly $1T market-cap slide has pushed its valuation back toward pre-AI-boom levels. Separately, China plans to let top AI firms buy a limited amount of Nvidia H200 Chips to offset an AI chip shortage. More from Reuters, Bloomberg, and The Information.

[4] 🤖 Meta introduced Muse Spark 1.1, but the bigger market takeaway is Zuckerberg returning to X after a roughly 3-year hiatus to announce it, while Meta’s new paid Model API gives developers access to its AI models and signals a potential new monetization path beyond ads and infrastructure scale. More from X.

[5] 🚀 SpaceXAI and Cursor unveiled the Grok 4.5, a new AI model aimed at legal and finance tasks, moving Elon Musk's AI stack deeper into high-value enterprise workflows and turning Cursor from a coding distribution channel into a broader professional-agent wedge. More from Bloomberg.

✨Open Source AI Development

[1] 🔒 China is weighing curbs on overseas access to its most advanced AI models after meetings with Alibaba, ByteDance and Z.ai, potentially limiting future open and closed models, tightening startup funding rules, and raising global AI costs as Beijing treats frontier AI as a national security asset. Separately, U.S. lawmakers are probing the growing use of Chinese AI models inside American companies, focusing on data, security and influence risks as cheaper Chinese open models gain enterprise traction. More from Reuters and CNBC.

[2] 💾 Deepseek is reportedly designing an inference-focused AI chip to reduce reliance on Nvidia and Huawei, after raising its first outside capital raise of over $7B at a $50B valuation as Chinese AI firms navigate U.S. export controls and intensifying chip self-sufficiency pressure. More from The Decoder.

[3] 📈 Zhipu, now branded Z.ai, is reportedly preparing to sell about $4B of shares after a 1,500% rally, highlighting investor appetite for Chinese open-source AI leaders as GLM models, especially the latest GLM 5.2, gain global traction and Beijing weighs tighter control over frontier model access. More from Bloomberg.

[4] 🚫 Alibaba will reportedly ban employees from using Anthropic's Claude Code starting July 10, classifying it as high-risk software and directing staff to use Qoder instead as U.S.-China AI access, security, and model-distillation tensions keep escalating. This news came out a day after it was reported that Claude Code had a hidden tracker quietly flagging users’ timezone, proxy, and potential connection to Chinese AI labs. More from TechCrunch and Ars Technica.

[5] ☁️Tencent officially launched Hy3, a 295B-parameter MoE AI model with 21B active parameters and 256K context, open source under a commercially friendly Apache 2.0 license, claiming flagship-level performance at lower cost as token usage rose 20x since preview. The hy3-preview is currently No.7 on OpenRouter in terms of token usage. More from Tencent.

Founder’s Corner

Seed Valuations are Flying

Source: X

What We Read This Week

  • SemiAnalysis says Meta may have the best shot at catching OpenAI and Anthropic as its superintelligence push combines a $14.3B Scale AI talent raid, 3,000 engineers focused on RL environments, and five 1GW+ AI “titan” clusters, making compute, data, and talent the core market watch items for Meta’s AI capex story. More from SemiAnalysis.

  • Meta plans to put its AI chip into production in September as it looks to double computing capacity, signaling another step toward reducing Nvidia dependence and lowering long-term inference/training costs as its 2026 AI capex ramps. (Reuters).

  • Anthropic’s Pentagon relationship reportedly frayed after months of emails between CEO Dario Amodei and Undersecretary Emil Michael over AI safety guardrails, exposing tensions between national security demands and responsible AI controls as defense adoption accelerates. (WSJ). 

  • Anthropic's rise is helping reshape New York's AI scene as capital, talent, and real estate attention shift toward frontier-model companies, showing how the AI boom is spreading beyond San Francisco while raising questions about whether NYC can sustain a durable AI startup ecosystem. (New York Times).

  • Nvidia’s Kyber NVL144 rack-scale AI system for Rubin Ultra chips has reportedly been delayed to 2028 due to PCB midplane manufacturing challenges, potentially giving AMD and Google a rare high-end AI infrastructure opening, though Nvidia says its roadmap remains intact and SemiAnalysis still sees data-center compute revenue running 20% above consensus in 2H FY2027. More from CNBC. 

  • Crusoe is in talks to raise about $3B in a new funding round that could value the AI data center operator at around $30B, roughly triple its $10B October valuation, underscoring surging investor appetite for AI infrastructure tied to major customers like Meta and Oracle. More from Bloomberg. 

  • Sysdig says JADEPUFFER may be the first agentic ransomware operation, exploiting an unpatched Langflow flaw to autonomously steal credentials, create admin access in 31 seconds, encrypt 1,342 configuration entries, and expose how weak credential controls can turn old security gaps into machine-speed extortion risk. (The Decoder).

  • Midjourney is pushing Hollywood studios to disclose broader internal AI usage in its copyright fight with Disney, Universal and Warner Bros., arguing the records could support its fair-use defense and highlighting how generative AI risk is becoming a strategic media issue. (TechCrunch).

  • Apptronik opened a 90K-square-foot Robot Park training facility in Austin where its Apollo 2 humanoids can collect real-world reps across logistics, manufacturing, and retail tasks, creating data the company says is already shaping Apollo 3. (YouTube).

  • Iluvatar CoreX is seeking to raise about $850M after a stock rally, highlighting investor demand for Chinese AI chip challengers as Beijing pushes semiconductor self-sufficiency and local alternatives to Nvidia. (Bloomberg).

  • Spotify asked Kalshi and Polymarket to remove its logo and clarify that there is no partnership after detecting 500,000+ artificial streams that boosted Malcolm Todd’s “Earrings” and helped settle a $3M Kalshi music-chart market, intensifying concerns that prediction markets can incentivize manipulation of real-world data. (Bloomberg). 

  • Bramshill Investments co-founder William Nieporte is suing after being fired for allegedly ignoring the $8B asset manager’s five-day return-to-office policy, claiming the mandate was used to strip him of a 12% stake and at least $30M in value, highlighting how RTO enforcement is reaching senior leadership and triggering governance disputes. (WSJ). 

  • Lenovo launched a ~$41 AI Student Phone in China with no games, browser or social media, combining a homework-help AI button with location tracking, classroom mode and parental controls as AI moves deeper into education-focused consumer hardware. (CNBC-TV18).

AI Fundraising News (July 2 — July 9)

  • Alsa: AI agent payments company, raised a $6.5M seed

  • Alta: AI sales agent automation platform company, raised a $25M Series A

  • ARMS Innovations: AI warehouse robotics orchestration company, acquired by Symbotic at an undisclosed price.

  • Bespoke Labs: AI agent training environments company, raised a $40M seed and Series A

  • Biren: AI chip company, raised an $892.5M HK IPO

  • Broccoli AI: AI operations platform company, raised $25M

  • Common Room: AI-native buyer intelligence, acquired by Zoom at an undisclosed price.

  • EdVisorly: AI admissions workflow automation company, raised a $13.3M Series A

  • EquiLibre: AI quant hedge fund company, raised a $125M Series A

  • Fleek: AI secondhand apparel pricing company, raised a $25M Series B

  • Hakimo: AI video monitoring company, raised a $12M Series A2

  • Kaon AI: AI-personalized story worlds company, raised $60M

  • Katalyze AI: AI pharma manufacturing agent company, raised a $10.5M seed

  • Kling AI: AI video generation company, raised $2.8B

  • LinqAlpha: AI investment research company, raised a $22M Series A

  • Lovable: AI text-to-software development company, raised $300M

  • Luffy AI: AI adaptive industrial controls company, raised a $10.8M Series A

  • Lyzr: AI enterprise agent startup company, raised a $100M Series B

  • Monogram: AI interactive interface generation company, raised a $40M seed

  • Mowito: AI robotic task learning company, raised a $3M pre-seed

  • Norm AI: AI legal agents company, raised a $120M Series C

  • Ollama: AI local model deployment company, raised a $65M Series B

  • Polysense: AI food factory inspection company, raised a $10.7M seed

  • Positron: AI chip startup company, raised $750M

  • Prime Intellect: AI agent infrastructure company, raised a $130M Series A

  • SambaNova: AI chip company, raised a $1B Series F

  • Savi Security: AI scam detection company, raised a $7M seed

  • Sherpa AI: AI privacy-preserving model training company, raised $18M

  • Skello: AI frontline workforce management company, raised $200M

  • Tangos: AI financial crime investigations company, raised a $20M seed

  • THEA: AI risk markets company, raised $8M

  • Tripo AI: AI 3D foundation models company, raised a $150M Series A3

  • Velocity: AI app monetization infrastructure company, raised a $27M seed

  • Worldmodeldata: AI game-based training data company, raised a $9.3M seedries B

  • Venice AI: AI private model access company, raised a $65M Series A