ARM Surges 35% in Two Days: Is the Agentic AI Story Worth Buying at $300?

ARM Holdings has done something extraordinary. In two trading sessions, the British chip designer added 35% to its market value, vaulting from the $175 range to an all-time high of $298. The stock now sits at $304 after-hours, with Bernstein calling for $300 and TD Cowen targeting $265. The catalyst is not a new product launch or an earnings beat. It is something far more powerful: a complete repricing of what ARM means in the age of agentic AI.

The question every trader is asking right now is the same one: is this a structural re-rating or a textbook overbought top?

The Bernstein Bombshell

The trigger was a single research note from Bernstein analyst David Dai. He initiated coverage with an Outperform rating and a $300 price target, forecasting that ARM's sales and profits will increase more than fivefold by 2030, reaching $26 billion in revenue and $9.83 in earnings per share. The current EPS is $1.77. That is a 5x earnings move in five years, and the market just decided to start pricing it in.

Dai's thesis rests on a paradigm shift most retail traders have not yet absorbed. The Gen AI paradigm is rapidly shifting from version 1.0, which was chatbots, to version 2.0, which is agents. Data centers focused on supporting agentic AI require four times the compute power of their traditional counterparts. Agentic AI consumes 1,000 times more tokens than first-generation generative AI applications. That math is what justifies the $137 billion server CPU market estimate by 2030, up roughly fourfold from current levels.

ARM is positioned to capture a disproportionate share of that explosion because of one critical advantage: power efficiency. As data centers hit physical power limits, every watt matters. ARM's architecture, originally designed for battery-powered mobile devices, is the only mainstream CPU design that scales without requiring nuclear-grade power infrastructure.

The Business Model Leverage

Here is what makes ARM unique among chip stocks. Unlike Nvidia which sells physical GPUs or AMD which competes on manufacturing, ARM is essentially a royalty business. The company designs chip architectures and licenses them to virtually every major chipmaker on Earth. Apple, Qualcomm, Amazon, Google, Microsoft, and even Nvidia all pay ARM royalties on chips built using ARM's architecture.

That royalty model creates exponential operating leverage. When AI inference volume explodes, ARM's revenue scales with chip shipments while its operating costs stay nearly flat. RBC Capital just raised its target to $260 from $175 after a stronger-than-expected Q4, citing a doubling of data center royalties. The firm expects over $2 billion in AGI CPU revenue in fiscal 2027-2028, with data center becoming ARM's largest segment.

The Counterpoint Research forecast is even more aggressive. They predict ARM will dominate the server CPU and custom AI processor market within three years, capturing 90% of that space. The custom silicon revolution at hyperscalers, where every major cloud provider is now designing their own ARM-based chips, has turned a smartphone royalty company into the architecture layer of global computing.

The Math on Whether $300 Holds

This is where experienced traders need to be careful. ARM is up 123.8% year-to-date. The stock has nearly doubled in five months. Three days ago it was at $223. The technical setup is screaming overbought.

But the fundamental setup is screaming structural shift.

The bull case is that ARM has the rare combination of clear technological moat, royalty business model leverage, and a market opportunity expanding faster than analyst models can keep up with. The Arm AGI CPU, unveiled in March, is the first dedicated agentic AI data center processor from the company. Production-ready silicon is no longer a roadmap item. It is here.

The bear case is purely about timing. Stocks that move 35% in two sessions almost always retrace 50-61.8% of that move within two weeks. That implies a pullback toward $260-275 is mathematically probable, regardless of whether the long-term story is real. The Jefferies $290 target and Bernstein $300 target are 12-month numbers, not next-month numbers. Buying at $304 means paying tomorrow's analyst target today.

How to Play This

There are three honest approaches.

For the conviction buyer, ARM at $304 is still cheaper than ARM at $475 in 2027 if the agentic AI thesis plays out. The fivefold profit projection by 2030 means even at current levels, the stock is trading at a reasonable forward multiple if you actually believe the 2030 numbers. Position size matters more than entry timing.

For the disciplined trader, waiting for the inevitable retracement is the smarter move. Set alerts at $275 and $260. A 10-15% pullback in a structurally bullish name is a gift. The stock has gone vertical without consolidation, and consolidation always comes.

For the cautious observer, take partial profits if you are already long. Selling 30-40% of a position into a 35% two-day rally is not greed-driven exit. It is risk management. The remaining 60-70% rides the long-term thesis. You cannot lose by booking gains.

The Broader Signal

ARM's move is not happening in isolation. AMD added 8.1% to $447.58, pushing its year-to-date gain past 100%. ASML jumped 6.2% to $1,550 after UBS reinstated it as their top European semiconductor pick. The entire AI infrastructure complex is being re-rated upward as the market finally accepts that this AI capex cycle is not a fad but a structural buildout extending through 2028.

ARM is the cleanest expression of that thesis. No manufacturing risk. No customer concentration. Every chip built with their architecture is a royalty paid. The agentic AI explosion is not theoretical anymore. It is happening, and ARM is collecting tolls on every transaction.

The question is not whether ARM is a great company. It clearly is. The question is whether $304 is the right price to start paying for that quality.

A 35% two-day rally rarely holds its full gain. But a 5x profit story rarely stays available at the same multiple. Pick your conviction. Size your position. The next 30 days will likely give better entries. The next 30 months will likely make those entries look like steals.

I am not a financial advisor. Trade wisely, Comrades.