Daily Brief - SiliconNoon (Page 83)

The Arena

Micron: HBM uses 3x the wafer of DDR5, and the gap keeps growing

AI's appetite for memory is colliding with a hard physical limit, and the chipmakers are running out of room to hide it. Two stories from the last 24 hours — one from a conference stage, one from a balance sheet — show the same problem from different sides. Micron used its Hot Chips 2026 presentation to make the case that the silicon cost of HBM, the stacked memory every AI accelerator relies on, has quietly tripled against ordinary DDR5, and the gap is widening with every generation. A single

The Arena

Unisound's agent business clears 5B yuan in H1, token revenue jumps 500%

Two Chinese AI stories landed today that point in opposite directions on what "working" looks like at scale: a Hong Kong-listed industry-agent company posted real growth on a Palantir-style model, and one of the country's largest healthcare AI products inked a deal to plug authoritative Western medical journals into its doctor-facing app. Hong Kong's "AGI first stock" delivered a clean H1: revenue of 5.62 billion yuan, up 38.7% year over year, with the agent business alone accounting for roughl

The Frontier

StreamPI teaches VLA models to think in time, not frames

Two robotics papers out of China this week are quietly rewriting what "embodied AI" actually has to know. A VLA model from Da Xiao and HKU learns to read continuous time, and a single-video pipeline from Niuwa, SJTU, and Shandong lets a robot push through a door it has only ever seen on tape. Both push past the single-frame ceiling — and both suggest the next wave of robot intelligence is less about bigger models and more about the dimensions they were ignoring. StreamPI gives VLA models a work

The Arena

OpenAI's Codex ships a 'Persistent' mode that runs until you stop it

The always-on agent race got a real code drop today. OpenAI is wiring a new "Persistent" setting into Codex's reasoning-effort menu — the closest a frontier lab has come to shipping a product around the idea of an agent that doesn't politely end its turn after a few minutes. Codex in Persistent mode is told to "continue working until put to sleep," according to a file in the Codex core that WIRED spotted in a recent pull request. That phrasing is doing the real work: today's reasoning modes sto

The Frontier

Tencent open-sources Hunyuan Hy4 preview — 770B params, 1M context

China's open-weight race just got its heaviest new contender, and Tencent is pitching it as a productivity workhorse rather than a benchmark-chaser. Tencent is open-sourcing Hy4 preview, a 770 billion-parameter Mixture-of-Experts model with a 1 million-token context window. The architecture activates only 49 billion parameters per token — a 78-layer stack where each layer beyond the first holds 256 routed experts plus one shared expert, with eight experts firing per token. It ships under the A

The Frontier

Ex-OpenAI VP: human AI researchers have 'two years' left in the driver's seat

The man who built OpenAI's o1 and o3 reasoning models now runs a lab whose stated goal is to make researchers like him obsolete — and he has put a number on how fast it will happen. Jerry Tworek, the former OpenAI research vice president who led the o-series reasoning models, said in a recent interview that human researchers have roughly two years left to play a meaningful role in AI research before the entire loop — from idea generation to validation — is automated. Tworek left OpenAI in April

The Arena

Nvidia pauses revenue-sharing deals with AI cloud firms

Nvidia's plan to become more than the chip supplier to the AI boom — to finance the data centers themselves and share in their revenue — just hit a speed bump. Nvidia has paused the revenue-sharing and credit-support deals it struck with AI cloud companies, people familiar with the matter told The Wall Street Journal, with Reuters confirming the move. The program, first announced in July, was Nvidia's bid to co-build "AI factories" with cloud providers — supplying Blackwell GPUs while taking a

The Frontier

Amap's ABot-Recon turns 12 frames into a 10,000-frame 3D scene

Two signals out of China this hour — one a new kind of spatial-intelligence model, the other proof that a frontier model can now run cheaply on domestic chips. Alibaba's Amap has released ABot-Recon, a streaming 3D-reconstruction model that rebuilds 10,000-frame scenes while looking at only 12 frames at a time. Announced August 28 and open-sourced with weights on Hugging Face and ModelScope, the model takes a single monocular RGB video stream — no depth sensor, no pre-known camera parameters —

The Arena

Alibaba Cloud opens Brazil data center, its second Latin America region

Alibaba Cloud has switched on its first data center region in Brazil, the company's second in Latin America. The new region, launched August 27, runs two availability zones offering compute, storage, and database services to enterprise and developer customers, with enterprise-grade AI services set to roll out in phases. It follows Alibaba Cloud's Mexico region, which went live in February 2025 — and which the company is already preparing to expand beyond its single availability zone. The expans

The Everyday

RayNeo's 34-gram AI glasses break the smart-glasses first-sale record

The camera-free AI glasses bet just got its first real proof of demand. RayNeo's iO is topping sales charts across China's three biggest e-commerce platforms within days of launch. The iO, a 34-gram pair of camera-free smart glasses priced from 1,996 yuan (about $280), swept the smart-glasses sales charts on JD.com, Tmall and Douyin in its first sales window, RayNeo announced August 27. The TCL-backed AR maker says the device broke 2026's record for a new smart-glasses product launch, calling i

The Frontier

Manycore's spatial-intelligence world model enters internal beta

The company behind Kujiale — one of China's biggest interior-design platforms — just made its clearest declaration yet that it wants to be infrastructure for the spatial-AI era, not just a design-software vendor. Its first interim results since listing show the AI business is now the only part of the company growing fast. Manycore Tech said its self-developed world model has entered internal beta, anchoring a pivot the company itself frames as becoming the "water seller" of spatial intelligence

The Guardrails

Judge strikes down Pentagon's blacklist of Anthropic

A federal judge has thrown out the Pentagon's designation of Anthropic as a supply-chain risk, handing the Claude maker a decisive win in one of the most closely watched legal fights between an AI lab and the US government. A California federal judge ruled Thursday that the Pentagon's blacklisting of Anthropic was unconstitutional, vacating the supply-chain risk label that Defense Secretary Pete Hegseth had placed on the company. US District Judge Rita Lin found that officials retaliated agains

See all