Arm doesn’t manufacture a single chip. It licenses architecture — the blueprint every mobile processor and, increasingly, every AI accelerator is built on. That business model just got a valuation that would make a hyperscaler blush.
Shares trade at $325.04, giving Arm a market capitalization of roughly $345.8 billion on a trailing P/E of 183.6x. Strip out the trailing noise and look at forward-year ratios and it gets more extreme: Arm’s fiscal 2026 price-to-sales sits at 32.7x, and its EV/EBITDA multiple is 137.3x. For context, most semiconductor franchises settle into the high single digits to low teens on EV/EBITDA once they mature. Arm isn’t priced like a mature royalty business. It’s priced like a moonshot.
The counterargument is that Arm’s economics are genuinely different. Gross margin is essentially software-grade at 97.5%, because there’s no fab, no wafer cost, no inventory risk — just intellectual property changing hands. Return on invested capital lands at 12.4%, respectable but not the kind of number that screams “this justifies a 137x EBITDA multiple on its own.” The bull case rests entirely on a forward narrative: every AI server rack, every custom silicon program at the hyperscalers, increasingly touches an Arm-licensed core. If that narrative compounds for another five years, today’s multiple looks cheap in hindsight. If it stalls even modestly, the multiple has nowhere to hide.
Wall Street’s own targets show the tension. The consensus rating is buy, with an average price target of $293.32 — which is actually below the current share price, implying a modest -9.8% gap between where analysts think fair value sits and where the stock trades today. That’s an unusual signal: a “buy” rating attached to a stock analysts collectively think is already ahead of their own price targets. Not every analyst agrees on how far this goes — the range runs from a low of $140 to a high of $500, a spread wide enough to reflect real disagreement about whether AI licensing revenue scales the way bulls expect.
That disagreement sharpened in June. TD Cowen lifted its target to $475 from $265, and UBS moved to $470 — both roughly 80% upward revisions inside a matter of weeks. When two major desks nearly double their targets simultaneously, it’s rarely because they modeled a small tweak to royalty rates. It signals a structural re-rating of how much of the AI compute stack analysts now expect Arm-based designs to capture, from data center CPUs to edge inference silicon.
The honest read: Arm is a call option on the AI chip architecture war dressed up as an equity. The company’s economics are excellent. The multiple already assumes those economics get significantly better from here. Anyone buying at these levels isn’t investing in what Arm is — they’re underwriting what Arm has to become.
