Huang confirms a $1B Hugging Face retention plan — and rival bidders

Share
Huang confirms a $1B Hugging Face retention plan — and rival bidders

Jensen Huang and Clément Delangue sat down together on CNBC's Squawk Box the morning after the deal, and the joint interview answered the questions the announcement blog post did not: what the price was set against, whether the talent is being paid to stay, and how Nvidia plans to explain owning the neutral ground to the labs that compete with it.


Huang confirmed on live television that there were other bidders for Hugging Face, and that $12.9 billion was simply the clearing price. "There are other bidders, and $12.9 billion is what it took to close the deal, and it's worth every single penny," he said — and when Becky Quick pressed him on who else was in the running, he shut it down with a line that will get quoted for a while: "It doesn't matter who the other bidders were. It only matters who wins." Delangue separately confirmed he approached Nvidia first, and declined to name the others. That reframes Wednesday's number: it wasn't a valuation Nvidia arrived at, it was a price it paid to win an auction.

He also confirmed the rumored retention package outright. Asked directly whether reports of a plan worth up to $1 billion to keep Hugging Face's staff were accurate, Huang answered "Yes," then added that "the talent is everything" and that Hugging Face is "all about the talent." Delangue started to say the company wasn't commenting before Huang talked over him. A nine-figure-per-year retention pool tells you what Nvidia thinks it actually bought: not three million models, but the few hundred people who maintain the platform everyone else depends on.

The more interesting thread was the security argument, which both men made at length. Delangue said the models used in this summer's attack on Hugging Face were unreleased and closed — "locking everything behind closed doors doesn't work" — and argued that open weights flatten the asymmetry between attacker and defender. Huang reached for the Linux analogy: there are far more people defending than attacking, so giving defenders frontier-capable open models produces what he called an asymmetric advantage. It's a self-serving case from a company that sells the chips either way, but it lands one day after OpenAI shipped a model whose chain of thought its own safety reviewers reportedly struggle to read. Delangue set the next target at 100 million builders on the platform, up from 18 million today.

Our read: two of the three things Huang said were reassuring, and one was a tell. The neutrality structure — Hugging Face keeps running as an independent platform, Nvidia compute stays optional — is the same language from Wednesday's blog post, now spoken aloud by the CEO, and it's worth exactly as much as it is enforceable. But "it only matters who wins" is the real message, and it's the one that should set the tone for regulators: Nvidia didn't buy Hugging Face to grow it, it bought it so nobody else could own it. We laid out why that matters when the deal was still rumor — Nvidia signs the Hugging Face deal at $12.9 billion.


Saudi Arabia's Humain is raising a $2.5 billion fund to build out data center capacity, Bloomberg reports. The vehicle is expected to be managed by BSF Capital, the investment arm of Banque Saudi Fransi, which will approach investors after clearing Saudi's Capital Market Authority — a process that could take two to three months. Financing is expected to mix debt and equity.

The money is earmarked for 250 megawatts Humain is building with Al Moammar Information Systems, a footprint that could eventually scale toward a gigawatt. Humain CEO Tareq Amin confirmed a fund was coming during LEAP 2026 earlier this week without giving a number, so this is the first real signal of how big Riyadh wants it to be. The pattern is now familiar: the Gulf is no longer just buying GPUs, it's building the financing structures — funds, vehicles, infrastructure debt — that let outside capital co-own the buildout rather than merely rent it.


What to watch: whether Hugging Face's neutrality gets written into anything binding, and whether Humain's fund clears the Capital Market Authority on the two-to-three-month timeline.

Is an auction-winning price the right way to decide who owns the open-model commons? Tell us in the comments.

Read more

Byteification turns Qwen3 and Llama 3 into byte-level models

Byteification turns Qwen3 and Llama 3 into byte-level models

The tokenizer is the last piece of the language-model pipeline that was never learned end to end — and a Nature paper out this week shows it can be swapped out for less than one percent of a pretraining budget. Researchers have retrofitted four open subword language models into true byte-level models using less than 1% of a typical pretraining budget, and the result — published 7 October in Nature — closes the performance gap that has kept byte-level models a niche curiosity for nearly a decad

OpenAI's dots go mobile and can start your Codex work

OpenAI's dots go mobile and can start your Codex work

One product update and two benchmark papers this hour, all circling the same question: how much of the work can you actually hand over, and what happens to the part you didn't? OpenAI's dots — the always-on agents that get their own cloud computer — can now be created and configured entirely from the ChatGPT app on iOS and Android, and a dot can start Codex work or pick up an existing Codex thread on your behalf. The October 9 release, billed as the first of a fresh run of updates for dots, mo

Coding agents as a hospital: Cockroach's five-month results

Coding agents as a hospital: Cockroach's five-month results

The agentic-coding debate keeps being framed as speed versus trust. Cockroach Labs just published the most detailed trust-first ledger yet — five months of numbers, including the parts that didn't work. Cockroach Labs ran its codebase like a teaching hospital for five months and says the pipeline merged well over a million lines of code with seven total reverts. In the setup the company calls MOLT Sinai, GitHub issues are patients, merging is discharge, and the humans in charge are Chiefs of M

Today in AI — October 10, 2026

Today in AI — October 10, 2026

A day where the boring machinery took center stage: proof checkers, order books, warehouses and the humble text message — plus fresh arguments about who gets to declare any of it safe. Models & Research * Mathematicians get a reliability primer for the Lean Theorem Prover. A guest post by Thomas Hales on Terence Tao's blog walks through what Lean actually guarantees — and devotes itself to the "Summer of Soundness Bugs," the string of kernel bugs found this summer that let false proofs thr