Conviction: Medium-High (upgraded from Medium)
Status: Researching — watching for entry
Editorial Note
ARM just confirmed the AGI CPU thesis is real — and now the stock has given you a 7% haircut from its RSI 84 peak to reload. The stock hit $166 on March 25 after the AGI CPU launch, RSI screamed to 84, and then spent this week cooling back to $148. That's not thesis invalidation — that's the FOMO washout you were waiting for. The custom silicon business model shift is intact. The entry zone from the last deep dive ($130-$145) is now in play. The question is whether to start nibbling here at $148, or wait for the full pullback.
The Story Right Now
Six days ago ARM launched its first-ever production silicon — the AGI CPU, a 136-core Neoverse V3 monster for AI inference, with Meta as anchor customer and OpenAI, Cloudflare, SAP, and SK Telecom signed on. The stock ripped from $135 to $166 in a single session (+19%), RSI hit 84, and then the normal hangover set in. Three sessions of distribution brought it back to $144 on March 27. Today it bounced hard back to $148.73 — a +8.6% recovery from yesterday's low of $140.
What's changed since the March 25 deep dive: price has pulled back 5% from the AGI CPU spike close, RSI has come in from 84 to 67. What hasn't changed: the strategic thesis. ARM is transitioning from pure IP licensor (royalties on other people's chips) to silicon vendor (selling its own chips). That business model evolution — from ¢¢¢ per chip to $$$ per chip — is the re-rating event. It's early innings.
The macro backdrop is key context. Per the March 26 market brief, ARM (+40% 3M) is one of only two names in the custom silicon cluster working in this war-driven market (MRVL +14% 3M is the other). Semis broadly are weak — SMH, SOXX both down 8-10% — but ARM is diverging on a genuine business model catalyst rather than sector momentum. That divergence says something important.
The Taiwan/TSMC risk is real and worth pricing in. ARM's AGI CPU is manufactured by TSMC. Any Hormuz-adjacent escalation into the Taiwan Strait would disrupt that supply chain. But this risk exists for the entire semiconductor industry — NVDA, AMD, QCOM all depend on TSMC. It's a reason to size conservatively, not a thesis-killer.
The Malaysia corruption probe (MACC investigating a $279M government deal) remains a noise event — ARM has not been named as a subject, and the stock has essentially ignored it through multiple sessions.
SoftBank's 75% ownership remains the structural overhang. Masa Son will eventually monetize this position. The question is timing — with ARM at a 5-year high and silicon revenue just beginning, SoftBank's interest is in maximizing the exit, not rushing it.
Quick Snapshot
| Signal |
Reading |
| Overall |
🟢 Bullish — historic catalyst intact, RSI cooling from extreme to healthy, entry zone approaching |
| Moat |
Wide — 99% smartphone CPU share, Armv9 royalty supercycle, developer ecosystem lock-in |
| Key insight |
RSI 84 → 67 pullback is healthy digestion; $130-$145 entry zone is now realistic on further consolidation |
Action Matrix
| Action |
Level |
Why |
| Current |
🔍 Starter position possible |
RSI 67 is healthy; down from 84; AGI CPU thesis intact |
| Entry Zone |
$138 - $148 |
Pullback zone; SMA20 ($130) is the deep support |
| Stop-Loss |
$110 (-26%) |
Below Feb lows; thesis-breaking level |
| Target |
$210 (+41%) |
Re-rate target as silicon revenue becomes material; prior 52wk high re-test then beyond |
Price Data
| Stock |
Price |
1D |
7D |
30D |
3M |
52wkHi |
RSI |
Trend |
Action |
| ARM |
$148.73 |
+8.6% |
-5.3% |
+19.6% |
~+38% |
-18.8% |
67.3 |
Strong-up |
🔍 Starter position / wait for $138-145 |
Key Technical Levels
| Level |
Price |
Notes |
| Support 1 |
$144-145 |
This week's consolidation low |
| Support 2 |
$130 |
SMA20 — prior resistance, now key support |
| Support 3 |
$111 |
Feb lows — major support / thesis invalidation |
| Resistance 1 |
$157-166 |
AGI CPU launch week highs |
| Resistance 2 |
$183 |
52-week high |
| BB Upper |
$156 |
Current upper band |
| BB Middle |
$130 |
SMA20 equivalent |
RSI Journey (March)
- Mar 2-13: Ranged $114-124, RSI not yet computed (below 14 day window)
- Mar 19-20: AGI CPU news beginning; $130-132, RSI 57-59
- Mar 25: AGI CPU launch day; $157 close, RSI 78.6
- Mar 26: Follow-through to $154.8, RSI 83.9 (peak overbought)
- Mar 27: First pullback $144.1, RSI 69.7
- Mar 30: Further cooling $137, RSI 61.5
- Mar 31: Bounce to $148.73, RSI 67.3
Company Overview
One-Liner
ARM designs the CPU architecture used in 99% of smartphones and is rapidly expanding into data center AI workloads — now making its own silicon (AGI CPU) for the first time in its 35-year history.
Business Model
| Question |
Answer |
| What they sell |
CPU architecture IP (instruction sets, core designs) + production silicon (AGI CPU, new) |
| Who pays |
Chip designers (Qualcomm, Apple, NVIDIA, MediaTek, Samsung) + now hyperscalers directly |
| Revenue model |
Upfront license fees + per-chip royalties + silicon sales (new, material in H2 2026+) |
| How sticky |
Extremely — 15M+ developers, decades of compiled binaries; switching costs are astronomical |
Key Segments
| Segment |
Revenue % |
Growth |
Notes |
| Royalty |
~59% ($737M Q3 FY26) |
+27% YoY |
Per-chip fees; Armv9 doubles rates |
| License & Other |
~41% ($505M Q3 FY26) |
+25% YoY |
Upfront design licenses, CSS adoption |
| Silicon (AGI CPU) |
Emerging |
N/A |
First customer deliveries H2 2026; revenue recognition TBD |
Geographic Mix
| Region |
Revenue % |
Notes |
| Asia-Pacific |
~75% |
Qualcomm, MediaTek, Samsung, TSMC customers |
| North America |
~15% |
Apple, NVIDIA, Qualcomm (US), hyperscalers |
| Europe |
~10% |
Infineon, NXP, STMicro |
Competitive Analysis
Industry Position
| Question |
Answer |
| Market share |
99% smartphones, ~15% data center (growing), ~60% automotive |
| Market size (TAM) |
$250B+ semiconductor IP + now addressable silicon TAM |
| Growth rate |
~25% CAGR (royalty acceleration from Armv9) |
| Key competitors |
x86 (Intel, AMD) in data center; RISC-V (open-source) long-term threat |
| Position |
Dominant mobile, growing data center, new silicon vendor |
Competitive Moat
| Moat Type |
Present? |
Evidence |
| Network effects |
🟢 |
15M+ developers, billions of compiled binaries |
| Switching costs |
🟢 |
Recompiling entire software stacks takes years |
| Cost advantages |
🟢 |
ARM inherently more power-efficient than x86 — critical for AI inference cost |
| Intangible assets |
🟢 |
6,000+ patents, 35 years of architecture refinement, Armv9 royalty premium |
| Efficient scale |
🟢 |
Natural monopoly in mobile CPU architecture |
Moat Assessment
Moat Width: Wide
Moat Trend: Widening — silicon vendor expansion adds margin capture on top of royalty stream
Summary:
ARM's moat is one of the widest in tech. The software ecosystem lock-in (15M developers, decades of compiled code) makes switching nearly impossible. The AGI CPU adds a new dimension: ARM now captures silicon margin directly from hyperscalers instead of waiting for downstream royalties. The RISC-V threat is real but 10+ years from matching ARM's ecosystem depth.
Management Assessment
Leadership
| Role |
Name |
Since |
Background |
Notes |
| CEO |
Rene Haas |
Feb 2022 |
7 years at NVIDIA (IP licensing), then Arm IP Products Group |
Architect of CSS strategy; driving silicon pivot |
| CFO |
Jason Child |
2023 |
Former CFO at Splunk, Opendoor |
Enterprise SaaS financial discipline |
| Chief Architect |
Richard Grisenthwaite |
20+ years |
Led Armv6 through Armv9 architecture evolution |
Technical brain behind the moat |
Founder Involvement
| Question |
Answer |
| Founder-led? |
No — SoftBank subsidiary (~75% ownership) |
| SoftBank ownership |
~75% |
| Skin in the game |
SoftBank's largest bet; Masa Son considers ARM his legacy asset |
Capital Allocation
| Metric |
Track Record |
| M&A discipline |
Good — focused on small IP acquisitions |
| R&D investment |
High — critical for architecture evolution and AGI CPU development |
| Debt management |
Conservative — minimal debt |
Red Flags
- SoftBank's ~75% control limits minority shareholder influence; lockup expiry risk
- Malaysia $279M deal under corruption probe (ARM not named as wrongdoer; stock shrugged it off)
- AGI CPU competes with largest licensees (Qualcomm, Broadcom) — relationship strain risk
Financials
Key Metrics
| Metric |
Q3 FY26 |
Q2 FY26 |
Q3 FY25 |
Trend |
| Revenue |
$1.24B |
$1.14B |
$983M |
📈 |
| Revenue Growth YoY |
+26% |
+23% |
+5% |
📈 Accelerating |
| Royalty Revenue |
$737M |
$620M |
$580M |
📈 Record |
| License Revenue |
$505M |
$515M |
$403M |
📈 |
Quality Checks
| Check |
Status |
Notes |
| FCF positive? |
🟢 |
Strong FCF generation |
| Profitable? |
🟢 |
Net income positive and growing |
| Debt manageable? |
🟢 |
Minimal debt |
| Cash runway |
Indefinite |
Self-funding; no dilution risk |
Leading Indicators
| Indicator |
Value |
Signal |
| Revenue Acceleration |
+26% YoY |
🟢 Accelerating (from +5% YoY two years ago) |
| Royalty Record |
$737M Q3 FY26 |
🟢 Armv9 penetration driving structural increase |
| Earnings Beat Rate |
75% |
🟢 Consistent execution |
| Next Catalyst |
~May 2026 Q4 FY26 |
AGI CPU pipeline commentary |
Valuation
Current Multiples
| Metric |
Current (at $148) |
Industry Avg |
vs Industry |
| P/E (trailing) |
~155x |
30x |
Extreme premium |
| P/E (forward) |
~50-55x |
25x |
Large premium |
| P/S |
~25x |
5-8x |
Extreme premium |
| Market Cap |
$158B |
— |
— |
Valuation Assessment
At $148 vs. the $161 peak, ARM's valuation has moderated slightly but remains extreme. The forward P/E of ~50x assumes 25%+ revenue growth continuing. The bull case is that silicon revenue — completely absent from current estimates — could drive significant upside to consensus. A $1B+ annual run rate in AGI CPU revenue by FY28 would represent ~20% of total revenue and justify a platform multiple re-rate. The bear case is that any quarter of growth deceleration at these multiples would be punished severely.
Bull Case
Why This Could Work
AGI CPU opens silicon-level margins — business model transformation
- Evidence: Meta (anchor), OpenAI, Cloudflare, SAP, SK Telecom as first customers
- Implication: ARM captures chip margin ($$$) not just royalty margin (¢¢¢) — structural P&L upgrade
Armv9 royalty supercycle is structural
- Evidence: Royalty revenue +27% YoY, Armv9 doubles per-chip royalty rates
- Implication: Revenue growth accelerates as Armv9 penetration deepens (currently ~30% of installed base)
Data center share gains from hyperscaler custom silicon
- Evidence: AWS Graviton, Google Axion, Microsoft Cobalt, Ampere — all ARM-based
- Implication: x86 data center erosion is a decade-long, structural tailwind
AI inference favors ARM's power efficiency
- Evidence: AGI CPU claims 2x performance per rack vs x86, with massive capex savings per GW
- Implication: As AI inference scales to billions of daily queries, power efficiency = ARM's killer advantage
Upside Scenario
| If This Happens |
Stock Could |
| AGI CPU hits $1B+ ARR by FY28 + Armv9 penetration |
$220+ (35%+ upside) |
| Data center royalty hits 25%+ of royalty revenue |
$200 re-rate to platform multiple |
Bear Case
What Could Go Wrong
Valuation compression on ANY growth miss
- How it plays out: At 50x forward P/E, even one quarter of 20% growth (vs. 25% expected) = -20% stock
- Probability: Medium — these multiples demand flawless execution
Taiwan/TSMC supply chain disruption
- How it plays out: AGI CPU is TSMC-manufactured; geopolitical escalation around Taiwan disrupts supply
- Probability: Low (6-12 months), but non-zero given geopolitical environment
AGI CPU licensee conflict
- How it plays out: ARM competing with Qualcomm, Broadcom via AGI CPU creates tension; licensees develop RISC-V alternatives
- Probability: Medium — this is the key strategic risk to monitor closely
SoftBank overhang
- How it plays out: SoftBank monetizes 75% stake gradually, creating persistent stock supply
- Probability: Medium — SoftBank has been monetizing winners across portfolio
Thesis Killers
- Revenue growth decelerates below 15% for two consecutive quarters
- Major licensee (Qualcomm/Apple) announces credible RISC-V migration timeline
- AGI CPU fails to gain traction beyond initial 5 partners in 12 months
Market-Moving News
Recent News
| Date |
Headline |
Source |
Impact |
Relevance |
| 2026-03-25 |
ARM launches AGI CPU — first in-house silicon in 35-year history; Meta, OpenAI first customers |
TechCrunch, CNBC |
🟢 |
Historic catalyst; business model evolution. +16% day. |
| 2026-03-25 |
Bank of America price target increase; new strategic partnerships |
Yahoo Finance |
🟢 |
Street validating the move |
| 2026-03-24 |
Stock rallied on AI server CPU demand + Malaysia probe context |
Motley Fool |
🟢 |
Pre-launch momentum building |
| 2026-03-23 |
+14.3% on AI server CPU optimism |
StockTitan |
🟢 |
Market was anticipating the announcement |
| 2026-03-22 |
Malaysia MACC corruption probe — $279M government deal |
Nasdaq |
🟡 |
ARM not named; stock shrugged it off |
| 2026-03-21 |
Cramer: "ARM always tight with NVIDIA, I think it gets tighter" |
MarketBeat |
🟡 |
Analyst sentiment bullish |
| 2026-02-04 |
Q3 FY26 earnings: $1.24B revenue (+26% YoY), record royalties |
ARM Newsroom |
🟢 |
Strong fundamental base before AGI catalyst |
News Patterns
ARM moves on (1) earnings beats driven by royalty growth acceleration, (2) hyperscaler adoption/partnership announcements, (3) architecture and product launches. The AGI CPU announcement is the biggest single-day catalyst since the IPO. Stock tends to overshoot in both directions around these events.
| Symbol |
Name |
Relationship |
30D |
Status |
Notes |
| SMH |
VanEck Semiconductor ETF |
Top holding |
-8% est. |
🔴 Weak |
ARM dramatically outperforming sector |
| SOXX |
iShares Semiconductor |
Contains ARM |
-10% est. |
🔴 Weak |
Broad semi weakness |
| QQQ |
Nasdaq 100 |
Top-50 holding |
-5% |
🔴 Weak |
ARM diverging from index |
| MRVL |
Marvell Technology |
Custom silicon peer |
+14% 3M |
🟢 |
Same custom silicon theme |
| NVDA |
NVIDIA |
Architecture partner |
~-10% 30D |
🔴 |
ARM ecosystem plays differently |
Sector Context
ARM is dramatically outperforming the semiconductor sector. SMH is down ~8% 30D while ARM is up +20%. This divergence is entirely catalyst-driven. The custom silicon theme (ARM, MRVL) is the only working corner of tech hardware in this war market.
Cross-References (Where This Appears in Our System)
| Location |
File |
Context |
| Watchlists |
focus, ai-scan, ai-infra, semis, tech-insider-buys, insider-universe |
Core tracking across 6 lists |
| Market Brief |
research-journal/2026-03-26-full-scan-market-brief.md |
Highlighted as custom silicon winner (+40% 3M) |
| Scan |
March 26 brief — "What's Actually Working" |
Part of custom silicon cluster alongside MRVL |
| Ticker |
Why Related |
Priority |
Notes |
| QCOM |
Largest ARM licensee; AGI CPU relationship to watch |
🟢 |
Watch for partnership announcements vs. conflict signals |
| AVGO |
Custom ARM chips for hyperscalers |
🟢 |
Could benefit from ARM ecosystem expansion |
| MRVL |
Custom ARM silicon for data center |
🟢 |
Closest peer in custom silicon theme |
| TSM |
TSMC manufactures AGI CPU |
🟢 |
Supply chain dependency; geopolitical risk proxy |
Catalysts & Timing
Upcoming Catalysts
| Date |
Event |
Impact |
Watch For |
| ~May 2026 |
Q4 FY26 Earnings |
🟢 |
First quarter with AGI CPU pipeline commentary; any revenue recognition |
| H2 2026 |
AGI CPU broad availability |
🟢 |
Revenue recognition timing; Meta deployment scale |
| Ongoing |
New AGI CPU customers |
🟢 |
Each new hyperscaler partnership = re-rating event |
| Ongoing |
Malaysia probe resolution |
🟡 |
Any naming of ARM as subject would be negative |
| Ongoing |
SoftBank share sales |
🔴 |
Monitor Form 4 / 13F filings for any stake reduction |
Entry Strategy
Position Sizing
| Conviction |
Allocation |
| Medium-High |
3-5% |
My conviction: Medium-High — historic catalyst is real, business model evolution is intact, pullback from RSI 84 creates entry opportunity
Target allocation: 4%
Entry Approach
| Strategy |
Details |
| Entry zone |
$138 - $148 (SMA20 to current consolidation) |
| Starter position |
40% of target at $148 on today's bounce confirmation |
| Add on |
Any pullback to $138-145 adds another 40%; RSI back at 55-60 |
| Full position at |
$130 (SMA20 test) — highest conviction add |
Technical Levels
| Level |
Price |
Notes |
| Support 1 |
$144-145 |
This week's pullback low |
| Support 2 |
$130 |
SMA20 — key mean-reversion target |
| Support 3 |
$111 |
Feb lows — thesis-breaking level |
| Resistance 1 |
$157-166 |
AGI CPU launch week range |
| Resistance 2 |
$183 |
52-week high |
| Target |
$210+ |
Re-rate on silicon revenue |
Short Interest
| Metric |
Value |
Signal |
| Short shares |
~14.7M (last reported) |
Moderate |
| Days to cover |
~4 |
Elevated — potential squeeze fuel |
| 10D avg volume |
10.8M |
Very elevated vs. 6.8M avg — institutions active |
Sources
Research Log
| Date |
Update |
| 2026-03-25 |
Created deep dive — AGI CPU launch catalyst, first silicon in 35 years, RSI 79 overbought |
| 2026-03-31 |
Refreshed — RSI cooling from 84 to 67, price from $166 to $148, entry zone now in play; conviction upgraded to Medium-High as thesis confirmed |
The Gold
Key Discoveries
| Discovery |
Implication |
| RSI 84 → 67 in 4 sessions |
Controlled consolidation, not breakdown; the FOMO buyers are flushed |
| AGI CPU: Meta co-developing, not just buying |
Deep integration = ARM's AI success is tied to Meta's AI roadmap |
| Today's volume at 6.9M (avg 6.8M) |
Institutional-level buying on the $140 bounce — smart money stepping in |
| TSMC dependency for AGI CPU manufacturing |
Geopolitical risk is priced in broadly; ARM shares this with all semis |
| 10D avg volume 10.8M vs. 6.8M avg |
Institutions are very active in this name post-AGI announcement |
Open Questions
- When will AGI CPU revenue hit the P&L? H2 2026 availability → revenue recognition cadence?
- How will Qualcomm/Broadcom respond to ARM competing via AGI CPU?
- What's the margin profile of silicon sales vs. royalty/license (silicon could be lower margin initially)?
- Will SoftBank use this rally to reduce its 75% stake? Monitor 13F/Form 4 filings.
- How many new AGI CPU customers get announced at Q4 earnings in May?