ARM surged 17% in a single day, with a 330x P/E ratio—what is there to buy?

NEXT PICK · Market Insights

ARM surged 17% in a single day, with a 330x P/E ratio—what is there to buy?

Monday, September 21, 2026

A British company founded in 1990 that doesn't manufacture a single chip today closed at $322.90, up 17.16% in a single day, adding about $50 billion in market value in one day — a figure larger than the total net worth of most A-share listed companies.

What drove it was not the earnings report or mergers and acquisitions, but Meta's Muse AI application exceeding expectations in server CPU demand, plus Piper Sandler issued an "Overweight" rating that day; On the same day, AMD rose 9.13% and Intel 14.16%, boosting the entire CPU sector together.

Trading volume has expanded to 2.4 times the daily average over the past three months, and the stock price has risen 55.8% above the 200-day moving average. But on the same quotation, the rolling P/E ratio is written at 330 times. So the question is very straightforward: what exactly are the people who bought in today?

Latest price $322.90 ▲ +17.16%

Price-to-Earnings Ratio (TTM) 330 times ▲ Forward 135 times

Distance from the 200-day moving average +55.8% ▲ Trading volume is 2.4 times

The rental business with a 97.5% gross margin and the portion of money it cannot recover

Arm doesn't make chips; it sells blueprints and passes: global chip manufacturers pay a licensing fee to receive CPU instruction sets and designs, then pay licensing fees for each chip after mass production, based on shipment volume.

In the most recent quarter, royalties totaled $715 million and licensing fees $574 million, totaling $1.289 billion, up 22% year-over-year. Rolling twelve-month revenue was $5.16 billion, with net profit of $1.04 billion.

The pinnacle of this business lies in a gross margin of 97.54%—selling an extra copy of IP adds almost no marginal cost. But on the flip side, if the gross margin is close to 100%, why is the operating profit margin down to only 17.3%?

The answer is R&D. Arm reinvested most of its gross profits back into architectural iterations, GPUs and neural processing units, and computing subsystems, which is both its cost and the way it renews its moat.

Rather than thinking of Arm as a chip company, think of it as the "landholder" of the entire mobile and embedded world: almost every smartphone's processor is on its ground, and the cost of moving isn't a new chip, but rewriting hardware verification, operating systems, compilers, and middleware for a full set of adaptations.

That's why it dares to reach into industrial IoT, cloud data centers, network equipment, automobiles, and robotics—one instruction set charges the entire industry's toll.

With a 330x P/E ratio and a cash flow yield of less than half a point, I'm buying in 2029

First, let's correct a widely circulated misconception

Arm is not a company with "missing financial data and unvaluable" assets; its accounts are quite clear, but the numbers don't look good.

Rolling P/E ratio is 330 times, forward P/E is 135 times, price-to-sales ratio (P/S) is 66.9 times, and enterprise value multiple is 321 times. Meanwhile, the median forward P/E ratio for the semiconductor industry is only about 22 times.

Cash flow is actually the company's most solid asset

Free cash flow for the most recent quarter was $665 million, with a rolling twelve-month flow of $1.51 billion; Cash on hand was $3.89 billion, total liabilities were only $485 million, and net cash was $3.4 billion.

The problem is, if you put $1.51 billion in free cash flow into a $344.9 billion market cap, the free cash flow yield is less than half a percentage point. Return on equity (ROE) of 13.35% is not impressive.

So clearly, what the market is buying isn't today's $5.16 billion annual revenue, but the version of Arm that will be realized in three to five years by data centers and custom chips.

So, how much of today's 17% bullish candlestick was realized? An uncomfortable reference is: the newly released bullish rating that day had its target price already crossed by the stock price by the close; while the average target price of more than forty institutions was actually about 10% lower than the current price.

Good news doesn't mean it hasn't materialized; it's realized faster than a researcher's pen.

Standing 55.8% above the 200-day moving average—is this momentum or overdraw?

The current price is about 23.7% above the 50-day moving average and 55.8% above the 200-day moving average, with trading volume expanding to 2.4 times the daily average over the past three months—all three features of a breakout phase are present: price momentum, volume coordination, and narrative upgrade.

The breakout stage also means the proportion of short-term funds in the buyer structure is rising rapidly. The meaning of the beta value of 3.89 is simple: for every 1% drop in the market, it drops nearly 4% on average.

Does the market still recognize this stage? This July, the answer was already given once—when the Philadelphia semiconductor index fell 20% from its peak, Arm fell 34% in a single month, and the most expensive valuation was always the first to be sold.

x86 is defending the city, RISC-V is laying the groundwork, NVIDIA is starting anew

AMD and Intel's x86 face-off with Arm in PCs, servers, and workstations. Their microarchitectures and software ecosystems are mature but burden with the heavy capital burden of wafer manufacturing; Arm only sells blueprints and operates lightly, but the price is that it never gets the most expensive segment of chip prices.

RISC-V takes a different path: open source, no patent fees, and has indeed seen growth in IoT, edge computing, and custom accelerators, but its toolchain, business support, and software ecosystem are still in its early stages and cannot replicate Arm's scale in the short term.

NVIDIA uses CPUs like Grace to enter data centers, posing a real threat, but its main battleground is ultimately GPUs and accelerated computing; Synopsys and Cadence's business models are most similar to Arm's—both rely on licensing fees—but they sell design processes, verification, and physical implementation tools, not the instruction sets themselves.

Remember this moat in one sentence

What competitors want to replicate is not just a single chip, but the path paved by thousands of customers, countless compilers, and operating systems over more than thirty years.

On the server and high-performance computing side, the path is clearly much thinner—x86 has a deep foundation, emerging acceleration architectures are constantly emerging, and Arm is the challenger rather than the renter.

A 50% share of the target market, and the yet-to-be-bitten segment of custom ASICs

The seller claims Arm holds about 50% of the CPU IP in its target market—almost a quantitative version of the word "de facto standard." But the truly valuable incremental value is not here.

Incremental growth is in custom ASICs. According to the same study's estimates, even capturing just 10% of the custom ASIC market share is enough to double Arm's current profitability.

This also explains why it chose to personally make AGI CPUs rather than quietly collecting licensing fees. In the most recent quarter, data center royalties more than doubled year-on-year, annualized contract value grew 13% year-on-year, and the direction is clear.

The company has not disclosed the segmentation of the overall market, serviceable market, and accessible market, so these can only be used as directional judgments—the global semiconductor IP licensing market is in the tens of billions of dollars, and Arm's ceiling depends on how far it can push its share in the cabinet.

The bulls' trump card: real cash flow, plus a second curve that is unfolding

The logic behind this round of bullish gains is no longer just the phrase "AI concept," but a set of verifiable anchor points:

▲ Bull Case

The bearish ledger: good company, and a good price that has already been pre-booked

Those bearish on Arm basically don't deny it's a good business; what they deny is the price:

▼ Bear Case

Three things will determine whether you continue to break through or return to your original state

In the following quarters, there aren't many events that truly have the power to change the narrative, but each one carries significant weight:

•	AGI CPU supply chain implementation progress: Continued verification in 2026 to see if it can support the $2 billion revenue target, with a bullish direction
•	Next quarterly report: Company guidance revenue is $1.38 billion, fluctuating by about $50 million, with non-GAAP operating expenses of about $780 million, serving as a direct check of valuation
•	Meta and OpenAI's custom CPU a surge in volume: Starting in 2027, fulfillment is expected to open up the possibilities for custom ASICs, with a more focused direction
•	AI server orders are propagating upstream: Over $60 billion in new orders from Super Micro will be delivered gradually over the next few quarters, with a focus on the main direction
•	Tight memory supply drives up the value of the entire AI chip supply chain: Around 2027, Arm will indirectly benefit as the IP layer, with a more neutral impact

It rose 126% in one year, but it crashed four times that year

Over the past twelve months, Arm's total return was +125.95%, but this curve is far from smooth: after the February 4 earnings report, licensing revenue fell 8% due to $505 million in licensing revenue falling short of the $520 million expectation, and in June, it fell another 10% amid a sell-off led by Korean semiconductor stocks.

The real test comes in July. On July 29, the company delivered a record-breaking quarterly report, with its stock price still down 8% that day, down 34% for the entire July—not because it did something wrong, but because when the Philadelphia semiconductor index pulled back 20% from its peak, the most expensive multiples took the biggest hit.

On September 14, another 7% drop occurred, citing overall risk aversion in the AI sector. On September 17, after the CEO's remark, the stock rose about 8%, and on September 21, Meta's demand signal sparked another spark, reaching 17.16%.

This historical judgment is actually quite clear

Arm's fundamentals are steadily paying off, but its stock price has never been priced by fundamentals—it is by the risk appetite of the entire AI sector.

⚠️ Risk Notice • Valuation extremes: rolling P/E ratio of 330 times, forward 135 times, with very little margin for error • Beta 3.89, with sector pullbacks usually dropping three to four times the market • The analysts' average target price is already below the current price, and the short-term rally has overdrawn sellers' expectations • SoftBank holds about 86.4% of shares, with concentrated control and a fragile tradable listing structure • Revenue spans China, the US, Japan, South Korea, and Taiwan, with export controls and geopolitical policies directly impacting licensing revenue • The $2 billion AGI CPU target has yet to be realized; real results will only be revealed in 2027

🟡 Neutral Business is paying off, and prices have already surpassed the next two years ahead of schedule.

💬 Discussion Good company, high price, don't chase highs. Real-time buy and sell positions post ARM to the official account.

Source

NextPick real-time market snapshot + Arm's latest quarterly earnings report and SEC filing + mainstream financial media public coverage, data reconciliation time after the close of the US Eastern Coast market on September 21, 2026.