Arm Holdings

The data-center transition is now visible in the numbers, not just the narrative: data-center royalty more than doubled YoY, Neoverse crossed 1.5B cumulative cores with the last 500M shipping in nine months, and AGI CPU booked demand doubled to >$2B in one quarter. A 98% gross margin, debt-free, $3.9B-net-cash IP model converts that into cash at a 27% TTM FCF margin.

• Royalty and licensing both set Q1 records and both grew >20% — the growth is not one lumpy license deal (royalty $715M +22%, license $574M +23%).
• Data-center royalty more than doubled YoY and, per the earnings coverage, is positioned to overtake smartphones as ARM's largest royalty pool — a mix shift toward higher royalty-per-chip.
• The AGI CPU move into production silicon is being pulled by customers faster than ARM planned: >$2B of demand across FY2027-FY2028 versus the $1B opportunity guided 90 days earlier, with manufacturing capacity already secured for the original $1B.
• Ecosystem lock-in keeps compounding at the top of the AI stack — NVIDIA Vera in full production on ARM, Google Axion as TPU host CPU, an AWS Graviton5 agreement with Meta for tens of millions of cores, Azure Cobalt 200, and Qualcomm's Dragonfly C1000 entering data-center CPU.
• Capital structure is a genuine differentiator: $3,888M cash and short-term investments, no borrowings, and operating cash flow of $902M in a single quarter (+172% YoY).
• The valuation reset is real — the stock is down 34% in July 2026 to $239.06, so the entry multiple is materially lower than it was before the drawdown even though the operating results improved.

GAAP economics are going the wrong way while the multiple still prices perfection: GAAP operating margin fell to 7.1% from 10.8% as opex grew 28% against 22% revenue growth, capex intensity jumped to 15% of revenue, and management trimmed the long-run royalty CAGR to high teens on a double-digit smartphone decline. At ~50x TTM sales and ~100x forward earnings, even a beat sells off.

• GAAP operating margin compressed to 7.1% (from 10.8%): GAAP R&D +29% and SG&A +22% on +22% revenue. The 41.2% non-GAAP margin exists mainly because $343M of quarterly share-based compensation is excluded — SBC is 27% of revenue.
• The long-run royalty growth expectation was cut to high teens from 20%, with the smartphone market expected to decline by double digits; royalty is guided to grow only 'low teens' in Q2 FY2027, so near-term growth leans on lumpy licensing (+~30% guided).
• Moving into production silicon structurally dilutes the model: ARM's own FY2031 frame is $15B of AGI CPU at >30% operating margin versus $10B of IP/CSS at >65%. Mix shift toward silicon lowers blended margin and adds inventory, foundry-allocation and supply-chain risk ARM never carried as a pure licensor.
• Capex intensity stepped up to 15.3% of revenue in the quarter ($197M) against a historical low-single-digit norm — the capital-light story is loosening, and the filing does not explain the step-up.
• Concentration is high and partly structural: $388M of $1,289M revenue (30%) is related-party, largely via Arm China, and ARM's own risk factors flag dependence on a limited number of customers plus SoftBank as controlling shareholder with potentially divergent interests.
• Valuation still discounts the plan, not the present: ~$255B market cap on $5.16B TTM revenue (~50x sales), ~244x trailing and ~100x forward earnings, against an FY2031 target of $25B revenue and >$9 non-GAAP EPS that is five years out. HSBC's July downgrade to Hold ($315 target) cited exactly this plus foundry-capacity constraints.
• Free cash flow quality: the $665M quarterly non-GAAP FCF was explicitly helped by 'favorable timing of receivables collections and tax payments' — the TTM 27% margin is the cleaner read, not the 52% quarterly figure.