Let me tell you a story about a number so large it stops sounding real. Nvidia — the chipmaker at the center of the artificial-intelligence universe — has agreed to pay $12.93 billion for Hugging Face, a platform for open-source AI models and datasets. Hugging Face generates roughly $150 million in annualized revenue. Do the division and you get a price of about eighty-six times revenue. For context, most healthy software companies trade at single-digit multiples. Nvidia just paid a multiple with two digits, for a company whose entire yearly revenue wouldn’t cover a rounding error on Nvidia’s own income statement.
This is not a story about one acquisition. It’s a story about what the AI boom has become: a land grab so urgent, so total, that the normal rules of valuation have been set aside like furniture before a flood.
The deal, announced September 3 and covered across this week’s newsletters, is one of Nvidia’s biggest ever. The strategic logic is actually legible: Hugging Face is where the world’s open-source AI models live — the shared workshop where researchers and developers build on each other’s work. By owning the workshop, Nvidia positions itself at the center of whatever gets built next. As Bloomberg’s morning coverage noted, the acquisition hints that open-source AI models could drive future demand for Nvidia’s chips. It’s vertical integration at planetary scale: own the models, sell the shovels.
But here’s what makes the week genuinely remarkable — the same days that brought us the Hugging Face deal also brought us the first serious warnings that the AI spending spree is getting dangerous.
S&P, the ratings agency, warned this week that AI financing is getting bigger and more complicated, and that the pristine credit ratings of the Big Tech hyperscalers are at risk. Read that again slowly. The most creditworthy companies on Earth — the ones whose balance sheets were considered bulletproof — are borrowing and committing so aggressively to fund AI infrastructure that the people who grade their debt are getting nervous. Axios’s Emily Peck flagged the S&P analysis, which uses an adjusted debt metric accounting for cash holdings, future lease obligations, and power purchase agreements — the hidden plumbing of the AI buildout.
And the scale of the commitments is staggering. A Morgan Stanley analysis, surfaced by Axios, estimates that Big Tech now carries roughly $3 trillion in off-balance-sheet AI infrastructure commitments — Alphabet, Microsoft, Nvidia, and their peers, promising trillions in future spending on data centers, chips, and power that doesn’t fully show up in the headline debt figures. Three trillion dollars. That’s not a budget line. That’s a bet on the future the size of a large country’s entire economy.
Oracle offered the week’s most vivid illustration of the tension. Its shares jumped in after-hours trading Thursday on strong revenue growth in the unit housing its AI hyperscaler business — the market loves the growth story. But S&P rates Oracle BBB-minus, just one notch above junk territory, and as Axios’s Matt Phillips noted, the company is “less financially strong than other tech giants sinking hundreds of billions into AI.” Growth is glorious. The balance sheet is the bill.
Then there’s Anthropic, the AI lab at the center of so many of this week’s stories. The Financial Times reported Sunday that Anthropic told investors it will be profitable for a second straight quarter, with gross margins above 80 percent — before accounting for revenue shared with distribution partners like Amazon and the enormous cost of training its models. Eighty-percent gross margins with an asterisk the size of a data center. Even the winners are spending at a pace that would make a drunken sailor blush.
And hovering over all of it is the essay that set the tech world arguing this weekend. Anthropic CEO Dario Amodei published “We Must Pace the Frontier” on his personal website Saturday, arguing that AI has grown “drastically faster” this summer, warning that within six to twelve months an AI “swarm” could be capable of taking over the entire internet and causing “hundreds of billions of dollars in damage.” He called for AI companies to accept “embedded evaluators” — outside overseers with desks in the office, access badges, and company laptops, like the examiners who sit inside big banks. Sam Altman, Elon Musk, and Demis Hassabis publicly backed him. Nvidia’s Jensen Huang scoffed, suggesting the industry was manufacturing hysteria to sell cybersecurity products: “What better way to create demand than to create a problem?” President Trump said he didn’t want to slow AI development, to keep ahead of China: “whoever wins AI wins.”
So here we are: the builders themselves are split between “slow down before it breaks the internet” and “speed up before China wins,” while spending trillions either way. If that doesn’t capture the strangeness of this moment, nothing will.
It’s not enough to just marvel at the thirteen-billion-dollar price tags and move on. It’s not enough to treat the AI buildout as someone else’s gamble — a story about billionaires and data centers that has nothing to do with us. We must listen to what the spending is telling us about where the economy is going, learn how the infrastructure being built today will shape the jobs and prices of tomorrow, and contribute our own judgment to the most important economic question of the decade: is this investment, or is it excess?
I want to be honest about both sides, because both deserve respect. The optimists see the railroads being laid — the defining infrastructure of the next fifty years, built by companies with the cash flows to fund it. The skeptics see the fiber-optic bubble of 1999, when trillions were spent laying cable for demand that took a decade to arrive, bankrupting the builders. Both can be true at different moments. The history of technology is the history of overbuilding followed by underappreciating what got built.
For the rest of us — the savers, the workers, the small-business owners watching from the sidelines — the practical wisdom is old and unfashionable: don’t confuse a boom with a plan. If you own the stocks, enjoy the ride but size your bets like someone who remembers 2000. If you’re building a career, notice where the money is flowing — into infrastructure, into power, into the unglamorous physical layer of the digital dream — because that’s where durable jobs live. And if you’re simply trying to understand the world your children will inherit, know this: the companies spending trillions today are betting that intelligence itself is about to get cheap. Whether they’re right or wrong, the attempt will reshape everything.
Thirteen billion dollars. One hundred fifty million in revenue. Eighty-six times. The number is absurd, and the ambition behind it is completely serious. That contradiction is the AI economy in a single deal — and it’s the story we’ll be living inside for years to come.
Written from the September 11–14, 2026 editions of Fortune CFO Daily, Axios Markets, Bloomberg’s morning coverage, the FT (via Reuters), Morning Brew, and WSJ What’s News. Deal terms, financial figures, and quotations are as reported by those outlets; my reflections are my own take.
