For three decades, Arm Holdings built one of the most profitable businesses in semiconductors without ever making a chip. It licensed its processor designs to the world’s biggest technology companies — Apple, Qualcomm, Samsung, Google — and collected royalties on every device sold. A toll booth on the information superhighway, as the old metaphor goes.
Now Arm wants to build the highway itself.
The Cambridge, England–based company announced this week that it will begin designing and manufacturing its own processors, a move that puts it in direct competition with the very customers that generate its revenue. The first chip, based on Arm’s own architecture and fabricated by Taiwan Semiconductor Manufacturing Co., is expected to ship in the second half of 2025, according to Yahoo Finance. The initial target: data center servers, the market where Nvidia and AMD currently dominate and where Intel has been hemorrhaging share for years.
This isn’t a tentative experiment. It’s a strategic escalation.
From Licensor to Competitor: Why Arm Is Making the Jump
Arm’s decision to enter the chip market directly has been telegraphed for months, but the formal announcement still sent tremors through the semiconductor industry. CEO Rene Haas has been vocal about his belief that Arm’s architecture — already powering virtually every smartphone on the planet and an increasing share of cloud computing workloads — is undermonetized. Licensing fees and per-unit royalties have made Arm consistently profitable, but the margins on selling actual silicon dwarf what any IP licensor can capture.
Consider the math. Arm currently earns a royalty of roughly 1–2% on the selling price of chips that use its designs. Qualcomm’s Snapdragon processors, Apple’s M-series chips, and Amazon’s Graviton server processors all run on Arm architecture. But the companies designing and selling those chips capture the lion’s share of the value. Arm gets pennies. They get dollars.
By selling its own chips, Arm can capture the full margin stack — design, architecture, and product revenue — rather than just the licensing slice. Wall Street appears to agree with the logic. Arm’s stock, which has roughly tripled since its September 2023 IPO, held steady following the announcement, suggesting investors had already priced in the strategic shift.
But the financial logic is only part of the story. The AI boom has created a once-in-a-generation demand surge for high-performance computing silicon, particularly in data centers. Nvidia’s dominance in AI training chips has made it the world’s most valuable semiconductor company, but the market for AI inference — running trained models rather than building them — is far more fragmented and potentially even larger. Arm believes its energy-efficient architecture gives it a natural advantage in inference workloads, where performance per watt matters as much as raw throughput.
The company’s first data center chip will reportedly target this inference market directly. And Arm isn’t going in alone. TSMC, the world’s most advanced chip manufacturer, will fabricate the processors — the same foundry that builds Apple’s and Nvidia’s most advanced silicon.
That partnership matters enormously. Access to TSMC’s leading-edge process nodes (likely 3nm or even 2nm) means Arm’s chips won’t be hobbled by manufacturing compromises. They’ll compete on equal footing, at least in terms of fabrication technology, with anything Nvidia, AMD, or Intel can bring to market.
The Customer Problem
Here’s where it gets complicated. Arm’s existing customers are also its future competitors. And several of them are already unhappy.
Qualcomm and Arm have been locked in a bitter legal dispute over licensing terms, a fight that went to trial in late 2024 and remains partially unresolved. The core disagreement: whether Qualcomm’s acquisition of chip startup Nuvia (founded by former Apple engineers) entitled it to use Arm’s architecture under existing licenses or required a new, more expensive agreement. A jury sided mostly with Qualcomm, but the case exposed deep tensions between the two companies.
Now imagine that dynamic multiplied across Arm’s entire customer base. Every major Arm licensee must now evaluate whether its chip design partner has become a chip design rival. Amazon, which has invested heavily in its custom Graviton processors for AWS data centers, may view Arm’s entry into the server chip market as a direct threat. So might Google, which has been developing its own Arm-based server chips and custom AI accelerators. And Microsoft, which recently unveiled its Arm-based Cobalt server processor for Azure, suddenly finds its architecture supplier competing for the same data center sockets.
The risk of customer defection is real but limited. Arm’s instruction set architecture is so deeply embedded in the global technology supply chain that switching away from it would take years and cost billions. RISC-V, the open-source alternative, has been gaining traction in certain markets — particularly in China, where geopolitical pressures have accelerated the search for non-Western chip architectures — but it remains years away from matching Arm’s performance and software compatibility in high-end applications.
Still, the threat is not zero. Every dollar Arm earns selling its own chips is a dollar its customers might have earned instead. And the licensing relationship, which depends on trust and mutual benefit, becomes harder to maintain when the licensor is also a competitor. Intel tried a version of this model for decades — designing, manufacturing, and selling its own chips while also offering foundry services to outside customers — and the inherent conflicts of interest contributed to its current strategic crisis.
Arm’s leadership is clearly aware of the tension. Haas has framed the company’s chip ambitions as complementary to, rather than competitive with, its licensing business. The argument: Arm’s own chips will expand the total market for Arm-based computing, which benefits all licensees through a larger software and developer base. It’s a plausible argument. Whether customers buy it is another question entirely.
The timing of this move also coincides with a broader industry trend. The line between chip designers, chip manufacturers, and chip consumers has been blurring for years. Apple designs its own processors. Amazon designs its own server chips and AI accelerators. Google designs its own tensor processing units. Tesla designs its own self-driving chips. The old model — where companies like Arm provided the intellectual property, companies like Qualcomm designed the chips, and companies like TSMC manufactured them — is giving way to something more vertically integrated and more competitive.
Arm’s move is both a response to this trend and an acceleration of it.
SoftBank, which owns approximately 90% of Arm following its IPO, has been pushing aggressively for the company to expand its revenue base. CEO Masayoshi Son has spoken publicly about his vision for Arm as a central player in the AI revolution, not just a behind-the-scenes IP provider. The chip initiative aligns with that vision. SoftBank reportedly plans to invest as much as $100 billion in AI infrastructure over the coming years, and having Arm produce its own high-margin silicon fits neatly into that strategy.
Investors should also watch the competitive response carefully. Nvidia, which has built its data center dominance on its CUDA software platform as much as its hardware, is unlikely to cede ground without a fight. AMD has been gaining share in data center GPUs and CPUs. And Intel, despite its well-documented struggles, still commands a significant share of the server processor market and has been investing billions in new fabrication capacity through its IDM 2.0 strategy.
Then there’s the question of software. Chips don’t sell themselves — they sell because developers write software optimized for them. Nvidia’s CUDA has become the de facto standard for AI development, creating a moat that no competitor has been able to breach despite years of trying. Arm will need to convince data center operators and AI developers that its chips offer a compelling enough advantage in performance, efficiency, or cost to justify the switching effort. That’s a high bar.
What Comes Next
Arm’s entry into the chip market represents the biggest strategic bet the company has made since its founding in 1990. The upside is enormous: if Arm can capture even a small percentage of the data center processor market directly, the revenue impact would dwarf its current licensing income. The global data center chip market is expected to exceed $150 billion annually by 2027, driven primarily by AI workloads.
But the downside risks are substantial. Customer alienation. Execution challenges in a market where Arm has no track record as a product company. The capital intensity of competing with Nvidia, AMD, and Intel, all of which spend billions annually on research and development. And the strategic complexity of simultaneously being a neutral IP licensor and an aggressive chip competitor.
The semiconductor industry has seen this movie before. When Samsung began making its own mobile processors using Arm architecture, it strained relationships with other Arm licensees who viewed Samsung as both a supplier and a rival. When TSMC’s dominance in manufacturing gave it enormous influence over chip design decisions, some customers began exploring alternative foundries to reduce dependence. The lesson: vertical integration creates value, but it also creates conflict.
Arm is betting that the value will outweigh the conflict. That its architecture is so dominant, its technology so advanced, and the AI opportunity so vast that customers will tolerate a new competitor in their midst rather than abandon the platform entirely.
It’s a bold bet. And it’s one that will define the next chapter of the semiconductor industry.
For now, the market is giving Arm the benefit of the doubt. But execution — delivering competitive silicon on time, at scale, with compelling performance — will determine whether this move looks brilliant or reckless in hindsight. The chips, quite literally, aren’t on the table yet.
WebProNews is an iEntry Publication