---
title: "ARM: 'Chip Tax' Stalls; Can an AGI CPU Justify the Valuation?"
type: "Topics"
locale: "en"
url: "https://longbridge.com/en/dolphin/post/44804833.md"
description: "Delegation can't hide the 'chip tax' slowdown"
datetime: "2026-11-05T08:29:30.000Z"
locales:
  - [en](https://longbridge.com/en/dolphin/post/44804833.md)
  - [zh-CN](https://longbridge.com/zh-CN/dolphin/post/44804833.md)
  - [zh-HK](https://longbridge.com/zh-HK/dolphin/post/44804833.md)
author: "[Dolphin Research](https://longbridge.com/en/dolphin.md)"
generator: "portal-rs"
---

# ARM: 'Chip Tax' Stalls; Can an AGI CPU Justify the Valuation?

ARM (ARM.O) released its FY27 Q2 results (through Sep 2026) in the U.S. after-hours session on the morning of Nov 5 Beijing time. Key takeaways below.

**1) Headline: revenue of $1.44bn (+27% YoY)**, above the Street ($1.38bn). **Upside was driven by License, while Royalty growth decelerated markedly**. GAAP GPM was 97.1%, remaining elevated.

**2) Biz. mix:** License rose to 54% of revenue, while Royalty fell to 46%. Mix shifted toward upfront License.

**a) License revenue was $770mn this quarter (+50% YoY)**, well ahead of consensus ($670mn). **The CFO noted License was about $100mn above internal plan, driven by stronger appetite for long-term agreements**: three new ATA (Arm Total Access, >60 ATAs now active, including a first-time ATA with a major customer) and four CSS licenses were signed (covering data center and smartphone chips). **SoftBank-related tech licensing and design services contributed $209mn**, up 17% YoY, and the company expects a similar run-rate per quarter in the back half.

**b)**$Arm(ARM.US) **Royalty revenue was $670mn (+8% YoY), below consensus ($707mn)** and under the prior company guide (+13%). **Growth slowed sharply as ongoing memory price hikes continued to depress low-end smartphone shipments**. Edge AI (phones, PCs, IoT) royalties were roughly flat YoY, and per-chiprate uplift from Armv9 and CSS only barely offset unit declines.

Helped by accelerating ARM-based server chip ramps at major cloud providers, **data center Royalty doubled YoY** this quarter. Physical AI (auto, robotics) Royalty rose 50% YoY. Partner shipments of ARM-based chips topped 10bn units in a single quarter for the first time, up 1.5bn YoY.

**3) Core metrics: (i) ACV reached $1.79bn**, **+12% YoY and +3.3% QoQ**, roughly in line with the Street ($1.80bn). (ii) RPO disclosure was discontinued starting last quarter.

**4) Opex:** R&D expense was $860mn (+24% YoY). SG&A was $340mn (+36% YoY). **R&D stepped up on next-gen architectures, CSS, and the AGI CPU product line**.

Net income was $130mn (-45% YoY), pressured by income tax expense ($78mn) and equity investment losses ($20mn). **Core OP was $200mn (+23% YoY), driven by topline growth and lower opex ratio**.

**5) Next-quarter guide: ARM expects FY27 Q3 revenue of $1.49–1.59bn**, with a midpoint of $1.54bn **(+24% YoY)**, in line with the Street ($1.53bn). Non-GAAP EPS is guided to $0.52–0.60, midpoint $0.56, **also in line with the Street ($0.55)**.

**Within that, License is guided to grow slightly above 50% YoY next quarter, while Royalty is guided to grow just over 10% QoQ and low single digits YoY**. Mix remains License-heavy near term.

**Dolphin Research view: License filled the gap, but the 'chip tax' is stalling**

Revenue and EPS both beat the top end of guidance, **but the upside was almost entirely from License, while Royalty, the 'chip tax', kept slowing**. Momentum remains uneven by component.

**License quality warrants a discount: management acknowledged the quarter benefited from timing of several high-value deals and backlog, implying more one-offs**. ACV, a gauge of License sustainability, was tepid, with YoY growth slightly softer vs. last quarter and broadly in line with expectations. A big License beat without ACV acceleration suggests timing of recognition, not a step-up in new deal scale.

Opex ratio fell and core OPM improved QoQ, largely on scale leverage rather than strict cost control. R&D is still growing fast.

Last quarter we argued ARM's 'rich valuation' requires 'superb execution'. The stock rallied above $330 on Meta validation, the CEO's confidence in AGI CPU, and an Agentic AI CPU-demand narrative, then gave back all gains. Volatility remained high.

**Management again reiterated the AGI CPU target of up to $2bn across FY27–28, boosting confidence, but the Royalty slowdown is a new negative. Shares drifted lower post-print (after-hours -3%)**.

The market is now focused less on the print and more on guidance, ACV, and the AGI CPU trajectory. These are the three key signposts.

**1) Guidance:** management guides $1.49–1.59bn for next quarter. **By component, License should be slightly above +50% YoY (~$800mn), while Royalty grows low single digits YoY (Dolphin est. ~$750mn)**.

In other words, next quarter's 'meet' has a similar setup as this quarter—**License backfills Royalty weakness**. Management also clarified AGI CPU shipments only begin in FQ4 (any FQ3 shipments would be minimal).

**More crucial is the full-year Royalty outlook.** Last quarter, the FY27 Royalty growth guide was trimmed to ~18%. Based on this quarter's actuals and next quarter's guide, **H1 Royalty is +15% YoY, FQ3 around +2%, and achieving +18% for the full year would require ~+40% in FQ4—unlikely. Management did not reiterate the full-year figure**. Dolphin estimates FY27 Royalty growth likely lands near ~10%, effectively a stealth downgrade.

**Management's view:** the memory price upcycle started to weigh on smartphones earlier this year. **Smartphone units are expected to be roughly flat in FY28**. With Armv9 penetration already above 50% and higher CSS rates, **phone Royalty could return to double-digit growth**.

**2) ACV:** a forward indicator for near-term revenue. ACV was $1.79bn this quarter, up 3.3% QoQ. Given the revenue mix, **Dolphin estimates roughly $430mn of the quarter's revenue came from prior contracts, while 'new contracts recognized in-quarter + Royalty' totaled about $1.01bn (vs. ~$870mn last quarter), with new contracts recognized in-quarter at roughly $340mn**.

Management indicated License should continue to rise QoQ into FQ3 and FQ4. ATA/CSS typically convert into design wins and Royalty after 12–24 months. **But ACV did not accelerate alongside License, so Dolphin remains cautious on the durability of the License surge**.

**3) AGI CPU:** beyond the core IP business, ARM plans to directly enter data center CPUs, with the first product introduced in Mar. This is a new revenue stream on top of existing licensing.

**Management raised the AGI CPU outlook:** ARM has signed purchase orders with multiple customers through end-2027 and secured supply agreements with ecosystem partners (wafers, substrates, test, memory, etc.). **Confidence in delivering 'up to $2bn' of AGI CPU revenue across FY27–28 has increased** (the CEO already cited $2bn at the Sep investor event).

**Initial commercial shipments remain planned for FY27 Q4 (Jan–Mar 2027), with Meta as the lead customer**, and others beyond Meta. Oracle OCI plans AGI CPU instances in 2027, and Supermicro and Lenovo will deliver full-rack systems.

**The incremental update is a tighter 'demand → orders → supply' chain**. Management expects first-gen GM at 30%–40%, with cost inflation in memory and other inputs partially passed through. **Reaching 50%+ industry-level margin likely awaits Gen 2 (with 18–24 month product cycles)**.

Since **ARM's AGI CPU ships in FY27, Dolphin's FY29 model already embeds the lift to growth and the dilution to GPM from CPU shipments (assuming CPU GM ramps from ~35% toward ~50%)**. Assuming revenue CAGR of +40%, GPM at 86%, and a 10% tax rate, ARM's current market cap (~$204.8bn) implies about 47x FY29 core after-tax OP.

There are multiple ways to value ARM. Some focus on FY28–FY29, while others anchor on the company's FY31 outlook.

**The prior FY31 outlook: total revenue of $25bn+, with $10bn from the legacy IP business and $15bn from AGI CPU, and EPS >$9**. Some institutions have even applied a single blended valuation to the whole company.

**Dolphin views an all-in-one valuation as too crude.** The IP business is a mature, high-GP licensing model, while the CPU business starts from zero and resembles a chip design house more closely. Thus we value the two segments separately.

Assume IP revenue CAGR of 17% over four years (FY27–FY31) with GPM at 97.5%. The new CPU business scales from near-zero (to an estimated $16bn), with GPM rising from 35% to 50%. FY31 opex ratio is assumed at 32% (allocated by revenue).

On FY31, the IP/CPU segments deliver core after-tax profit of $6.8bn/$2.6bn, with 20–40x and 15–30x PE respectively. **This implies $136–272bn for the legacy IP business and $39–78bn for the new CPU business, bringing total value to $175–350bn**. **Discounted back at 11%, that equates to roughly $115.5–231bn today**.

**Overall, this quarter is a 'headline beat, weaker mix' print**. Revenue/EPS beat the top end, and the AGI CPU outlook points to up to $2bn over FY27–28, which looks solid. **But the long-term value driver—Royalty—slowed to 8%, next-quarter guide implies low-single-digit growth, the ~18% full-year framework is unlikely to hold, and the License beat was timing-heavy with no ACV acceleration**.

Dolphin sees limited near-term catalysts until Royalty re-accelerates (currently only single digits). **AGI CPU won't meaningfully show up in P&L until FY28, and lower first-gen margins will weigh on profitability**.

**AGI CPU prospects are constructive, so the CPU segment can earn an upper-range multiple**. But slower Royalty growth dampens long-term IP growth expectations, with near-term growth already down from double digits to single digits. The next key windows are the FQ3 print in Feb (AGI CPU update) and the Arm Everywhere mid-year LT target update. **Before then, the stock likely oscillates within the valuation band and struggles to revisit prior highs**.

\<End>

Dolphin Research on ARM

Sep 28, 2026 First Take '[Muse goes viral: Is this the ChatGPT moment for CPUs?](https://longbridge.com/en/dolphin/post/44095954?channel=OWNN00210)'

Jul 30, 2026 Trans '[Arm (Notes): Core biz is hot; new biz demand surging](https://longbridge.com/en/dolphin/post/43054762?channel=OWNN00210)'

Jul 30, 2026 Earnings Take '[ARM: Big Tech pulls back; sky-high valuation cools first](https://longbridge.com/en/dolphin/post/43069298?channel=OWNN00210)'

May 7, 2026 Trans '[Arm (Notes): AI CPU to generate revenue by end of next fiscal year; LT targets unchanged](https://longportapp.cn/en/topics/40497905)'

May 7, 2026 Earnings Take '[ARM: 'Short-term noise' can't hide the edge; AI narrative underpins LT premium?](https://longportapp.cn/en/topics/40496999)'

Feb 5, 2026 Trans '[Arm (Notes): Limited exposure to smartphones; FY27 growth to stay ~20%](https://longportapp.cn/en/topics/38492483)'

Feb 5, 2026 Earnings Take '[ARM: AI tailwinds persist, but smartphones are the drag today?](https://longportapp.cn/en/topics/38489939)'

Nov 6, 2025 Trans '[ARM (Notes): All incremental compute built on Arm architecture](https://longportapp.cn/en/topics/36069902)'

Nov 6, 2025 Earnings Take '[ARM: Backed by SoftBank, a coveted link in the AI chain?](https://longportapp.cn/en/topics/36069546)'

Risk disclosure and statements for this article: [Dolphin Research disclaimer and general disclosure](https://support.longbridge.global/topics/misc/dolphin-disclaimer)

### Related Stocks

- [SOBKY.US](https://longbridge.com/en/quote/SOBKY.US.md)
- [ARM.US](https://longbridge.com/en/quote/ARM.US.md)
- [META.US](https://longbridge.com/en/quote/META.US.md)
- [00354.HK](https://longbridge.com/en/quote/00354.HK.md)
- [SFTBY.US](https://longbridge.com/en/quote/SFTBY.US.md)
- [METAM.US](https://longbridge.com/en/quote/METAM.US.md)
- [9984.JP](https://longbridge.com/en/quote/9984.JP.md)
- [FBL.US](https://longbridge.com/en/quote/FBL.US.md)
- [METU.US](https://longbridge.com/en/quote/METU.US.md)
- [METD.US](https://longbridge.com/en/quote/METD.US.md)

## Comments (2)

- **deskok · 2026-11-08T23:37:49.000Z**: ok
- **Holmas · 2026-11-05T23:33:43.000Z**: It's over.


---
> **Disclaimer: This article is for reference only and does not constitute any investment advice.**