Money, plumbing, an ultimatum: OpenAI lowers its price, Agent Skills joins the Linux Foundation, Meta and October 30

Three signals this Friday: money, plumbing and an ultimatum. The largest private fundraise in history just lost 150 billion dollars of valuation. The agent economy just got a neutral home. And Meta gives businesses seven days to decide what their content is worth.

Signal one: money

OpenAI is closing a raise of up to 30 billion dollars at a pre-money valuation of 1.25 trillion. The round would be led by Coatue and investors from the United Arab Emirates, and OpenAI told investors it wants signatures by the end of the following week. A separate commercial deal would cover more computing capacity in the UAE. OpenAI plans 600 billion dollars of compute spending by 2030.

The story is in the gap. On October 20 I reported, citing Reuters, that OpenAI was seeking the same 30 billion at a valuation above 1.4 trillion, to set its price before Anthropic's IPO. Bloomberg had reported that figure a month earlier. A few days and a more nervous market later, 150 billion has evaporated from the asking price.

Why it matters

Valuation is a weather report. At this price, OpenAI would become the most valuable private company in history, and it still had to lower its price. The room has cooled. The Nasdaq fell 2.1% this week, its worst stretch since June, on a specific worry: the giants are financing their AI data centers with debt, and the debt keeps growing. Meta fell 4.4%, Alphabet 2.9%, Micron and Intel more than 5%.

Both sides of the wall

Two days after Washington ordered federal civilian agencies to remove DeepSeek's technology from their systems, Bloomberg reported that the Chinese lab was raising about 15 billion dollars, one of the largest rounds in Chinese history, with CATL and Tencent among potential investors. A ban on one side. A record check on the other.

Meanwhile Anthropic, which raised 65 billion at a 965 billion valuation in May, is reportedly targeting 1.5 to 2 trillion for its IPO.

The AI frontier has become a technology race and a capital race at once, funded in part by sovereign money.

What changes

For anyone choosing an AI vendor, the purchasing question widens. You buy a model, and with it a balance sheet, a funding pipeline and a geopolitical camp. Ask who funds the model, because the answer tells you which doors will stay open.

Signal two: plumbing

On October 22 in San Jose, Anthropic announced it is donating Agent Skills, the open standard that gives AI agents know-how and working methods, to the Agentic AI Foundation, under the Linux Foundation.

It was not an isolated gesture. At the same event:

• OpenAI contributed its Open Responses standard;
• Hugging Face contributed OpenEnv;
• Alibaba contributed OpenSandbox.

Four companies that fight every day just agreed on common rules. The foundation now hosts ten projects, up from three a year ago, including the Model Context Protocol (MCP), the protocol that lets models connect to tools and data, donated by Anthropic in December 2025.

An analogy helps. The internet works because nobody owns HTTP or email: they are shared standards. Agents need the equivalent to know how to learn a trade (skills), how to respond, how to run in a safe environment. These four projects cover those layers.

Why it matters

This is where the agent economy is really decided. Launch keynotes make headlines. Boring infrastructure makes the market. More than 70 platforms have integrated Agent Skills since December. Anthropic says developers have published a million skills, installed nearly 280 million times. Those are Anthropic's figures, not independently verified, but the direction is clear: how agents learn new trades is becoming shared plumbing.

The uncomfortable question

Why do companies give their technology to neutral foundations? Usually so the whole industry can build on it without fear. The donation includes the name, the logo, the GitHub organization and the agentskills.io domain, and anyone can use or contribute without joining the foundation. Anthropic is telling the market: stop wondering who owns it, and build. A standard everyone adopts is worth more to its creator than a proprietary one few adopt.

What changes

For brands, the agent layer stops being an experiment and becomes infrastructure. A product that cannot be found, described and used with confidence by someone else's agent, through shared standards, risks not existing in the channel that will matter tomorrow.

The question for every brand team has changed. Building for agents is a given. What remains is choosing which standards to build on. I would start with a simple question for your technical team: are our products, prices, availability and terms accessible to an agent through a standard method, or only to a human who clicks?

Signal three: the ultimatum

On October 30, Meta's new commercial terms take effect. And they change what your website is for.

The clause

Any business with a Meta business account, meaning anyone running a Facebook page, an Instagram profile or a WhatsApp business channel, will be deemed to have authorized Meta's automated crawlers, using "test credentials," to access and analyze the domains and URLs associated with its business. Meta may use that content to provide, develop, improve, research and maintain its products, AI included.

An update on the October 17 episode

On October 17, I described Meta backing down after Nieman Lab's report. The retreat is real, but partial. Meta removed the word "train" after publishers objected. According to lawyers, the crawling permissions survived. Three attorneys who reviewed the text for Nieman Lab believe it still grants broad crawling rights, potentially behind logins and paywalls, with no opt-out. The terms say it bluntly: if you do not agree, stop using Meta products for business purposes.

So I correct my reading: the line moved less than the first statement suggested.

Why it matters

This is the content bargain, rewritten. A media lawyer called it remarkable to bury such a clause in terms of service, when that access would be worth millions of dollars in a licensing market. And Meta charges some accounts, through its Meta One subscription, for the right to post external links to the very content it crawls for free.

What changes

Every brand now needs two content strategies. The first, for its customers. The second, for machines, because everything on a website becomes raw material for someone else's model.

Before October 30, the practical steps are simple:

1. check the robots.txt file and what it allows;
2. decide what the brand would never want used, and put it out of reach;
3. reread the terms of every platform where the brand has a business account, with the October 17 question: what are we giving away that is new?

The thread of the day

The price of intelligence, marked down. The agent economy, placed under neutral governance. And your website, seven days from becoming training data.

Three movements that answer one another: money turns cautious, standards get shared, and content becomes the asset everyone wants without paying for it.

The noise filtered. The signal kept.