Thursday, November 5, 2026

Back to the front page

From X

@convequity 1 click

Convequity on X

AI swarms could make network security more important. Anthropic's October 12 research ran an AI model across 15 cheap Android TV boxes at 0.39 tokens per second. The small models still couldn't reliably complete the full replication cycle. I think we're still early. Inference on this hardware is painfully slow. But the idea is interesting because human-level intelligence isn't necessary for something to reproduce and spread. Think of insects: limited individual intelligence, yet populations can spread widely. An AI worm could carry its model with it, adapt to unfamiliar devices and use compromised hardware to run the next copy. A model too large for one device could be split across several. Better processors and more memory in future cameras, TV boxes and other connected devices would make more local inference possible, wherever attackers can gain access. That creates a security problem beyond installing software on individual devices. You're not going to put an endpoint security agent on every camera or TV box. You need to identify those devices through the network, restrict what they can communicate with and isolate them when compromised. A perimeter firewall alone cannot inspect traffic that never passes through it. I think $FTNT could benefit here. Its agentless device discovery and network segmentation fit this problem, especially across distributed sites. More capable AI on unmanaged devices could increase demand for those controls. The opportunity also extends to networking vendors that integrate security into switches and wireless access points, where they can control how devices connect and communicate.
@convequity 1 click

Convequity on X

$NVDA's falling unit share in China's data-centre AI chip market overstates how much business Nvidia has lost to Huawei. Three things explain why. 1. IDC counts units, and did not say whether that count includes the Blackwell parts barred from China and reportedly smuggled in. Chinese demand also reaches Nvidia outside China. Xiaomi or ByteDance rents a Blackwell cluster in Malaysia from a local cloud provider. That is Nvidia revenue from a Chinese buyer that never appears as a Chinese unit. So Nvidia's share of what Chinese buyers spend on AI compute is higher than its share of counted units. 2. Huawei's volume growth is real, and memory explains how. Huawei cannot get enough stacked high-bandwidth memory, the HBM that top-end AI chips need. The 910C ran on an older grade, HBM2E, that Huawei had been able to mass procure. The new 950PR uses lower-bandwidth single stack memory instead of stacked HBM, so the 950PR does not depend on HBM supply. Each chip moves data more slowly, and Huawei can make many more of them. Also, we expect 950DT to ramp up quite quickly alongside 960DT coming earlier in 1Q27 instead of original 4Q27 timeline due to Huawei's own HBM variant that is able to ramp up production quicker. 3. So Nvidia's drop is not Nvidia failing to supply China. In our view, it has three causes: US export rules that bar its best parts, Beijing promoting domestic chips and requiring approval to import Nvidia's compliant ones, and a 950 series that beats the 910 series it replaces and is production-wise attractive for model serving. For Nvidia, where Chinese compute spend goes matters more than the China unit count, and much of it still goes to Nvidia hardware, in China or in Malaysia.

Trading Post November 5, 2026

MercadoLibre’s logistics expansion cut margins by 60 basis points, while Mercado Pago grew credit 62% with 7% NPLs.

Intel: The CPU Comeback, Discounted

Intel's Q3 2026 results were hit by memory costs and an Israel shutdown, but enterprise AI inference could lift CPU demand as server…

Problems at Twist Bioscience (TWST)

Twist Bioscience’s $200 rebound followed an Anthropic-linked short squeeze, yet its DNA and protein services face broad competition while…

VSH Earnings Review

Vishay reported record Q3 revenue of $979 million, up 24%, and guided Q4 to about $990 million as rising backlog and margins contrast with…

Thursday Tailwinds #13: Nov 5, 2026

Florida marinas are 92% full as boat counts rise and permits constrain new supply, creating pricing power for existing operators such as…

eToro: Expanding Beyond Crypto

EToro trades at 2.89x EV/FCF with $304 million in trailing free cash flow, expanding from crypto into equities, commodities, currencies and…

SILC · long

Silicom LTD

Silicom (SILC) is pitched as inexpensive on its existing hardware business, with upside framed as a cheap call option.

Publications we follow

Every publication on this site, hand-picked.

129 publications

Newsletters 109

X accounts we follow

Every account on this site, hand-picked.

91 accounts

Investor letters we follow

359 firms

Sectors

Ordered by how often our editor reaches for them.