Top News
Nvidia forms AI security industry alliance: The chip making announced the Open Source AI Alliance, a collective that will bring together representatives from Adobe, CrowdStrike, Hugging Face and Dell Technologies. Their post-OpenAI sandbox incident goal: building shared tools for identifying and capturing rogue agents. Nvidia will contribute access to its open source project, Object-Oriented Agent, to help companies monitor and keep tabs on their own agentic activity. In an open letter last week, co-signed by a number of major tech firms, Nvidia argued that blanket restricts on open weight models are doomed to backfire, concentrating excessive power in a handful of closed providers.
Also, Nvidia drops major investment in Safe Superintelligence: The chip company invested a “substantial” but undisclosed amount into the AI lab from former OpenAI chief scientist Ilya Sutskever, as part of a long-term overall partnership. (According to WSJ, Nvidia reached the deal after getting a sneak peek into the relatively covert state of the startup’s research.) Safe Superintelligence will now receive access to a lot of GPU hardware — enough to increase its resources by “orders of magnitude.” There’s not a TON of insight available yet into exactly what Sutskever and his team are building, but their public goal is “straight shot” research into the a secure form of AGI. Essentially, AI that can do whatever a human could do, without the unpredictability of pesky consciousness or free will.
Finally, Nvidia may team on OpenAI data center: The chip company that makes all the headlines is also in negotiations to guarantee $250 billion in new financing for a 10-gigawatt OpenAI data center in Southern Ohio. The project already has backing from Japan’s SoftBank, and could carry with it a total price tag of over $500 billion. (As OpenAI lacks an investment-grade credit rating, this backstop allows it to borrow on favorable terms.) Nvidia also has plans to finance another $350 billion for OpenAI chip purchases, effectively backing its own sales, in a structure that certain skeptical investors have mocked as overly circular: “around and around we go.”
TWiST 500
Bloomberg reports that Chinese open source AI giant DeepSeek has told prospective investors that it’s suspending its second fundraising round for now. Backers had expected to sign agreements in the coming days. The company may resume the process at some point in the future.
The likely trigger, though this is not yet confirmed, was an alleged leaked transcript that went viral online, purportedly capturing DS co-founder Liang Wenfeng speaking candidly with investors about China’s reliance on Nvidia chips, and the country’s continued lag behind frontier US AI models. Bloomberg has yet to verify the transcript’s authenticity, and DeepSeek has not commented. Still, the timing feels a touch suspect.
The pause comes fast on the heels of DeepSeek’s first-ever round, which closed in June and raised a total of $7 billion from backers including Tencent and CATL, valuing the company at roughly $50 billion. Round 2 was aiming even higher, at a $67 billion valuation. DeepSeek has also been prepping for a possible IPO coming as soon as this year, a landmark for any Chinese AI company of its size.
The stakes here obviously go beyond any one company’s cap table. DeepSeek has been treated as clear proof that Chinese open weight labs can compete with the best models from Silicon Valley, despite US expert controls on the most advanced chips. Any stumble in its fundraising path could be read as a signal about the state of the broader global AI race. – Zach
A message from Superhuman
Get the AI that works where you work. Find out more at superhuman.com.
This Week in Startups
E2317: Most lab-grown meat companies have already gone under, even those with a lot of funding and early buzz. Cultivated meat is expensive to produce, and so far, there’s little demand from consumers who fear the “mad science” connotations of growing tissue in a lab. But Japan’s IntegriCulture is still standing, in part because of an innovative deal with cosmetics companies that license their technology. Go deep on this little-discussed niche industry with IntegriCulture founder Dr. Yuki Hanyu.
E2316: We’ve been told that quantum computing has been “5-10 years ago” for 30 years. On this edition of TWiST Tokyo, Jason sits down with Yaqumo CEO Kazu Nakashoji to discuss why now may be the moment that things start to change. Find out why quantum machines are so error prone, why quantum computers have such an unconventional supply chain, and and go deep on the cryptography reckoning that may be coming for your Bitcoin.
E2315: In the first episode of our week-long summer trip to Tokyo, Jason is joined by CEO Bilal Kharouni of Ekei Labs. He’s using AI and epigenetics to develop longevity treatments in Okinawa, one of the world’s legendary Blue Zones noted for its residents’ tendency to live long, healthy lives. He discusses why a smaller raise can go much further in Japan than Silicon Valley, why the best moat for biotech companies is not an AI model but their data, and why Ekei killed a burgeoning side business in order to maintain focus.
TWiST Partner Offers
Every.io: For all of your incorporation, banking, payroll, benefits, accounting, taxes or other back-office administration needs, visit every.io.
MongoDB: AI-assisted and agentic coding is helping you build faster than ever. Start building at MongoDB.com/ai
Agree.com: Stop chasing invoices and automate your entire contract-to-cash stack. Go to https://agree.com and tell them Jason sent you to get 50% off for life!
Subscribe to Us On Your Favorite Podcast Platforms
SF Live-Work Space Now Available
Need a flexible living and working environment in San Francisco? This thoughtfully designed loft-style residence at 787 Bryant St., the heart of the vibrant SOMA district and the city’s creative hub, is now available for rent or purchase. Check the listing for more details.
The TWiST500 newsletter is the new, updated, and improved TWiST Ticker.
