The consensus is wrong. Arm is not pivoting to manufacturing. It is pivoting to control.
When a CFO of a 96% gross margin company tells analysts they are "looking at transactions" in chip manufacturing, the market hears ambition. I hear something else entirely. I hear a company that has reached the logical ceiling of its current business model and is now forced to play a game it was never designed for.
The narrative is seductive. Arm, the British IP giant that powers 95% of the world's smartphones, is "moving into chipmaking." The stock jumps. The talking heads celebrate. But this is a category error. Arm is not becoming a manufacturer. It is becoming something far more dangerous to its competitors: a gatekeeper.
History doesn't repeat, but it rhymes. The last time a high-margin IP company tried to move down the value chain into manufacturing, it was Rambus in the late 1990s. The result was a decade of litigation, regulatory scrutiny, and value destruction. Arm is smarter than that. But the market is not.
Hook: The CFO's Words, Not the Headlines
Let me start with the data point that matters. In Arm's most recent earnings call, CFO Jason Child was asked about the company's capital allocation strategy. His response, parsed carefully, was not a declaration of intent to build fabs. It was a hedge. "We are always looking at transactions that could accelerate our growth," he said. "But we are very disciplined about capital allocation."
That is not the language of a company about to spend $20 billion on a fab. That is the language of a company that knows its stock is priced for perfection and needs to keep the narrative alive.
The real story is buried in the supply chain. Over the past 12 months, Arm has been quietly building a "Total Design" ecosystem. This is not about manufacturing chips. It is about commoditizing the design-to-manufacturing handoff. Arm wants to be the operating system for chip creation, not just the instruction set.
Context: The Architecture of Dependence
To understand what Arm is actually doing, you must first understand the structural asymmetry of the semiconductor value chain.
Arm today sits at the highest margin point in the industry. Its 96% gross margin is a function of a simple reality: it owns the instruction set architecture that nearly every mobile device and an increasing number of servers rely on. But owning the architecture is not the same as owning the value.
The real value in AI chips is captured by the full-stack players. NVIDIA does not just design GPUs. It owns CUDA, the software ecosystem, the networking stack, and the supply chain relationships. Arm, by contrast, licenses an IP block and walks away. The customer then takes that block, designs a chip around it, and sends it to TSMC or Samsung for fabrication.
Arm captures the architect's fee. NVIDIA captures the entire building.
The pivot to "manufacturing" is actually a pivot to full-stack. Arm wants to offer its customers a complete design-to-fabrication service. This is not a new idea. Marvell and Broadcom have been doing this for years with custom ASICs. But Arm has something they don't: the CPU architecture that 99% of those custom chips will be built around.
Based on my experience auditing over 200 ICO whitepapers in 2017, I learned that the most dangerous narrative is the one that sounds most plausible. The Arm-to-manufacturing narrative is plausible. It is also wrong.
Core: The Virtual Fab Model
Let me walk you through the actual data. Arm's FY2024 capital expenditure was less than 5% of revenue. A typical semiconductor manufacturer like TSMC spends 35-50% of revenue on capex. If Arm were to build a single advanced fab, its capital intensity would increase by an order of magnitude. Its gross margin would collapse from 96% to the 30-40% range. Its valuation multiple would follow.
Arm's market cap is roughly $120 billion. A single 3nm fab from TSMC costs approximately $20 billion. Arm would need to dilute its equity by nearly 20% to fund one fab. And that fab would take three to four years to reach volume production. By then, the technology cycle would have moved on.
This is not a strategy. This is a suicide pact with shareholders.
The actual strategy is far more elegant. Arm is building a "virtual fab" model. Here is how it works:
- Arm locks in advanced capacity at TSMC and Samsung through pre-payment agreements.
- Arm offers its large CSP customers (AWS, Google, Microsoft) a bundled service: Arm IP + reference design + guaranteed capacity.
- Arm charges a premium for the capacity allocation, effectively turning its IP licensing into a capacity brokerage.
The margins on this model are lower than pure IP licensing, but they are still 50-60% gross margin. More importantly, this model increases switching costs. A customer that has designed a chip around Arm's IP and secured capacity through Arm's relationships is far less likely to migrate to RISC-V.
Risk isn't what you can see. It's what you don't. The risk here is not that Arm builds a fab. The risk is that Arm becomes a bottleneck for capacity, and the CSPs start building their own direct relationships with TSMC and Samsung, cutting Arm out of the middle.
Contrarian: The Real Enemy is RISC-V, Not NVIDIA
The market narrative frames Arm's move as a competitive response to NVIDIA. This is theatrical misdirection.
NVIDIA is Arm's largest indirect customer. The Grace CPU is an Arm architecture product. NVIDIA's AI servers are full of Arm-based management controllers. NVIDIA and Arm are not competitors. They are partners in a common ecosystem.
The real threat is RISC-V.
RISC-V is an open-source instruction set architecture. It is free. It is gaining momentum in the IoT and embedded markets. And it is now targeting the data center. Ventana Microsystems has already demonstrated a RISC-V server-class chip that outperforms Arm's mid-range offerings.
The economics of RISC-V are devastating to Arm's business model. If a customer can design a chip using a free instruction set, why pay Arm a royalty? The answer today is software ecosystem lock-in. The answer tomorrow may be very different.
Arm's move into manufacturing is a defensive play against RISC-V. By offering a complete design-to-fabrication service, Arm increases the cost of switching. A customer that has committed to Arm's total design ecosystem, including capacity guarantees, cannot simply swap in a RISC-V core. The entire supply chain relationship would need to be rebuilt.
Volatility is the fee for admission to the future. Arm is betting that the fee for switching to RISC-V will be high enough to keep its customers locked in.
The Geopolitical Layer: Friend-Shoring as a Service
There is another dimension to this strategy that the market is ignoring. The US CHIPS Act and the global push for semiconductor self-sufficiency are creating a fragmented supply chain. Arm's CSP customers need to source capacity from geographically diverse locations to comply with regulatory requirements.
Arm can become the "friend-shoring" broker. It can lock in capacity in the US (TSMC Arizona), Europe (TSMC Dresden), and Japan (Rapidus). Then it can offer its customers a portfolio of capacity options, all pre-integrated with Arm's IP.
This is a powerful value proposition. A CSP like AWS needs to build its own chips. It needs Arm IP. It needs TSMC capacity. And it needs to demonstrate that its supply chain is not concentrated in Taiwan. Arm can solve all three problems with a single contract.
Code is law, but capital decides who writes it. Arm is using its capital position to rewrite the rules of the semiconductor game.
The Financial Reality: Valuation Strain
Let me be blunt about the numbers. Arm's current valuation is insane. A 70-80x P/E multiple is pricing in 20-25% revenue growth for the next five years. If Arm moves into manufacturing, even the virtual fab model, its margins will compress. The market will re-rate it from a software-like multiple to a hardware-like multiple. That is a 50% downside risk.
I have seen this movie before. In 2020, during the DeFi summer, I identified unsustainable yield rates in lending protocols. I moved capital out of high-yield farming and into protocol-generated revenue streams. The subsequent exploits proved the thesis. The same logic applies here: when the market is pricing in perfection, the downside is asymmetrical.
Arm's CFO knows this. That is why he is talking about "discipline" and "transactions." He is managing expectations. The actual pivot will be slow, incremental, and carefully calibrated to avoid shocking the valuation.
Takeaway: The Cycle Positioning
The semiconductor cycle is entering a new phase. The AI-driven boom has pushed advanced capacity to full utilization. The next two years will see a capacity crunch, followed by a potential overbuild. Arm's window to execute this strategy is narrow.
If Arm can lock in the capacity relationships and the total design ecosystem before the next downturn, it will emerge as a more powerful gatekeeper. If it fails, it will be stuck with expensive capacity commitments and a business model that no longer works.
The market is betting on the former. Based on my experience navigating the 2022 Terra-Luna liquidation, I know that the market is often wrong about the timing, but rarely wrong about the direction.
Arm is going to move into manufacturing. Just not in the way the headlines suggest. The real story is about control, not concrete. And in the semiconductor game, control is the only thing that matters.
The question is not whether Arm can build a fab. It is whether Arm can build a wall around its customers before the RISC-V erosion begins. That wall will be made of capacity, not silicon. And it will be expensive.
History doesn't repeat, but it rhymes. The last time a company tried to turn its IP into a vertically integrated monopoly, it was called Intel. We know how that ended.
