Arm and the smartphone curse: when AI hype isn't enough to save the stock

Today's tech market looks like a casino where every table is betting on the same two letters: AI. Arm, the British holding company whose architectures sit in the pocket of nearly every person on the planet, looked like the perfect favorite in that game. But the company's latest earnings report hit optimists like a cold shower. Even though the company is setting records in the server segment, its shares fell 8%. It is a classic story of inflated expectations colliding with the inertia of the real world.

Let's work out what went wrong. On paper, Arm looks decent. Revenue is growing, and the move to the new Armv9 architecture is running ahead of schedule. For those who don't follow core revisions: v9 lets Arm collect twice as much in licensing fees per chip as the previous generation. That is a "goose that lays golden eggs," especially at a time when Apple, Samsung and other giants are trying to squeeze the most out of on-device neural network processing. You'd think everything is fine, but there is one catch: market size.

The problem is that Arm has become a hostage of its own dominance in smartphones. That market matured long ago, if not aged. People no longer replace their phones every year, and the "revolutionary" AI features in new models have not yet made mainstream buyers rush to the store. The result is a paradox: every chip sold earns Arm more than before, but not enough chips are sold to satisfy Wall Street's appetite. Investors got used to AI being a rocket, while Arm now looks more like a heavy truck trying to pick up speed on a steep climb.

Meanwhile, the company's server division is doing genuinely well. Cloud giants such as Amazon, Google and Microsoft are increasingly dropping traditional x86 solutions in favor of their own chips based on Arm Neoverse. In data centers, where electricity bills make up the lion's share of costs, Arm's energy efficiency is becoming the deciding factor. Training and running large language models demands enormous resources, and here the British architecture wins the optimization battle. But, unfortunately for SoftBank (Arm's owner), server successes cannot yet fully offset stagnation in the mobile sector, which has been the company's foundation for decades.

The psychological factor also matters. SoftBank chief Masayoshi Son long promoted Arm as the "center of the AI universe." When the bar of expectations is raised into the stratosphere, any slowdown is perceived as a catastrophe. The market expected Arm to show growth rates like Nvidia's, but Arm is not Nvidia. It does not sell ready-made shovels for the gold rush; it sells the blueprints for those shovels. That is a more stable but less explosive business. The current share correction is not a sign of collapse but a return to reality after a long period of euphoria.

There is also the threat of RISC-V looming on the horizon. It is an open architecture that lets companies design chips without paying royalties to Arm. For now it is popular mainly in controllers and embedded systems, but Chinese companies, fearing sanctions, are actively investing in it. If Arm cannot prove that its license is worth the money it asks for, competition will get even tougher. The company is at a transformation point: it needs to stop being a "phone" company for good and become an "infrastructure" company.

The bottom line: AI really does help Arm earn more from every server, but the company still depends too heavily on whether you decide to buy a new smartphone this year. Can the v9 architecture become the standard for PCs and servers faster than the mobile market deflates?