Arm Holdings has had a week where it has gone from "the next name in the AI chain" to one of the hottest titles on the market. The share price jumped roughly 46.5% in just five days, moving to a new all-time high of around $306 apiece. Behind such a sharp rise was not just emotion, but a combination of several specific factors: a very optimistic outlook from Bernstein, billions of dollars in orders for a new AI processor for data centers, and record quarterly results that backed up this optimism with numbers.
In other words, the market got a clear AI story, fresh contracts and strong fundamentals all in one moment - and quickly overestimated how big a role Arm could play in AI infrastructure in the coming years. The result was an aggressive repricing valuation that added dozens of percent in one week, while opening up the question of where the growth story ends and the bet on the perfect scenario begins.
What set the stock in motion: 46.5% for the week and a new ceiling
During the last five trading days of the previous week, Arm Holdings $ARM stock jumped roughly 46.5% to finish at all-time highs around $306.5. The main immediate impetus was the news that customers had committed in advance to about $2 billion of Arm's new autonomous silicon designs for the coming years, double the original internal expectations. Growth was thus not just based on "AI hype" but on a concrete backlog of orders for the new product direction.
These numbers came shortly after the company reported record fourth fiscal quarter results, so the positive fundamentals and new outlook combined into one big "perfect storm" moment for the bulls.
Bernstein: fivefold earnings growth and the return of CPUs to the center of AI
The catalyst for sentiment was new coverage from investment bank Bernstein. The latter initiated a recommendation on Arm with an "Outperform" rating and a $300 price target, slightly below where the stock ultimately ended up after the rally, but clearly above the then-market price at the time of the report. The key thesis was:
Arm's profits could grow roughly fivefold by 2030
Arm has room to significantly increase its share of the CPU market in servers and data centres
and the total addressable market it is targeting is estimated at around $137 billion
At the same time, analyst David Dai pointed out that the generative AI paradigm is shifting from "chatbots" (AI 1.0) to "agents" (AI 2.0), autonomous systems that plan, decide and execute tasks on their own. This, he says, brings central processing units (CPUs) back to the center of the action, as they are the ones that coordinate, control and complement specialized accelerators (GPUs, NPUs) in complex AI pipelines. This is where Arm is uniquely positioned, thanks to its architecture and years of experience with efficient CPU cores.
New AGI CPU: $2 billion in orders in just a few months
An important concrete pillar of the story was the new datacenter processor that Arm unveiled in the spring. It was the first datacenter CPU designed specifically for "autonomous AI" tasks - that is, for scenarios where AI systems run large, long, and with high coordination and control requirements.
The company said in March that it had pre-confirmed orders for this AGI CPU of around $1 billion over a two-year horizon. In the ensuing period, that customer commitment has increased to more than $2 billion, double the original expectation. Management explained at the results that data centers designed for autonomous AI require several times more computing power than traditional servers - and that Arm's new generation of CPUs are optimized for these deployments.
The analytical comments added an interesting detail: in so-called agent-based AI systems, the volume of processed "tokens" grows orders of magnitude more than in first-generation generative AI. This means a much higher load on the infrastructure - including CPU - and therefore a higher potential for companies that can effectively design this type of computing architecture.
Record Q4 2026: revenue +20%, profit +49%
Fundamentally, Arm had a very strong fourth fiscal quarter 2026, which provided a solid foundation for the rally. For the period ended 31 March, the company reported:
Revenue of around $1.49 billion, which was roughly 20% year-over-year growth and a new record for quarterly revenue
net income of about $313 million versus $210 million in the same period a year ago
Significant improvement in licensing revenue, which grew by a low double-digit percentage to about $670 million
Licensing revenue was driven by continued adoption of the newer Armv9 architecture and Core System Solutions (CSS), as well as increasing deployment of Arm-based chips in data centers. Importantly, royalty revenues in the data center segment in particular roughly doubled year-over-year - which fits with the thesis that Arm is rapidly moving from the mobile and embedded world into the "heavy" data center business.
CPU market share: quadruple and a $137 billion "pie"
Bernstein estimated in his analysis that Arm's share of the server CPU market could quadruple to roughly fourfold in the next few years. Combined with the fact that the overall addressable CPU market for data centre and AI applications is growing, the bank said this created scope for a multi-fold increase in profits by 2030.
A key competitive advantage that analysts highlighted was energy efficiency:
Arm-based processors have a reputation for being extremely energy efficient
In the era of AI, where data centres face limits on power consumption and heat dissipation, low power consumption is becoming one of the main competitive weapons
this was already reflected in the numbers: licensing revenues in the data centre segment roughly doubled year-on-year, indicating that Arm was really making inroads in this market, and it was not just a "powerpoint" story
SoftBank: from 40 to 260+ billion and a return of over 500%
The clear winner of the rally was Masashi Sonna's SoftBank Group, which held approximately 87% of Arm.
SoftBank had acquired Arm in a 2016 privatization deal for around $32 billion and later increased and restructured its stake, bringing the cumulative investment to an estimated $40 billion
After the IPO and subsequent share price growth, the book value of its stake climbed to over $220 billion
this meant a gross return of around 500-550% on a mark-to-market basis, before any further sales and taxes
What an investor could take from this
Several lessons were offered from the dramatic week around Arm:
AI infra isn't just GPUs - while the market's attention has long been focused mainly on Nvidia-type accelerators, Arm reminded that CPUs that can efficiently run AI agent systems in data centers will also play a key role.
Valuation may be breaking away from reality - 46% growth in a week and a valuation of over $300 billion stands on a combination of real fundamentals and very aggressive expectations through 2030. This is both an opportunity and a risk for investors.
SoftBank's "single bet" worked, but with a lot of attrition - it's not realistic for retail to have 80+% of assets in one company, but Arm's example showed how a concentrated bet on the right technology at the right time can work out.