This week, the AI economy started behaving like a real economy. Companies borrowed billions against chips, rivals agreed on shared plumbing for shopping agents, and investors — public and private — began asking what the numbers underneath really mean. The through-line: accountability arrived faster than the hype cycle planned for.
1. SpaceX wants $40 billion in debt to buy Nvidia chips
- What’s happening: SpaceX is in talks to raise about $40 billion — roughly $10 billion in bank loans plus $30 billion in investment-grade debt — to buy Nvidia AI chips, a deal first reported by the Financial Times. The same week, Waymo closed its first-ever debt deal, a $5 billion term loan led by PIMCO, Blackstone and Sixth Street.
- Why it matters: The AI buildout is shifting from equity, which forgives, to debt, which doesn’t. Morgan Stanley estimates AI infrastructure will need $1.5 trillion in external financing by 2028.
- Zilck’s take: Debt disciplines faster than any regulator. Once coupons come due, “compute demand” has to convert into revenue on a schedule. Watch utilization rates replace ambition as the metric lenders — and soon everyone — care about.
2. GPT-6 lands — and OpenAI’s revenue takes a $20 billion haircut
- What’s happening: OpenAI rolled GPT-6 into ChatGPT on October 7 with “Intelligent UI” — answers built from charts, forms and calculators — for its 1.2 billion weekly users. A day later, the Financial Times reported OpenAI’s annualized revenue is approaching $50 billion, about $20 billion below the figure previously circulated.
- The friction: Nothing declined; the labs count differently. Anthropic books cloud-partner sales; OpenAI doesn’t — yet AI stocks, including Nvidia, Oracle, and CoreWeave, sank on the report.
- Zilck’s take: When a definitional footnote moves public market caps, the sector is being priced on vibes-based accounting. The quieter, bigger news is the product: interactive answers turn ChatGPT from a chat window into a surface where transactions can happen.
3. Walmart, Shopify and Stripe back a passport for shopping agents
- What’s happening: Meta and Sierra published the Personal Agent Protocol on October 6 — an open, OAuth-based standard for how personal AI agents identify themselves to businesses — with Shopify, Stripe, Walmart, Genesys and Rocket on board and a v0.1 spec due later this month.
- The quote: “When customers send an agent, they expect the same service they’d get themselves,” said Stripe payments lead Kevin Miller.
- The e-commerce angle: Today, merchants see agents as unlabeled bot traffic. PAP turns them into identified, permissioned customers — guest or signed in, read-only or allowed to transact, on the merchant’s terms.
- Zilck’s take: Standards win by default adoption, and Walmart plus Shopify plus Stripe covers a vast share of US checkout. If you sell online, being readable to agents just moved from thought experiment to roadmap item.
4. Gartner: only 34% of shoppers would let AI pick their gifts
- What’s happening: A Gartner survey of 1,004 US consumers, released October 7, found 34% are willing to use GenAI for holiday gift recommendations — while 77% plan to shop in stores and 75% on Amazon.
- Context: Shoppers who use AI for gift inspiration consult 4.48 sources on average, versus 3.24 for those who don’t. AI is adding a discovery layer, not replacing one.
- The quote: “Marketers should not treat AI shopping tools as a standalone channel,” said Gartner senior principal analyst Emily Weiss.
- Zilck’s take: Read items 3 and 4 together: the infrastructure is sprinting ahead of consumer trust. The brands that win this holiday season will optimize for agents and humans at once — accurate, structured product data serves both.
5. Firmus pulls its $5 billion IPO after investors balk
- What’s happening: Nvidia-backed AI data-center operator Firmus withdrew its roughly $5 billion Sydney listing on October 8 — it would have been Australia’s second-largest ever — after tepid demand at a $30.6 billion valuation, nearly triple its $10.5 billion private mark from August.
- Why it matters: Public markets are the one AI venue still willing to say no. Firmus says it will now “pursue capital from the private markets” instead.
- Zilck’s take: A valuation that triples in eight weeks is a narrative, not a price. The counterargument — that volatility, not fundamentals, killed the deal — only holds if the next private round confirms the mark. Founders: a hot private round that sets expectations a public book won’t ratify is a liability, not a win.
The Narrative Thread
Five stories, one plot: the AI economy is institutionalizing. Debt is replacing equity, protocols are replacing scraping, audited definitions are replacing headline ARR, and buyers — retail investors and gift shoppers alike — are getting a vote. The public mood matches: a new AP-NORC poll found 64% of Americans think AI is developing too fast. And Amazon ran its Prime Big Deal Days sale straight into that trust gap this week, kicking off a holiday season where Gartner says AI will assist discovery far more than it closes sales.
The takeaway for founders, marketers, and operators is the same move three ways: run your business as if diligence has already started. Structure your product data so an agent can find and transact it, report metrics you could defend under someone else’s accounting, and plan growth against revenue a lender would believe. This week proved someone will eventually check — a bondholder, a protocol, a public book, or a shopper with 4.48 open tabs.
Discussion