HONEST VERDICT: ARM = exceptional business, indefensible valuation, weakening software moat, AI bypassing its architecture, and a pivot to silicon that trades 35 years of neutrality for revenue that won’t matter until 2028.
WHAT $ARM IS: ARM doesn’t manufacture chips. It designs the blueprint (ISA) that tells every processor how to work, licenses it to Apple, Nvidia, Qualcomm. 280B+ chips shipped. Royalty on every one. 2/20
THE CHINA RISK: 22-25% of ARM revenue from China. Beijing mandating RISC-V adoption for geopolitical independence. Even if Chinese RISC-V underperforms ARM by 30%, state policy overrides economics. This is the most credible structural revenue risk. 19/20
THE CONSEQUENCE: ARM just gave its three largest data centre licensees maximum motivation to accelerate RISC-V. Eliminate the royalty payment AND escape a new competitor in one move. The long-term royalty losses may dwarf the silicon revenue gains. 18/20
THE FATAL STRATEGIC IRONY: For 35 years ARM was the neutral arms dealer — licensed to everyone, competed with nobody. The AGI CPU makes ARM a combatant against Amazon, Google, Nvidia, and Microsoft. All pay ARM royalties. All now face ARM as a competitor. 17/20
HAS NVIDIA DONE IT? YES — 2 YEARS EARLIER. Grace CPU shipped 2023. Vera CPU (2026) targets agentic AI. Amazon (Graviton), Google (Axion), Microsoft (Cobalt) all built their own ARM chips long ago. ARM arrived late to its own party. 16/20
THE AGI CPU — HISTORIC SHIFT: March 24 2026: ARM built its first chip in 35 years. 136 cores, TSMC 3nm, 300W TDP, 800GB/s memory bandwidth. Targets agentic AI data centers. Meta is lead customer. Revenue impact: NOT UNTIL 2028. 15/20
LAYERS 3 + 4 — STILL REAL: 10,000+ patents create legal cost. Apple’s A-series has 6 years of ARM-specific OS and compiler tuning. Migrating = billions + years of performance regression. These switching costs are genuinely high. 14/20
THE REAL INSIGHT: Big tech isn’t REPLICATING ARM. They’re BYPASSING it. Google TPU for training. RISC-V + AI compilers for inference. Amazon Graviton, Google Axion, Microsoft Cobalt — all ARM-based but internally built. No ARM silicon purchase needed. 13/20
LAYER 2 — SOFTWARE MOAT: 35 years of compiler optimizations. WAS the deepest moat. AI eroded it — Google’s MLIR/TVM auto-generates optimized code for ANY ISA. What took decades can now be built in 2-3 years. 12/20
LAYER 1 — ISA: ALREADY REPLICATED. RISC-V hit 25% global processor market share by Jan 2026, growing 30% CAGR. The instruction set itself is no longer proprietary in the market. But it’s ARM’s least valuable asset. 11/20
THE MOAT HAS 4 LAYERS — each with different replicability:
1.ISA instruction set
https://t.co/GHmKcWm2Qa ecosystem
3.Patent portfolio
4.Customer switching costs
One is already replicated. Two is being eroded. Two remain real. 10/20
THE AGENTIC AI THESIS: Training AI = GPU-heavy → Nvidia wins. RUNNING AI agents continuously = CPU-efficient → ARM wins. Power-efficient CPU orchestration at inference scale is ARM’s lane. This thesis is genuinely new and credible. 9/20
THE TECHNICAL SETUP: Fell from $417 ATH to $298 low in 7 days (-27%). Confirmed pivot top sell signal June 3. All short-term MAs acting as resistance. 13.29% short interest. SoftBank -9% amplifies every down day via 90% ownership link. 7/20
THE MOST ALARMING SIGNAL: 25 analysts rate ARM “Buy.” Their consensus 12-month target is 16% BELOW the current price. When the bull case is below current levels, the fundamental analysis is complete. 6/20
THE VALUATION PROBLEM: 486x trailing P/E. 189x forward P/E. PEG ratio 6.16. Earnings yield 0.21%. The risk-free 10yr Treasury pays 4.54%. You earn 22x MORE lending to the US government than ARM earns per dollar invested. 5/20
THE FINANCIALS: FY2026 revenue $4.92B (+23% YoY). Net income $904M (+14%). But EPS grew only 5% last year and MISSED estimates by 7.7%. Revenue is growing faster than profits — margins compressing despite the 97% gross margin. 4/20
THE BUSINESS MODEL: 97% gross margins. Net cash $3.08B. Zero meaningful debt. Asset-light royalty machine — costs almost nothing to license one more chip design. One of the finest business models in technology. 3/20
The market is celebrating record Q1 beats while the majority of companies guiding for the rest of the year are telling you earnings will disappoint. That is the central contradiction of this entire earnings season, sitting in the FactSet database in plain sight.