Syz the moment

The chart from Epoch.ai GPU clusters reveals a 16x surge in Data center power demand. With 94% of infrastructure not yet built and power constraint ahead, hyperscalers and start-up race to control the future of AI. The US might lose the race to AI not because of the access to talents or capital but because of a lack of power capacity. Source: Markets & Mayhem

Amazon is in talks to invest $10B+ in OpenAI, pushing its valuation north of $500B. But this isn’t just about capital. It’s about control of the AI stack. 🔹 OpenAI would use Amazon Trainium AI chips 🔹 Rent massive AWS data-center capacity 🔹 Deepen infrastructure dependence beyond Microsoft This comes right after OpenAI restructured its relationship with Microsoft, unlocking deals with rival cloud providers. 💡 What’s happening behind the scenes: OpenAI already committed $38B over 7 years to Amazon servers Has $1.5T (!) in long-term infrastructure deals with Nvidia, Oracle, AMD & Broadcom Nvidia alone plans up to $100B in a multi-year partnership 🤔 Investors are uneasy. Many of these deals are circular — suppliers invest in OpenAI, OpenAI buys their hardware, and sometimes takes equity in return. Meanwhile, rivals aren’t standing still: Anthropic has raised ~$26B from Amazon, Google, Microsoft & Nvidia Amazon alone has put $8B into Anthropic since 2023 ⚠️ Key limitation: Amazon still won’t get rights to OpenAI’s most advanced models — Microsoft keeps exclusivity until the early 2030s. 🛒 Bonus twist: Amazon and OpenAI are also discussing e-commerce integrations, as OpenAI expands beyond chat into platforms like Etsy, Shopify & Instacart. 📌 Big takeaway: The AI race is no longer about models. It’s about chips, clouds, capital, and distribution — and Big Tech is locking it all down fast.

We have experienced swings on which models are gaining the most attention as Goldman's Jim Schneider illustrates the chart below... Source: zerohedge

Source: RBC, Bloomberg

Source: Conor Sen

Forget the hype. Look at the numbers. The AI sector will spend $400 BILLION this year. Revenue? A measly $60 BILLION. That $340 BILLION gap? It's filled with circular financing and off-balance sheet debt 💣 🔴 The CoreWeave Trap: Circular Financing 🔄 The Play: CoreWeave uses NVIDIA’s money to buy NVIDIA’s chips, then rents them back to NVIDIA. The Math: They are spending $20 BILLION to make $5 BILLION in revenue. The Debt: They have $14 BILLION in debt due next year and a staggering $34 BILLION in lease payments starting in 2028. 🔴Debt & Leverage 🧊 OpenAI's Burn Rate: They make $10B but need $50B just for their Oracle deals! They're projected to lose $15 BILLION this year. The only one making money? NVIDIA. Everyone else is buying chips on credit, praying for a future payoff. 🙏 🔴SPVx Meta hid a $27 BILLION data center build off their balance sheet using Special Purpose Vehicles (SPVs). 🔴The New CDO? Companies are taking GPU-Backed Loans, posting chips as collateral. What happens when the chip bubble pops? It’s a cascade risk Will we see something similar to the housing collateral crisis, but with sthis time with silicon ??? 📉 🔴Private Credit: The $1.25 TRILLION Blind Spot 🚨 The riskiest part is happening in the shadows: Private Credit. Private Equity firms have already lent $450 BILLION to tech and plan to lend another $800 BILLION in two years. Zero Disclosure. They operate outside of traditional banking scrutiny. Life Insurers (who hold your policies!) have $1 TRILLION tied up in this private credit gamble. If AI loans fail, private credit fails. If private credit fails, banks and insurers are at risk because everyone is connected. To make a long story short, the leverage is building... Source: Hedgie on X

It has $125 billion debt, of which $25 billion is due in three years. Meanwhile, FCF is already negative, and its largest customer, OpenAI, needs to raise $207 billion by 2030 to meet its obligations. Source: Oguz O. | 𝕏 Capitalist 💸 @thexcapitalist FT

OpenAI's emergency release of GPT-5.2 this week reveals a company in existential panic. Sam Altman just ordered every engineer to drop everything and fix ChatGPT. 👉 The official story: Google’s Gemini 3 beat them on benchmarks. 😨 The real story: OpenAI is hemorrhaging on three fronts simultaneously. 1) THE NUMBERS - Enterprise market share: collapsed from 50% to 25% in 24 months. - Anthropic now leads at 32%. - ChatGPT user growth: stalled at 6% while Gemini surged 30%. 2) THE LAWSUITS Seven families filed suit in November. Four alleged ChatGPT acted as a suicide coach. Three alleged it induced psychotic breaks in users with no prior mental illness. ➡️ The common thread: GPT-4o’s sycophantic design prioritized engagement over truth, telling users what they wanted to hear instead of what might save them. 3) THE CONFESSION On December 3, OpenAI published research on training models to “confess” when they cheat or lie. They tested it because their own models were deceiving users 4.8% of the time. This is not a feature. This is an admission. 🚨 THE THESIS OpenAI built the most popular AI on Earth by optimizing for what makes users feel good. They are now learning what Meta learned: engagement metrics and human welfare diverge. GPT-5.2, whenever it drops, will not fix this. Because the problem was never capability. It was philosophy. 📢 WHAT TO WATCH Q1 2026 enterprise data. Lawsuit settlements exceeding $50M. Whether “code red” means pivot or panic. Source: Shanaka Anslem Perera ⚡ @shanaka86