Nvidia wants to loan OpenAI $250 billion to buy Nvidia chips.

This edition of the Future Blueprint was brought to you by:

Hi AI Futurists,

OpenAI wants to build a data center that costs more than $500 billion. Nvidia has offered to guarantee the bill, then lend it another $350 billion to buy the chips that go inside. One famous investor already has a name for this. Let's take a look.

Our agenda.

Best,
Lex Sokolin

P.S. Hit reply with suggestions or feedback!

The biggest challenge isn't the apps you use. It's the work that happens between them.
Eney is a proactive AI assistant that works across the tools you already use.

It connects to Google Workspace, Notion, Confluence, Jira, Dropbox, and more to understand your context, surface what needs your attention, and complete tasks with your approval.

From planning your day and following up after meetings to organizing files and updating your work across apps, Eney helps you stay focused on meaningful work instead of managing it.

Try Eney Free >

🧠 Moonshot dropped the biggest open-weight AI model ever (TECHi). Kimi K3 is 2.8 trillion parameters and 1.4 terabytes, free to download. Your laptop is not invited.

🕵️ The White House says Moonshot copied a US model to build Kimi K3 (CyberScoop). Officials claim it distilled Anthropic's Fable and trained on restricted chips routed through Thailand. The open-weight gift came with a receipt.

🤖 Anthropic shipped Claude Opus 5 at the old price (MarkTechPost). It beats OpenAI's Sol on the hardest reasoning benchmark and holds a million-token memory, still $5 per million tokens. Same sticker, sharper knife.

☁️ Microsoft is rationing compute and serving itself first (The Next Web). Its own Copilot products get capacity before Azure cloud customers do, even as it lifts sales quotas 30%. Selling seats it cannot fill.

🇨🇳 Moonshot and DeepSeek are both racing to go public (Fortune). Moonshot eyes a Hong Kong listing near $50 billion, DeepSeek wants Shanghai at up to $71 billion. The price war has an IPO tab now.

💼 OpenAI launched ChatGPT for Small Business (9to5Mac). Its work agent and Codex crossed 10 million users, and OpenAI is running in-person AI academies for owners. The corner store is the new enterprise.

⚡ OpenAI is planting a 3.2-gigawatt data center in rural Georgia (CBS News Atlanta). Project Camellia would power up in phases through 2032, with $80 million pledged to local residents. Every model needs a hometown now.

🦾 Europe got its first humanoid-robot unicorn (RoboticsTomorrow). London's Humanoid raised $152 million at a $1.35 billion valuation, with Bosch and Schaeffler backing it. The robots are European now.

🔩 A chip startup hit $10.3 billion a month out of stealth (CryptoBriefing). Etched raised $300 million from Sequoia for chips that run only transformer models, with $1 billion in contracts already booked. Betting the whole factory on one AI shape.

🧪 An AI-for-materials startup raised $450 million (Electronics Weekly). Cambridge's CuspAI hit a $2.6 billion valuation designing materials nobody has made yet, from carbon capture to chips. AI is done writing emails, now it invents matter.

OpenAI is pursuing a 10-gigawatt data-center campus in Piketon, Ohio, on the site of a former Cold War uranium-enrichment plant. The project could cost more than $500 billion. OpenAI does not have the investment-grade credit rating lenders typically want for financing on that scale, so Nvidia has reportedly offered to guarantee about $250 billion in lease and construction debt. SoftBank’s energy arm would build the site.

A separate proposal would have Nvidia finance another $350 billion in chip purchases for OpenAI. Nvidia would be financing the chips that fill a building whose debt it is also guaranteeing.That structure is the source of the concern. Nvidia has used variations of this arrangement before: equity investments in AI startups, support for customers’ cloud spending, and now real-estate guarantees. The recipient then spends heavily on Nvidia hardware. The money returns as Nvidia revenue, supporting its valuation and creating room for the next deal.

Michael Burry put the criticism bluntly after the news broke: “Around and around we go.” He then increased his bet against Nvidia. The question is whether demand for Nvidia’s chips is entirely independent, or whether Nvidia is helping create part of the demand it reports. That matters beyond the stock market. Much of the AI economy rests on the assumption that infrastructure spending will continue rising. If OpenAI becomes Nvidia’s long-term anchor customer, the arrangement could secure years of demand. If the financing proves too aggressive, the consequences would reach companies and valuations built around the same buildout.

Model costs are falling. The physical infrastructure behind them is getting more expensive, and an increasing share of that cost appears tied to Nvidia’s balance sheet.

Adomate: turn your own data into winning ads

For marketers and DTC brands who need fresh ad creative without an agency.

Robynn AI: a website that audits and improves itself every week

For founders and marketers who own a site but have no time to tune it.

Estera: an AI receptionist for calls and WhatsApp, around the clock

For small businesses and founders losing leads to missed calls.

This 158-page research report provides the first comprehensive taxonomy of public companies, private ventures, and tokenized protocols building the infrastructure for autonomous AI systems. Compiled by Lex Sokolin, former Chief Economist at ConsenSys, fintech strategist at Autonomous Research, and current Managing Partner at Generative Ventures, this report delivers institutional-grade analysis of 100+ companies across 14 critical infrastructure layers. Learn more here.

That’s all for today, folks!