Arm Holdings (ARM 0.49%) has been a strong stock to own since it went public in September 2023. After hitting the market at $51 per share, Arm's stock has since tripled in price to $159 as of this writing.

Although Arm's valuation looks sky-high at 99 times forward earnings, investors are taking an optimistic view, likely in anticipation of Arm-based chips taking market share in data center computing. Known for their power efficiency, Arm-based chips have long dominated in smartphones. Yet with power concerns coming to the fore as AI data centers require massive amounts of electricity, Arm-based chips are now penetrating this high-growth vertical.

But Arm executives apparently don't think licensing its architecture is enough anymore. Reports are that Arm is about to embark on a huge strategy change -- and it could be an extremely risky bet.

Arm seeks to make its own chips

Last week, Reuters reported that Arm was attempting to poach top executives from some of its biggest customers, with the goal of designing its own Arm-branded chips for data centers. According to people familiar with the matter, the new Arm chip is supposed to be ready by this summer. Following the Reuters report, the Financial Times then reported Meta Platforms (META -2.76%) had already signed up to be the first customer for the new Arm chip. In addition to poaching executives at top tech companies, it has also been rumored that Arm could acquire private chip designer Ampere to bolster the effort.

Up until now, Arm has only licensed its architecture to other chipmakers. Arm typically charges an upfront licensing fee to its customers that ranges from $1 million to $10 million, then a royalty fee on every sale of an Arm-based chip, which typically ranges from 1% to 2% the chip's revenues. Current customers include industry heavyweights such as Nvidia (NVDA 0.40%), which licenses Arm architecture for its Grace CPUs, and Apple, which makes Arm-based PC and smartphone chips, among many others.

However, it appears current CEO Rene Haas and Softbank (SFTB.Y 2.21%) Chairman Masayoshi Son don't believe licensing revenues is enough. Of note, Softbank still owns some 90% of Arm Holdings, so Softbank can dictate the direction in which Arm will go.

While the licensing model is very capital-efficient and low-risk, royalty revenue pales in comparison to the revenue and profits one can make from selling an expensive AI chip at full price.

Son sees an AI opportunity too big to ignore

The move specifically into data center chips could be Masayoshi Son's big AI gambit. After all, Son has recently made bold and ambitious comments regarding the future of artificial intelligence. Late last year, Son claimed "artificial super-intelligence" was coming by 2035, and that it would cost $9 trillion in investment to realize, noting:

I have predicted it would take 400GW of AI datacenter power, that's bigger than the total US electricity supply and it will require 200 million chips. ... The cumulative capex is $9 trillion -- how do you recoup? It's too much investment for many people's view, but I say it's still very reasonable capex: $9 trillion is not too big, maybe too small.

Then in early 2025, Son, in conjunction with OpenAI's Sam Altman and Oracle's Larry Ellison, announced Stargate, a massive $100 billion AI data center project, which also has the potential to grow to $500 billion. Softbank is also reportedly interested in making a direct investment in OpenAI.

Given his view on AI, it's no wonder Son wants to participate in a bigger way.

Image source: Getty Images.

But this move could be highly risky

Son is clearly bullish on AI and the massive investments needed to reach artificial super-intelligence. However, Arm's new move could be highly risky.

After all, the move means Arm will now be competing against its customers. That could potentially limit Arm's appeal versus the competing x86 architecture. Large tech companies that design their own chips tend to prefer neutral third-party partners. For instance, Arm's customers tend to use use Taiwan Semiconductor Manufacturing for their manufacturing, as TSMC is a pure foundry that doesn't compete against its chipmaker customers.

If Arm does go ahead to produce its own chip, it's possible these third-party chipmakers could begin to make custom x86 chips. Of note, Amazon (AMZN -0.89%) announced back in September 2024 that it would design its own custom Xeon processor based on the x86 architecture.

While Arm-based chips have typically excelled in low-power applications and x86 has typically dominated high-performance computing applications, x86 chipmakers have also recently announced "e-cores" over the past few years that will make x86 chips more power-efficient, thereby closing the gap with Arm in that characteristic.

Is Son over-compensating for a huge mistake?

Of note, Softbank sold its entire stake in Nvidia back in 2019 for $4 billion. Had Softbank held onto shares, that stake would be worth more than $150 billion today. Moreover, Softbank didn't have any notable AI bets when ChatGPT shook the world back in late 2022.

Thus, Son may be directing Arm down a risky path in order to make up for Softbank's huge past mistakes. And making investment decisions out of fear of missing out or to compensate for a past mistake is probably not the best strategy.

As has been the case with Masayoshi Son's boom-or-bust career, Arm's entrance into designing chips is a big bet, but one that could also blow up in Arm's face as well. Arm investors should therefore monitor these developments closely, and watch out not only for adoption of the new Arm chip, but also if existing customers begin designing more on the competing x86 architecture.