Arm Holdings plc (NASDAQ: ARM) — Equity Research Briefing
Sector: Semiconductors / AI Hardware (IP Licensing)
Price (as of 2026-06-26): ~＄352 · 52-wk range: ＄100–＄453 · Market cap: ~＄378B
Rating context: Street consensus "Moderate Buy"; average price target ~＄280 (below current price)
Prepared: 2026-06-26 · Educational / informational — not investment advice (see disclaimer)
1. Executive Summary
Arm Holdings is the architectural "tollbooth" of modern computing: it does not manufacture chips but licenses CPU designs and
collects a royalty on virtually every chip built around them. The business is structurally excellent — ~98% adjusted gross margin,
recurring royalties that compound for 10–20 years per design, three consecutive years of 20%+ revenue growth, and a genuine new
growth engine in AI/data-center silicon.
After a parabolic run from ~＄120 (March 2026) to an all-time high near ＄445–455 in mid-June 2026, the stock corrected ~21% to
~＄352 amid a macro-driven, AI-trade unwind (hawkish Fed + a sharp Asian tech sell-off on 6/23). Crucially, ARM fell more than its
semiconductor peers because of its very high beta (~3.8) — the basis for the "relative opportunity" thesis explored in this report.
Bottom line: The structural growth story is real, but valuation is the binding constraint (forward P/E ~200x, P/S ~76x — the most
expensive large-cap chip name, with consensus target below the current price). This is a high-quality, high-beta name best
approached with staged entry and disciplined position sizing, not a one-shot "buy the dip."
2. Company Overview
Business model — two revenue streams:
• Licensing fees — upfront payments when a customer takes an architecture/design license.
• Per-chip royalties — a small % of each chip's selling price, collected for the product's life (often 10–20+ years). This is the
recurring, high-margin engine.
Key economics:
• Armv9 (current flagship architecture) carries roughly ~2x the royalty rate of Armv8 and is still ramping toward a long-term
attach target of ~60–70% of royalty revenue.
• CSS (Compute Subsystems) — pre-integrated, validated IP "blueprints" — carry a higher per-chip rate (>10%) and already represent
~20% of royalties and growing.
Reach: Arm designs power ~99% of the world's smartphone application processors and a fast-growing share of data-center, automotive,
and IoT/edge silicon.
Ownership: SoftBank Group holds ~87% of ARM — a small public float that amplifies volatility.
Strategic shift: In March 2026, ARM shipped its first in-house production chip in ~35 years — the "Arm AGI CPU" (a data-center CPU
for AI inference), with Meta as lead customer. This is the biggest strategic change in company history (and, as noted in Risks, a
source of channel conflict with its own licensees).
3. Financial Performance (FY2026, fiscal year ended 2026-03-31)
• Total revenue: ＄4.92B (+23% YoY)
• Royalty revenue: ＄2.61B (record) (+21% YoY)
• Licensing revenue: ＄2.31B (record) (+25% YoY)
• Non-GAAP EPS: ~＄1.77 (record)
• Adjusted gross margin: ~98%
• Non-GAAP operating margin: ~47%
• Q4 FY2026 (quarter ended 2026-03-31): record revenue ~＄1.49B (+20% YoY); data-center royalty more than doubled YoY; Arm's
self-reported CPU share among top hyperscalers reached ~50%.
• Three straight years of 20%+ revenue growth — growth and profitability simultaneously.
• ARM guides only one quarter ahead; next earnings catalyst ~2026-07-29.
Note on margins: GAAP operating margin has compressed (rising R&D/opex from the in-house chip push), even as the IP model keeps
gross margin near 98%.
4. Growth Engine — AI & Data Center
The data-center/AI opportunity is the real story behind the re-rating:
• Data-center royalty grew >2x YoY — the fastest-growing part of the business.
• Arm CPUs sit inside the leading AI systems: Nvidia Grace and next-gen Vera (Rubin platform), Amazon Graviton, Microsoft Cobalt,
Google Axion — all Arm-based.
• Hyperscaler share: Arm's own metric puts Arm-based CPUs at ~50% of compute among top hyperscalers. Note: this is a narrow,
self-defined cut; independent trackers put the broad data-center CPU market closer to 20–25%. Both can be true (different
denominators).
• Own-silicon optionality: the Arm AGI CPU (Meta as lead customer) reportedly has >＄2B of FY27–FY28 bookings. However, material
revenue is not expected until ~2027 — today it is optionality and narrative, not earnings.
• Capex tailwind: combined 2026 hyperscaler capex is guided around ＄725B (+~77% YoY); every Arm-based data-center CPU (whether
Nvidia Grace, a hyperscaler's custom chip, or Arm's own) carries a royalty.
5. Valuation
ARM is the most expensive large-cap semiconductor name on every multiple:
• ARM: ~＄352 · Mkt cap ~＄378B · Trailing P/E ~414x · Fwd P/E ~200x · P/S ~76x
• NVDA: ~＄196 · Mkt cap ~＄4.7T · Trailing P/E ~30x · Fwd P/E ~20x · P/S ~19x
• AMD: ~＄519 · Mkt cap ~＄847B · Trailing P/E ~173x · Fwd P/E ~60x · P/S ~23x
• AVGO: ~＄382 · Mkt cap ~＄1.8T · Trailing P/E ~64x · Fwd P/E ~24x · P/S ~24x
• Industry-average semiconductor P/E ≈ ~24x — ARM trades several times richer than peers.
• Street consensus: "Moderate Buy" (roughly 19 Buy / 7 Hold / 1 Sell), but the average price target (~＄280) sits ~15–20% BELOW the
current price — the market has already priced in near-perfect execution. High target ＄500 (Mizuho/Bernstein); low ~＄140.
• Mid-June saw a wave of bullish PT raises (＄460–500 from UBS/BofA/Mizuho/Bernstein/TD Cowen) alongside at least one
valuation-driven downgrade (New Street, Buy→Neutral).
Interpretation: The quality is not in dispute; the price of that quality is. At ~200x forward earnings, the stock needs the
AI/data-center ramp and the AGI-CPU optionality to largely come good just to justify today's level.
6. Macro Backdrop & The Recent Drawdown
The move: ~＄120 (Mar 2026) → all-time high ~＄445–455 (6/17–6/18) → ~＄352 (6/25); roughly +235% YTD at the peak, then a ~21%
peak-to-trough drawdown.
Why it dropped — macro/sector-led, not company-specific:
1. Hawkish Fed (6/17): the dot plot removed the expected 2026 rate cut and tilted toward a hike — a direct headwind to
long-duration, high-multiple growth stocks.
2. Asian tech sell-off (6/23): Korea's KOSPI fell ~10% intraday (Samsung/SK Hynix −12%+, SoftBank −15%), dragging the global AI
trade.
3. AI-trade unwind / rotation into defensives; no ARM-specific bad news drove the decline.
The "fell more than its sector" thesis — single-day move on 6/23:
• ARM: −10%
• SOXX (semis ETF): ~−8%
• AMD: ~−6%
• NVDA: ~−4%
Peak-to-6/25, ARM (~−21%) fell roughly 2x the semiconductor index (~−8% to −12%). This supports the "relative opportunity" framing
— ARM is cheaper relative to where it was and dropped more than peers.
The balanced view: ARM also rose far more than peers on the way up (high beta ~3.8 cuts both ways). Falling more after rising more
is symmetric high-beta behavior, not proof of mispricing — even after the drop, ARM remains the most expensive chip name and trades
above consensus fair value. "Fell more = bargain" is a higher-risk, higher-conviction bet, not a data-confirmed fact.
7. Technical Analysis (Short / Medium / Long Term)
Levels derived from the March→June price action; directional, not precise.
Resistance (overhead):
• ＄400 — first overhead hurdle / round number on any bounce.
• ＄440–455 — the all-time-high zone; major resistance (＄450 psychological).
Support (below):
• ＄350 — current level and round-number support; the immediate battleground (short-term).
• ＄282 — ~50-day moving average and roughly the analyst consensus target (medium-term).
• ＄172 — ~200-day moving average / long-term trend reference.
Read: Price sits far above its 200-day MA (~＄172), reflecting how extended the rally was — a wide gap between price and trend that
reinforces the mean-reversion / high-volatility risk.
• Medium term: ＄282 (50-day MA ≈ consensus target) is the key value reference.
• Long term: the uptrend is not threatened unless price loses the ＄200/＄172 region.
8. Bull Case vs Bear Case
Bull case:
• Royalty flywheel: Armv9 (~2x rate), CSS (>10% rate), and data-center royalties (>2x YoY) all ramping simultaneously.
• Direct beneficiary of the ~＄725B (+77%) 2026 AI capex wave, regardless of which hyperscaler wins.
• Own-chip optionality (AGI CPU with Meta, >＄2B FY27–28 bookings) plus ~98% gross margin and recurring revenue.
• The correction left ARM down more than its sector — a relatively cheaper entry vs the peak.
Bear case:
• Valuation: ~200x forward P/E, ~76x sales — priced for perfection; consensus target below price.
• High beta (~3.8) → brutal in any AI-trade unwind (already −21% from the high).
• Regulatory/legal overhang: FTC antitrust probe (2026), prior Qualcomm litigation loss, and channel conflict from competing with
its own largest licensees.
• Customer concentration (top 5 ~57%; Arm China ~16% and not controlled), ~18% China exposure, smartphone cyclicality, and customer
in-housing.
• AGI CPU contributes ~zero revenue until ~2027 — the near-term multiple rests on the existing royalty/licensing engine.
9. Investment Thesis & Recommendation
ARM is a structurally outstanding business at a demanding price. The "AI tollbooth" thesis is intact and the data-center ramp is
real, and the recent sector-relative drawdown does create a more attractive entry than the June peak. But valuation and beta
dominate the near-term risk/reward: with the stock above consensus fair value and earning ~200x forward, downside in any further
AI-trade wobble is meaningful.
Suggested approach (risk-management framework, not a directive):
1. Stage entries — avoid a single full-size purchase; scale in over time/levels.
2. Size small — treat as a high-beta satellite position, not a core holding.
3. Watch the levels — reassess if ＄350, then ＄282, give way.
4. Mark the calendar — next earnings ~2026-07-29 is a key volatility event.
10. Key Figures at a Glance
• Price: ~＄352
• 52-week range: ＄100–＄453
• Market cap: ~＄378B
• FY2026 revenue / growth: ＄4.92B / +23%
• Royalty / Licensing: ＄2.61B / ＄2.31B (both records)
• Adjusted gross margin: ~98%
• Forward P/E / P/S: ~200x / ~76x
• Consensus rating / avg target: Moderate Buy / ~＄280 (below price)
• Beta: ~3.8
• SoftBank ownership: ~87%
• Peak→trough drawdown (Jun 2026): ~21%
• Next earnings: ~2026-07-29