Chips & Infrastructure

The Arena

Today in AI — October 19, 2026

Monday was a capacity day: Meta published the infrastructure behind its new models, Intel finally named its inference card, and the policy side kept tightening around Chinese models and federal networks. Models & Research * Tencent's agents rebuild their own environment. A Tencent Hunyuan Frontier paper, written with researchers from NUS, Maryland, Georgia and Indiana, argues that most agent failures are environmental rather than architectural: real workspaces scatter evidence, mix in…

The Stack

Postgres 19 delayed to November 19 — AI code review found too many bugs

Three releases, three different reasons to stop and look: a database delayed because AI review found more bugs than anyone could fix in time, an export-control draft that decides which models China may ship, and a browser assistant that answers without being able to read the question. PostgreSQL 19 will not ship this month.

The Stack

Houmo moves compute inside memory — and A20 enters production

Beijing chipmaker Houmo used its partner conference this week to lay out two compute-in-memory routes at once — and to say the first chip built on the newer one is already in production. Houmo's A20, its first chip built on a 3D compute-in-memory architecture, has entered production. The part supports up to 16 chips fully interconnected, claims a 50% improvement in compute utilization over the previous generation, and covers models from 3B to 30B parameters, according to co-founder and…

The Frontier

Brain organoids classified Braille at 83% — in a trio, not alone

Two stories today: human brain tissue on a chip decoded a tactile signal, and the data-centre company behind some of the largest AI contracts in Europe turns out to be six months behind on its flagship site. Human forebrain organoids grown on a microelectrode array classified Braille letters with 61 percent accuracy on their own and 83 percent when the responses of three organoids were combined, in a peer-reviewed study from the University of Bristol.

The Frontier

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute

Microsoft finished pre-training its biggest in-house model yet — then used it to make a point about the company it bankrolls. Microsoft AI completed pre-training of MAI-Base-2, a mixture-of-experts foundation model with 80 billion active and 2.8 trillion total parameters, and says the recipe is 4.2–12.3x more pre-training FLOPs-efficient than DeepSeek-V4-Pro-Base.

The Guardrails

Zero-click spyware rode ad auctions into 45 Al Jazeera phones

Researchers now have hard technical proof that simply showing someone an ad can hand their phone to a spyware vendor — and the ad network that did it was carrying the industry's own "certified vendor" stamp for six years. Also today: Google's cross-retailer Universal Cart went live in the US. Citizen Lab says dozens of devices at Al Jazeera were infected with mercenary spyware from Candiru, now rebranded Integrity Labs, through ordinary targeted ads — no tap, no download, no prompt.

The Stack

AI 101 — What is a chiplet?

A chiplet is a piece of a processor that is manufactured as its own small chip, then packaged alongside other small chips so that the finished part behaves like one big chip. Instead of printing one enormous processor, you print several modest ones and wire them together inside a single package. Every flagship AI accelerator shipping today is built this way. It matters right now because the constraint on AI hardware has moved.

The Arena

TSMC's sixth straight record quarter — 58.7% margin, 68% from HPC

The chip cycle's upstream end and its downstream end told two different stories on the same morning: TSMC booked a sixth consecutive record, while the lab that buys its wafers told investors its growth rate is falling fast. TSMC's third-quarter net profit rose 68.7% to NT$763.1 billion (about $23.9 billion), a sixth consecutive record, on revenue of NT$1.494 trillion — up 50.9% year over year and roughly $46.7 billion.

The Stack

Linus Torvalds ships his first LLM-assisted kernel patch series

Torvalds used Claude Code to write a kernel patch series — and the kernel's own reviewers found a race in it within a day. Also: Anduril prices a $1.3 billion employee tender, and the Baker lab publishes two Nature papers on AI-designed enzymes. Linus Torvalds has posted his first patch series written with a large language model — and he says so in the commit message.

The Guardrails

Nvidia puts $250M of compute behind UK safety labs

Two halves of the same hour: Nvidia turned a European tour into a safety-research funding pledge, and Brussels went looking for a sanctions trade to keep a lithography supplier alive. Nvidia is donating $250 million of compute — $50 million a year for five years — to UK agentic safety and security research, delivered through UK-based Nvidia Cloud Partners with the infrastructure hosted in Britain.

The Arena

Salience Labs raises $125M for all-optical AI data center switches

The AI buildout's bottleneck moved from the processor to the wire, and today's money followed it: a British photonics startup pulled in a nine-figure round with both Intel and AMD on the cap table, while two smaller bets went to the less glamorous end of the agent economy. Salience Labs raised an oversubscribed $125 million Series B led by Bullhound Capital, with Intel Capital and AMD Ventures both participating — two chipmakers that compete for the same accelerator sockets, funding the same…

The Guardrails

Deep Dive — Drones just proved AI compute is a wartime target

Four days of Ukrainian drone strikes took out half of Yandex's data centers, including the buildings that hold all three of its training supercomputers. The rest of the industry is now asking a question it had been avoiding: what happens when the machines that make AI are also the easiest thing to hit. The four days Russia started it.

See all

Today in AI — October 19, 2026

Postgres 19 delayed to November 19 — AI code review found too many bugs

Houmo moves compute inside memory — and A20 enters production

Brain organoids classified Braille at 83% — in a trio, not alone

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute

Zero-click spyware rode ad auctions into 45 Al Jazeera phones

AI 101 — What is a chiplet?

TSMC's sixth straight record quarter — 58.7% margin, 68% from HPC

Linus Torvalds ships his first LLM-assisted kernel patch series

Nvidia puts $250M of compute behind UK safety labs

Salience Labs raises $125M for all-optical AI data center switches

Deep Dive — Drones just proved AI compute is a wartime target

Get SiliconNoon in your inbox