ARM · Arm Holdings plc · Stock research

Refresh of sustained coverage. Reference close $238.78 on 2026-08-24, split/dividend-adjusted daily series — 45.7% below the $439.46 52-week high, RSI(14) printed 27.3, a three-month price change of -25.7%.

ARM — Arm Holdings plc

Current Thesis

The leg on offer is unchanged: a royalty annuity on data-center CPU, priced as though the observed compounding rate persists while the handset base decays. Q1 FY27, reported 2026-07-29, put royalty revenue at $715M (+22% YoY) and licensing at $574M (+23% YoY) on total revenue of $1.289B against $1.262B consensus, adjusted EPS $0.45 versus $0.40, net income $270M. Data-center royalty grew more than 100% year over year for a second consecutive quarter on cumulative Neoverse shipments past 1.5 billion cores, with committed customer demand for the Arm AGI CPU above $2B across FY27–FY28 and manufacturing capacity secured behind more than $1B of it.

What advanced since the last note is one level. The prior note identified $239.06 — where the shares stood before the 2026-08-04 session that added more than 11% on data-center royalty commentary and carried them to $280.56 — as the floor of the post-print range. The 2026-08-24 close of $238.78 sits below it. The entire post-print advance has been retraced and the launch shelf surrendered on a closing print, with the $255 area (where JPMorgan, Guggenheim and Needham all struck targets on 2026-07-30) unreclaimed since it was lost.

Nothing Arm-specific has been disclosed since 2026-07-29. The next company-controlled datapoint is the Q2 FY27 report, roughly ten weeks out. That leaves Nvidia's Q2 FY27 print on 2026-08-26 — guided to $91B ±2% against $46.7B in the year-ago quarter — as the nearest demand read-through available, and it is a third party's number.

The narrative is saturated. Dating: 2026-08-18 through 2026-08-24. A name carried by 42 covering analysts fell 8.02% on 2026-08-18 while the Technology Equipment sector fell 3.80%, bounced 0.55% on 2026-08-20 into a level it could not close above, gave back roughly 3% into 2026-08-21, and then closed through the August launch shelf on 2026-08-24 — the whole sequence on zero company news. July frames the regime: the shares lost 34% across that month and the post-print revision cluster ran one way (RBC $475→$340, Wells Fargo $350→$280, HSBC $315→$230, all 2026-07-30; Rosenblatt to $250 on 2026-07-31; UBS to $320 on 2026-07-30). The reported fundamental line has not failed, which is why the label is not dead. Every directional move of size since 2026-07-14 has originated with a third party.

Bullish and bearish views on Arm Holdings plc

The model's bull view on Arm Holdings plc (ARM), in brief: Q1 FY27 (2026-07-29) cleared both lines: revenue $1.289B vs $1.262B consensus, adjusted EPS $0.45 vs $0.40, net income $270M. The bear view: The named bear argument is a supply constraint, not a demand one. Both cases follow in full.

Bull Case

• Q1 FY27 (2026-07-29) cleared both lines: revenue $1.289B vs $1.262B consensus, adjusted EPS $0.45 vs $0.40, net income $270M. The $1.330B–$1.430B Q2 revenue guide brackets consensus; the $0.43–$0.51 EPS guide starts at it.
• Data-center royalty more than doubled YoY for a second consecutive quarter (JPMorgan note, 2026-07-30), on cumulative Neoverse shipments past 1.5 billion cores.
• Third-party corroboration of the Neoverse line: Amazon disclosed Graviton, Trainium and Nitro together passing a $25B annual revenue run rate, with Graviton commitments nearly tripling quarter over quarter (compiled 2026-08-20).
• A quantified forward book: AGI CPU committed demand above $2B across FY27–FY28, with manufacturing capacity secured behind more than $1B of it (2026-07-29 disclosure).
• TAM support from a fresh third-party estimate: BofA Securities on 2026-08-13 sized server CPU demand at $210B by 2030 on AI-agent workloads, naming Arm alongside AMD and Intel.
• Coverage still skews long into the drawdown: 20 Strong Buy / 7 Buy / 13 Hold / 1 Sell / 1 Strong Sell across 42 analysts, average 12-month target $285.72 (stockanalysis.com, as of 2026-08-24). William Blair's Sebastien Naji maintained Buy on 2026-08-14; New Street's Pierre Ferragu upgraded to Buy at $260 on 2026-07-30.

Bear Case

• The named bear argument is a supply constraint, not a demand one. HSBC's Frank Lee cut to Hold on 2026-07-14 on TSMC 3nm capacity bottlenecks limiting Arm AI server CPU shipments, expecting meaningful additions only in H2 2027 and TSMC prioritising existing customers on new allocation. Shares fell more than 6% that session. HSBC then took its target from $315 to $230 on 2026-07-30. That argument sits on top of Arm's own qualifier: roughly half the $2B-plus AGI CPU book is capacity-covered, the balance is not.
• The royalty guide decelerates by Arm's own numbers. Q2 FY27 royalty growth is guided to roughly 13% against 22% delivered in Q1 FY27.
• Target dispersion is extreme and the low end is not academic: the 42-analyst range runs $125 to $500, with Goldman Sachs at Hold/$150 (2026-07-30) and Morgan Stanley at Hold with its target lifted only to $212 from $202 on the same date.
• Valuation is rate-exposed: a 257.2x trailing P/E and a 0.39% earnings yield at the 2026-08-20 close, against a 30-year yield near 5.329% on 2026-08-18.
• Related-party concentration: SoftBank contributed $193M of Q1 FY27 revenue against $574M of total licensing; BofA sized SoftBank at 25–30% of licensing revenue on 2026-01-17.
• x86 is defending the socket with capital: AMD priced a record $5B bond on 2026-08-13 for AI infrastructure and detailed a target of 80%+ H2 2026 server revenue growth backed by cloud commitments.

Setup & Price Structure

The 2026-08-04 gap has been fully closed. $280.56 was the high of that session; $239.06 the pre-gap reference; the 2026-08-24 close of $238.78 is beneath both, so the August range no longer has a defended floor inside it. $255 — the densest post-print target cluster — was lost on a closing print and has not been reclaimed on any subsequent close.

RSI(14) printed 27.3 on 2026-08-24, below the conventional 30 line. That is a condition of the tape, not a reason: a momentum name 45.7% off its high can hold sub-30 readings through a decline, and the last three sessions produced no company disclosure that would arrest one.

Crowding and positioning observables, stated as observables:

• 42 sell-side analysts cover the name; the $285.72 average target sits roughly 19.7% above the last close, and that gap widened because price fell, not because targets rose — every revision since 2026-07-30 that carried a number moved down except Morgan Stanley's $202→$212 and New Street's initiation at $260.
• Short interest was 12.0% of float at the 2026-07-15 settlement with days-to-cover of 2.92. No settlement since then has been read against the August round trip in this note.
• No US Form 4 data exists for this name. As a foreign private issuer Arm reports on 6-K; third-party "insider selling" claims about ARM have no US filing behind them.
• SoftBank holds a controlling stake and runs a $20B margin facility secured against it; loan-to-value was 13.0% against a 25% policy ceiling at the 2026-08-06 disclosure.
• A major sector print lands 2026-08-26 after the close, one session out.

Catalyst Calendar (next 30 days)

• 2026-08-26 — Nvidia Q2 FY27 results (quarter ended 2026-07-26), after the close. Guidance was $91B ±2%; previews compiled 2026-08-19 put the street near $91.7B revenue and about $2.08 non-GAAP EPS.
• ~2026-08-26 (est.) — FINRA short-interest dissemination for the 2026-08-14 settlement. First scheduled read on whether the 12.0%-of-float base covered into the 2026-08-04 advance or added into the retrace.
• ~2026-09-08 (est.) — Apple annual September product window. Apple is an architecture licensee and the largest single handset volume signal ahead of the December-quarter royalty pool Arm guided to roughly 13% growth.
• ~2026-09-15 (est.) — FINRA short-interest dissemination for the 2026-08-31 settlement. Second consecutive read, permitting a trend rather than a single observation.
• ~2026-11-04 (est.) — Arm Q2 FY27 results. Outside the window, and the first test of the $1.330B–$1.430B revenue guide and the ~13% royalty growth guide. Nothing company-specific is scheduled before it.

Elapsed catalysts

• ~2026-09 (est.) — Arm Holdings 2026 AGM: remuneration policy, auditor renewal, board slate (set by the 2026-08-10 6-K). Governance, not demand; matters only on material opposition, which SoftBank's vote makes unlikely. (passed 16d ago)

What Would Change Our Mind

The structure that carried the post-print bid is already gone: $255 was lost and never reclaimed, and the 2026-08-24 close of $238.78 took out the $239.06 level the 2026-08-04 advance launched from. Confirmation that this is a trend rather than a flush would be a weekly close below $230 — HSBC's 2026-07-30 target, which would then read as an anchor rather than a floor, with Morgan Stanley's $212 the next struck level beneath it. Secondary: Nvidia's 2026-08-26 print landing with hyperscaler capex-digestion language, or passing without a data-center read-through that Arm can borrow.

What would argue the other way, and is checkable: a FINRA dissemination showing the short base materially above 12.0% of float into the retrace; a Q2 FY27 report on or near 2026-11-04 describing data-center royalty in the same 100%-plus terms used on 2026-07-30; or the capacity-covered share of the $2B-plus AGI CPU book rising above the roughly half disclosed on 2026-07-29, which would take the direct rebuttal out of HSBC's argument.

Correlation Notes

• NVDA — the 2026-08-26 print is the proximate driver of the ARM tape this week, and Nvidia's Windows-PC and RTX-class platforms are built on Arm architecture, so the read-through runs through both capex language and design wins.
• AVGO — JPMorgan reiterated Overweight on 2026-08-20 citing an intact Google TPU roadmap and $56B+ projected AI revenue. Custom-silicon strength is a partial substitute for merchant CPU, so the two do not always move together.
• AMD — the direct socket competitor. The 2026-08-13 $5B bond and the 80%+ H2 2026 server growth framing are the specific numbers to watch against Neoverse share claims.
• TSM — HSBC's bear case is entirely a TSMC 3nm allocation argument; TSMC capacity commentary moves ARM without touching Arm's own disclosure.
• SoftBank Group (9984.T) — controlling holder, $20B margin facility against the stake, and a Q1 result on 2026-08-06 that included an $8.2B gain on Intel. Its balance-sheet stress and ARM's price are mechanically linked through the facility's LTV.
• Long-end rates — a 257.2x trailing multiple prices duration; the 30-year near 5.329% on 2026-08-18 coincided with an 8.02% single-session fall on no company news.