Megawatts, Money, and Moats - AI Update October
OpenAI reached a $500 billion valuation. That makes it the world’s most valuable private company, far ahead of Elon Musk’s SpaceX and TikTok’s parent ByteDance (as of October 2025).
OpenAI reached a $500 billion valuation That makes it the world’s most valuable private company, far ahead of Elon Musk’s SpaceX and TikTok’s parent ByteDance (as of October 2025). The company’s ambition now extends far beyond paid seats. It is wiring a full-stack commerce engine into ChatGPT, aiming to convert conversational queries into direct purchases. The goal is to master the journey from vague user intent to specific product, a place where today’s social discovery feeds still struggle. This could be a boon for Etsy and Shopify merchants that live in the long tail. But this vision is gated by a brutal physical reality: it requires an AI power grid of unprecedented scale. This has forced a sophisticated and risky $100 billion partnership with NVIDIA, built on complex financial engineering and a high-stakes bet on the future of energy. Oh, and did we mention that new deal with AMD, where OpenAI will take up to a 10% stake? 1. ChatGPT goes after commerce Based on an internal forecast reported by The Information in April, OpenAI targets $125 billion in revenue by 2029. ChatGPT alone is expected to reach $60 billion in revenue by 2029, primarily from paid subscriptions (Plus, Pro, Business, and Enterprise). It’s not unprecedented if you use Microsoft’s Productivity & Business suite as a comp ($120 billion in revenue in FY25), but it would require universal adoption in the corporate world at a time when competition is fierce in agentic work. For perspective, Meta needed 12 years to go from ~$4 billion to ~$135 billion in annual revenue, across a family of apps used by 3+ billion people each month. Source: Fiscal.ai The bigger story is monetizing the free audience, starting with native shopping. Why assistants can move the needle Search and social feeds push you toward what’s already popular. An assistant starts earlier in the funnel. You describe a problem, not a brand. That’s where long-tail inventory wins. The assistant can surface a perfect niche product you didn’t know to search for, and then close the loop inside the chat. What was announced so far Instant Checkout inside ChatGPT: Users can discover and buy without leaving the assistant (rolling out with Etsy listings in the US, with Shopify next). Even if the first version of checkout is basic, ChatGPT would offer the shortest path to fixing a problem with a purchase. It’s the on-ramp to monetizing the free audience with commerce. Merchant rails and payments: OpenAI published an open standard so sellers can make their products shoppable inside ChatGPT. When multiple merchants offer the same item, enabling Instant Checkout becomes the way to be the link that converts. OpenAI says results aren’t pay-to-play, but the incentive is clear: wire your catalog and let the agent complete the funnel. Upstream supply & siting: OpenAI announced memory-supply letters in South Korea (Samsung, SK Hynix) and discussions about hosting data centers there. It’s a wafer-to-watt hedge. The goal is to lock HBM and locations so inference capacity actually shows up as adoption grows. That brings us to the massive deal with NVIDIA. 2. The OpenAI-NVIDIA deal OpenAI is building an AI factory of unprecedented scale, and NVIDIA is fronting the $100 billion bill. The two giants plan to deploy millions of GPUs, totaling at least 10 gigawatts of computing power, to train OpenAI’s next-generation models. The build-out will be progressive, with NVIDIA’s investment unlocked as each gigawatt comes online. The first phase is slated for the second half of 2026, launching on NVIDIA’s upcoming Vera Rubin platform. This partnership underlines a simple, brutal truth of the AI boom: it’s a game of staggering capital intensity with no guaranteed payback. It also reveals why a company worth over $4 trillion, like NVIDIA, needs needle-moving deals, even if it means financing a key customer to lock in demand at a planetary scale. The deal is a strategic financial loop. OpenAI pays NVIDIA cash for hardware, and in return, NVIDIA reinvests that cash into non-voting equity in OpenAI. This ties NVIDIA’s financial return directly to the successful deployment of its own hardware. An initial ~$10 billion tranche will be triggered once definitive purchase agreements are signed. Translation: Capacity is financed in stages, upside is shared, and governance remains firmly with OpenAI. The scale: This is utility-grade power. The plan is a multi-year ramp that aligns NVIDIA’s hardware and software roadmap with OpenAI’s model stack, with commitments tied to real-world deployment milestones. The context: OpenAI is simultaneously de-risking its infrastructure plan by signing massive, parallel deals. Its contract with CoreWeave has ballooned to over $22 billion, while projects with Oracle and SoftBank are securing physical locations. Together, these moves push the broader Stargate initiative toward a multi-gigawatt scale, ensuring it has locked down compute, power, and real estate from a diverse set of…
Read on MPC Markets Mosaic
