---
type: "Topics"
locale: "en"
url: "https://longbridge.com/en/topics/40316981.md"
description: "Over the past two years, the market's imagination of AI computing power has almost entirely been placed on the cloud—H100s in queues, power shortages, and increasingly staggering CapEx numbers. Edge AI has always been the other, overlooked line. The reason isn't complicated: the models couldn't run. Stuffing a flagship large model with 70B parameters into a phone couldn't pass the memory and power consumption tests. In April 2026, Google open-sourced Gemma 4. With four tiers from E2B to 31B, the lowest requiring only 3.2GB of memory to run. This is a watershed moment—not because Gemma itself is exceptionally powerful, but because it shows that model lightweighting has caught up with hardware. Phones, PCs, wearables, and in-vehicle systems, terminals once considered "lacking in computing power," are starting to gain the conditions to truly run flagship-level models this year. $Qualcomm(QCOM.US) surged after announcing its latest earnings report! Could Qualcomm become the leader in mobile ASICs? After all, Qualcomm's CEO just announced: "We will also enter the custom chip (ASIC) field, starting with collaboration with a leading hyperscale cloud service provider, with the first shipments expected by year-end." After edge AI is implemented in the future, the beneficiaries may not be the same companies as in cloud AI. The cloud competes on HBM, advanced process nodes, and rack power density; the edge competes on low-power NPUs, on-chip memory, low-latency communication, and sensors. Simply put, those analog, RF, MCU, and memory interface companies that haven't risen much in the past two years might get their turn in this round. Currently, the stock list (US stocks) covers several categories:1. Edge AI Chips: $Qualcomm(QCOM.US) $Apple(AAPL.US)2. IP & Design Services: $Arm(ARM.US)  $CEVA(CEVA.US)  $Synaptics(SYNA.US)3. Memory & Interface: $Micron Tech(MU.US) $Rambus(RMBS.US)4. Communication & RF: $Skyworks Solutions(SWKS.US) $QORVO(QRVO.US)5. Foundry: $United Microelectronics(UMC.US)  $Intel(INTC.US)5. Device Terminals: $Dell Tech(DELL.US)  $HP(HPQ.US) etc. The rotation within the semiconductor sector this year is evident, and this kind of "waiting for the wind" alpha opportunity might appear in these segments."
datetime: "2026-04-30T07:59:24.000Z"
locales:
  - [en](https://longbridge.com/en/topics/40316981.md)
  - [zh-CN](https://longbridge.com/zh-CN/topics/40316981.md)
  - [zh-HK](https://longbridge.com/zh-HK/topics/40316981.md)
author: "[官方股单 ](https://longbridge.com/en/profiles/8795959.md)"
---

# Over the past two years, the market's imagination …


### Related Stocks

- [QCOM.US](https://longbridge.com/en/quote/QCOM.US.md)
- [SWKS.US](https://longbridge.com/en/quote/SWKS.US.md)
- [UMC.US](https://longbridge.com/en/quote/UMC.US.md)
- [MU.US](https://longbridge.com/en/quote/MU.US.md)

## Comments (4)

- **宇宙大爆炸 · 2026-04-30T15:49:56.000Z · 👍 2**: I've already ordered the Pixel 10 to experience on-device AI. From a cost-efficiency perspective, simple tasks use on-device models, while complex tasks are handled in the cloud.
- **uRock · 2026-04-30T15:22:03.000Z · 👍 1**: I'm optimistic about low-power ARM. x86 is not only weak in mobile phones, but it's even weaker in cloud servers, and it's power-hungry.
  - **Vito_zhang** (2026-04-30T15:34:01.000Z): Positive for liquid cooling
- **奇迹的交易员cola · 2026-04-30T15:18:55.000Z · 👍 2**: This watchlist is very valuable💫