ARM just made its move. See my blog from last week on this profound step: A View from the Watchtower: ARM - From Architecture to Platform
ARM stepped up the stack — from architecture to platform — signaling that the future of AI infrastructure will be defined at the system level, not just at the chip level. But at the same time, something equally important is happening beneath it. The industry is beginning to ask a different question: What if no one should own the blueprint at all?
We are entering a phase of the semiconductor industry where multiple architectural models are not just coexisting — they are all accelerating at once. ARM is moving up the stack into platform ownership. RISC-V is expanding as a sovereignty-driven alternative. x86 (Intel and AMD) continues to anchor enterprise and cloud infrastructure. And NVIDIA has effectively defined the modern AI system architecture — where GPUs, interconnect, and software form the center of gravity. But this is not fragmentation. This is convergence under a single economic force: AI.
AI is not just another workload. It is the defining economic driver of the next decade:
And in that world, no single architecture is sufficient. Instead, we are seeing a new equilibrium emerge:
Different roles. Same system.
Imagine the system now being deployed in a modern AI data center. It is no longer a single architecture. It is a composition:
Each of these elements comes with its own:
Individually, they work. But together, they form a system that no single vendor fully owns — and no traditional design flow was built to manage. So the question is no longer:
Can we design each component correctly?
The question becomes:
How do we coordinate the integration, validate the full system behavior, and support it over time — across architectures, vendors, and geopolitical boundaries?
That is a system problem.
For decades, the semiconductor industry operated on a simple premise: You could trust the foundation. ARM became the most successful embodiment of that idea:
It wasn’t just technology. It was institutional trust. Design teams didn’t need to question the ISA. They didn’t need to validate the foundation. They could build. And that model scaled to hundreds of billions of devices.
RISC-V doesn’t just introduce a new architecture. It removes the idea that architecture must be owned.
At first, this looked like a cost play. It isn’t. It’s a control play. Viewing recent announcements from the Watchtower makes this very clear.
If sovereignty is becoming a design requirement, China is the clearest example of why. Export controls and access restrictions have forced a fundamental shift in strategy: Dependency on foreign-controlled architectures is no longer acceptable. RISC-V provides a path forward:
This is not a theoretical shift. It is already happening: Chinese hyperscalers and semiconductor companies are actively investing in RISC-V-based designs — including AI-oriented processors — as part of a broader push toward technological independence. And this has ripple effects across the entire industry:
This doesn’t mean RISC-V replaces ARM or x86. It means the world is no longer operating under a single trusted blueprint. RISC-V is gaining traction not because it’s “better” in a traditional sense. It’s gaining traction because the world has changed.
In that environment, dependency becomes risk. And RISC-V offers something fundamentally different: The ability to build without permission.
If ARM’s model was built on centralized trust, then RISC-V’s model introduces distributed responsibility. Anyone can implement it. Anyone can extend it. Anyone can modify it. That is certainly a powerful concept. But it creates a new question: Who validates the system?
In a world of AI systems and chiplet-based architectures, this question gets sharper:
This is not just about design. This is about provenance. The lineage of the system. The traceability of decisions. The integrity of the lifecycle. And as systems become more distributed, provenance becomes harder to establish — and more important to prove.
This is not a future scenario. This is the system being built today.
We are now entering a fascinating world with two competing models:
Each solves one problem. Neither solves the whole system. Because the real challenge has shifted.
For decades, the industry focused on:
What architecture should we build on?
Now the question is different:
How do we coordinate and validate systems built across multiple architectures, vendors, and organizations?
Because modern systems are:
And increasingly:
This is not an ISA problem. This is not a traditional EDA problem. This is a coordination and traceability challenge at the system scale. EDA 1.0 enabled design. EDA 2.0 optimized the flow. EDA 3.0 begins when:
The system — not the chip — becomes the unit of design
And the ecosystem — not the company — becomes the unit of execution
In that world, what matters is not just:
If ARM represents control through ownership, and RISC-V represents freedom through openness, then the next layer must solve something different: Trust across independence. Because in a heterogeneous system:
But you are still responsible for the outcome.
The industry’s first great blueprint was architectural. It allowed the world to scale compute. The next blueprint will be different. It won’t define instructions. It will define ecosystem coordination.
ARM’s move up the stack signals a world where platforms matter more. RISC-V’s rise signals a world where control is being redistributed. Together, they point to the same conclusion: The semiconductor industry is no longer just designing technology. It is managing trust at scale. And in a world of open architectures, distributed systems, and geopolitical pressure:
Trust no longer comes from ownership.
It comes from traceability.
That is both the shift and the gap. And that is where the next era of semiconductor design — EDA 3.0 — will be defined.