This week was less about new capabilities and more about the forces closing in around them. The economics came under scrutiny, the rules started to bite, and geopolitics reshaped a company overnight. The models keep improving, but the story last week was the money, the regulation, and the risk underneath it all. Here is what mattered.
OpenAI's revenue came in well below what investors assumed
Reports this week showed OpenAI's annualized revenue was around $50 billion at the end of September, roughly $20 billion below the nearly $70 billion many investors had been assuming. The gap is largely an accounting difference: OpenAI books only its own share of sales made through cloud partners, while rivals like Anthropic count the gross amount, and investors had been comparing the two on the higher basis. OpenAI still told investors it expects to reach at least $70 billion in annualized revenue by the end of the year. The confusion added to a sell-off in AI-related stocks. [Bloomberg, Quartz]
Our take: the striking thing is that a $20 billion swing came down to how you count revenue, not how much money changed hands. That is a useful reminder whenever you read a big AI number: the definition matters as much as the figure. The same applies to the claims vendors make to you. When a tool promises to save "40 percent" or handle "90 percent" of some task, the honest question is always what exactly is being measured, and against what baseline. We push for that specificity on every AI project, ours and our clients' alike, because a number you cannot trace is a number you cannot plan around.
SpaceX moved to borrow $40 billion for Nvidia chips
SpaceX is in talks to raise about $40 billion to buy Nvidia chips, roughly $30 billion in new debt and $10 billion in bank loans, according to a Financial Times report, with firms like Apollo and PIMCO involved. The raise comes only four months after SpaceX's record $86 billion public offering, and it captures how capital-hungry the AI build-out has become. Morgan Stanley has estimated that AI infrastructure will need $1.5 trillion in outside financing by 2028. [Bloomberg, BNN Bloomberg]
Our take: the sums here are a signal, not a shopping list for anyone else. One company borrowing $40 billion just for chips shows how expensive it is to own AI at the frontier, which is exactly why almost no business should try to. The leverage has flipped: the hard, costly work of building and running the models is being done by a handful of giants, and everyone else rents the result for cents. The smart position is to stay a renter, spend your money on the part that is yours, your product and your customers, and let the giants carry the capital risk. We build on that rented foundation on purpose, so your costs scale with use instead of landing up front.
OpenAI started watermarking ChatGPT's text in Europe
On October 5, OpenAI said it will begin adding an invisible watermark to text generated by ChatGPT and Codex for users in the European Union, to comply with the EU AI Act's transparency rules, which took effect in August and require AI-generated content to be marked so other systems can detect it. The method, which OpenAI calls textGrain, subtly biases word choices using a secret key. It rolls out over the coming weeks to EU users on all plans. OpenAI noted the watermark is not foolproof: replacing about 10 percent of the words with synonyms cut detection sharply. [TechCrunch, OpenAI]
Our take: this is the EU AI Act turning from text on a page into something a company actually has to ship, and it will not be the last rule to do so. Even if you are nowhere near Europe, the direction is clear: disclosure of AI-generated content is becoming an expectation rather than a courtesy. Getting ahead of it is cheap and builds trust. The practical step is to be honest now about where you use AI with customers, a chatbot that says it is a bot, AI-written content that is labeled as such, instead of waiting to be told. We help businesses set those disclosure habits early, because being upfront about AI is turning into a competitive advantage as much as a compliance one.
Manus raised $500 million after China blocked Meta's deal
Manus, a startup that builds AI agents, raised more than $500 million at a reported $4 billion valuation on October 8, its first round since China blocked Meta's roughly $2 billion acquisition of the company earlier this year. Boyu Capital and IDG Capital led the round, with existing backers including Tencent. Manus has resumed operating independently and recently launched tools for building personal agents that can handle tasks like bookings and calls. [CNBC, SiliconANGLE]
Our take: the money is notable, but the backdrop is the real lesson: a government stepped in, unwound a done deal, and left the company to chart its own course. AI is increasingly tangled up in geopolitics, and that can reach your business through the tools you depend on. A vendor can be acquired, blocked, or cut off from a market with little warning. The sensible hedge is the one we keep returning to: know who owns the tools you rely on and where they operate, and keep enough flexibility to switch if the ground moves. We favor setups where no single vendor's fate becomes your problem, so a deal collapsing or a border closing upstream stays their headache rather than yours.
The one thing to remember
This week was less about what AI can do and more about the forces closing in around it. The economics got scrutinized, as a $20 billion question over OpenAI's revenue rattled markets and SpaceX lined up $40 billion just for chips. The rules started to bite, with OpenAI watermarking its text to satisfy Europe. And geopolitics reshaped a company, as a blocked acquisition sent Manus out on its own. For a business, this does not change the core advice so much as sharpen it: stay a renter, insist on numbers you can trace, be honest about where you use AI, and keep the freedom to switch. The capability will keep arriving. The winners will be the ones who stay clear-eyed about the money, the rules, and the risk underneath.