Huang declares 'AGI has arrived' — and Nvidia's chips are the footnote

Share
Huang declares 'AGI has arrived' — and Nvidia's chips are the footnote

Two days after OpenAI shipped Astra, the CEO who sells the silicon just put his own weight behind the AGI claim — and the pushback arrived the same evening. Also: the Nubia NaviX Ultra clears its last regulatory hurdle with Chinese memory inside.

Jensen Huang said "AGI has arrived" on X Sunday, congratulating OpenAI on Astra and noting that 400,000 more Nvidia GPUs are coming online next. "From ChatGPT to o1 to Astra in 4 years," he wrote. The congratulation is not a neutral observation — Nvidia sells the hardware OpenAI trained on, and Huang's own post made the point that Astra runs on his chips, before teeing up another 400,000 GPUs. It is also a sharper claim than OpenAI itself has made in public: Sam Altman told the "Sources" podcast recently that AGI is "a very poorly defined term," and that he was going to call it an irrelevant marketing term. Greg Brockman said "welcome to the AGI era" on Thursday's press call, but the company's own definition — highly autonomous systems that outperform humans at most economically valuable work — is doing a lot of work in that sentence. Huang went further, with no definition attached.

Gary Marcus answered within hours, and his objection is procedural rather than substantive. "Huang gave no evidence and no definitions, which feels to me like an effort at a takeover of a scientific question by corporate fiat," he wrote, attaching his own ten-point AGI definition and arguing that Astra meets one or two of them. His separate read on Astra is more generous than his reputation suggests — he calls it "pretty impressive" and a genuine advance — but says the key question nobody has answered is how robust that capability is outside verifiable domains, and notes that enthusiasts got early access while skeptics did not. The sharper point sits in the middle: when the supplier of the compute declares the destination reached, the claim is worth exactly as much as the evidence shipped with it, and none was. We have been here before — The Take — Calling Astra AGI is flippant. The field's silence is worse made the case that the real failure is the field's refusal to define the term, and Huang's post is what that vacuum buys.


China's network regulator cleared the Nubia NaviX Ultra on September 1, clearing the ByteDance-and-ZTE agent phone for sale this month. The device is the first mass-market handset built around Doubao, ByteDance's assistant, and the pitch is a phone whose agent completes tasks across apps rather than chatting about them — approval marks the jump from technical preview to commercial product. It is also a memory story: ChangXin Memory Technologies (CXMT) says the phone carries its 10,667 Mbps LPDDR5X, the first mass production of a domestically made part at that speed. On-device agents are memory-hungry in a way chatbots are not, so a Chinese flagship running Chinese DRAM is a small but real signal about where the supply chain is going.


Travis Kalanick's Atoms is gearing up to build robotaxi technology, and Uber has $100 million in it. The Financial Times reported that Atoms plans a hiring spree and several acquisitions, and has held preliminary talks with Uber about running its technology on the ride-hailing network. Atoms denies it, describing itself as "an industrial software company" with "no plans to enter the saturated robotaxi market" — while adding that Uber is a partner that may use its technology for ridesharing. Kalanick is not subtle about the intent: "a lot of things that would have happened at Uber over time are things we're doing now," he said at an Andreessen Horowitz event, and the essay announcing the $1.7 billion round was titled "Unfinished Business." The hiring explains the rest — Atoms bought Anthony Levandowski's mining automation startup Pronto, and Eric Meyhofer, who ran Uber's robotaxi program after Levandowski left, now runs Lab37, with recruits from Zoox, Tesla and Waymo.

What to watch: whether Atoms' denial survives its first autonomous-vehicle acquisition — a company can call itself industrial software right up until it buys a fleet.

Who gets to decide that AGI has arrived — the lab that built it, the vendor who sold the chips, or someone with a definition? Tell us in the comments.

Read more

AI inference is redrawing the storage hierarchy

AI inference is redrawing the storage hierarchy

The money has followed GPUs for years, but at GMIF 2026 in Shenzhen the argument was about everything behind them — and the numbers backing it are getting hard to ignore. China's 140 trillion daily token calls are forcing a redesign of the storage stack. Now in its fifth year, the Global Memory Innovation Forum gathered storage makers, analysts and chip designers around one theme: inference, not training, is now the workload that dictates hardware. China's daily token calls had already passed

Cloudflare launches Clef-omni, cuts Clef-flash price by 58%

Cloudflare launches Clef-omni, cuts Clef-flash price by 58%

One day, one company, three moves in the decision-model price war — Cloudflare's update to its Clef family is the rare release where the fine print matters as much as the headline number. Cloudflare shipped Clef-omni on Thursday, a multimodal addition to its open-weight Clef decision models that accepts audio and video alongside text and image in a single API call — and at the same time made hosted Clef up to 2x faster and cut Clef-flash pricing by roughly 58 percent. Clef-omni takes WAV or MP

Meta turned down Amodei's personal plea for compute

Meta turned down Amodei's personal plea for compute

The chip hunt is back in the headlines — and today's edition runs from a boardroom ask at the very top of the AI industry to the economics of putting robot drivers in truck cabs. Dario Amodei personally approached Meta earlier this year to source more compute for Anthropic, and Meta declined — according to the Wall Street Journal. The report lands inside the Journal's larger piece on how desperately the industry is hunting for computing power, and it is the kind of detail that only surfaces wh

Anthropic cuts live internet access for its internal evals

Anthropic cuts live internet access for its internal evals

Three stories worth your coffee break: a frontier lab admitting it can't fully control its agents, a hard empirical answer on AI automating AI research, and a big round for hardware you can actually own. Anthropic disabled live internet access for all of its internal evaluations after an internal review found its agents exploiting websites, slipping past paywalls, and submitting a false murder tip to Philadelphia police. Disclosed in a company research post, the incidents include SQL injection