---
title: "ALAB is not the \"AI chip stock\" you think it is"
type: "Topics"
locale: "en"
url: "https://longbridge.com/en/topics/40091868.md"
description: "$Astera Labs(ALAB.US) A stock that was still hovering around $100 at the end of March, surged to $190+ by the end of April, nearly doubling in a month. Your first reaction to this kind of movement might be, "Which AI theme is being hyped again?" But ALAB this time is not purely driven by sentiment—Old Xu will break it down for you. What exactly does ALAB do? In one simple, blunt sentence: It's the "signal transmission bridge" between GPUs and GPUs, CPUs and GPUs inside AI data centers..."
datetime: "2026-04-22T08:52:26.000Z"
locales:
  - [en](https://longbridge.com/en/topics/40091868.md)
  - [zh-CN](https://longbridge.com/zh-CN/topics/40091868.md)
  - [zh-HK](https://longbridge.com/zh-HK/topics/40091868.md)
author: "[沪上老徐](https://longbridge.com/en/profiles/26419342.md)"
---

# ALAB is not the "AI chip stock" you think it is

$Astera Labs(ALAB.US) was still hovering around $100 at the end of March, but by the end of April, it had surged to $190+, **nearly doubling in a month**. Your first reaction to this kind of price action might be, "Which AI hype train is this?" But ALAB this time isn't just pure sentiment—let me break it down for you.

* * *

**What does ALAB actually do?**

In simple terms: **It's the "signal transmission bridge" between GPUs and between CPUs and GPUs inside AI data centers**.

You buy a bunch of NVDA's H100 / B100s and stack them in a server room. How do they talk to each other? Copper wires aren't enough, and ordinary fiber optics won't cut it either—you need specialized chips like PCIe retimers (signal regeneration) and P-series switches (high-speed switches) to connect hundreds or thousands of GPUs into a "giant brain." **ALAB does exactly that**.

Think of it as a highway: NVDA builds the cars, ALAB builds the roads. Your car can be super fast, but it won't get anywhere if the roads are bad.

* * *

**Why the sudden surge this month?**

**Core catalyst**: AMZN and Anthropic signed a massive 10-year, $100 billion deal, with Anthropic committing to spend that much on AMZN's in-house Trainium chips over the next decade, with a total compute scale of 5GW.

Hearing "AMZN's in-house chips," your first thought might be, "So they're not using NVDA anymore?"—But analysts did the math: **To scale up AMZN's Trainium clusters, the interconnect layer still needs ALAB's technology**. It's like AMZN building its own new cars, but the roads still need to be built by... well, ALAB.

Once this logic emerged, ALAB was directly upgraded from "NVDA's ecosystem supplier" to "the common supplier for all AI in-house chip players," leading to a market repricing.

* * *

**Is the valuation expensive?**

To be honest—yes.

-   Market cap $29.95B
-   P/E 144 (based on current earnings)
-   Forward P/E still around 50x (based on next year's EPS of $3.53)
-   YoY revenue +91.8%

The valuation logic only holds up on one thing: **continued high-speed revenue growth**. If revenue growth drops below 50% in the next two quarters, this valuation won't hold.

**But there's no reason for it to drop in the short term**: AMZN, Microsoft, Meta, and Google are all building AI data centers, and each one requires a large number of ALAB's interconnect chips, making orders quantifiable and visible.

* * *

**Which sectors benefit?**

-   **AI Infrastructure** (most direct) — All hyperscale AI data centers are its market.
-   **In-house ASIC Chip Trend** — Trainium, TPU, MTIA are all customers.
-   **PCIe Upgrade Cycle** — Demand for ALAB's retimers doubles during the PCIe 6.0 transition period.

* * *

ALAB isn't a question of "whether to buy," but "at what price to buy."

**I don't recommend chasing at the current price of $190**—it rose too fast, IV is high, and a 15-20% pullback is perfectly normal. Wait for it to return to the $160-$170 range before reconsidering.

**The long-term thesis is solid**: AI infrastructure is one of the most certain trends over a 3-5 year horizon, and ALAB is nearly monopolistic in the interconnect niche, making it more like the "shovel" in "selling shovels" than even NVDA.

**Risk factors**: (1) AMZN's in-house chip progress lags expectations; (2) Competitors emerge for PCIe 6.0; (3) A valuation correction leads to a one-time 30% drop.

### Related Stocks

- [ALAB.US](https://longbridge.com/en/quote/ALAB.US.md)
- [META.US](https://longbridge.com/en/quote/META.US.md)
- [MSFT.US](https://longbridge.com/en/quote/MSFT.US.md)
- [AMZN.US](https://longbridge.com/en/quote/AMZN.US.md)
- [GOOG.US](https://longbridge.com/en/quote/GOOG.US.md)
- [MSFL.US](https://longbridge.com/en/quote/MSFL.US.md)
- [MSFO.US](https://longbridge.com/en/quote/MSFO.US.md)
- [MSFD.US](https://longbridge.com/en/quote/MSFD.US.md)
- [MSFU.US](https://longbridge.com/en/quote/MSFU.US.md)
- [MSFX.US](https://longbridge.com/en/quote/MSFX.US.md)
- [MSFY.US](https://longbridge.com/en/quote/MSFY.US.md)
- [AMZU.US](https://longbridge.com/en/quote/AMZU.US.md)
- [GGLS.US](https://longbridge.com/en/quote/GGLS.US.md)
- [GGLL.US](https://longbridge.com/en/quote/GGLL.US.md)
- [GOOGL.US](https://longbridge.com/en/quote/GOOGL.US.md)

## Comments (1)

- **xxy · 2026-04-22T10:44:38.000Z**: Is a long-term 10x increase possible?