Semiconductors made using the Intel 18A production node benefit from both Intel’s RibbonFET implementation of gate-all-around transistor technology and the PowerVia backside power delivery architecture. (Credit: Intel)
The defining constraint in modern computing is no longer demand for compute. It is the ability to deliver exponentially more compute within a fixed power, thermal, and physical envelope.
AI — particularly generative and emerging agentic workloads — accelerates this tension. Datacenters are already power-constrained, with many facilities operating at or near their electrical and cooling limits. At the same time, enterprises and cloud providers are under pressure to deploy increasingly capable AI platforms that demand higher performance, greater memory bandwidth, and sustained throughput. The result is a structural mismatch: AI wants scale, but infrastructure is boxed in.
This tension cascades up the stack. Advanced AI platforms — whether traditional servers, accelerated systems, or purpose-built AI factories — must deliver more compute density per rack, per watt, and per square foot. Achieving that density is not a system-level problem alone. It ultimately depends on progress at the lowest levels of the stack: transistor architecture, power delivery, and manufacturing technology.
With all this said, the implication is clear that semiconductor process leadership is not some abstract race for cool node naming rights. It is a real requirement for enabling the next generation of AI systems to exist at all. This is where Intel’s foundry and process roadmap becomes relevant, not as a competitive talking point, but as a case study in how advanced transistor and manufacturing technologies are being brought to bear to address real, physical constraints.
This analyst insight will dig deeper into how Intel has long been on the front end of transistor and manufacturing technologies, and why this matters more than ever in the sub-2nm semiconductor world.
A History of Firsts That Solved Real Scaling Limits
Intel has continually been a leader in transistor and manufacturing technologies. Across its history, the company has taken industry-first positions in high-volume CMOS manufacturing, copper interconnects, strained silicon, and high-k metal gate technologies. Each of these breakthroughs was introduced to overcome specific physical limits related to leakage, signal delay, or power efficiency. These were not cosmetic improvements; they were structural changes that allowed performance and efficiency to continue scaling when prior approaches had stalled.Just over a decade ago, Intel was the first manufacturer to bring FinFETs (tri-gate transistors) into high-volume production. That transition fundamentally changed transistor geometry and extended the industry’s ability to scale performance per watt well beyond the limits of planar designs. Within a few years, FinFETs became the industry standard.
The pattern is consistent: Intel repeatedly introduces new transistor and process technologies, proves they can be manufactured at scale, and in doing so sets a direction the broader industry eventually follows.
What Intel 18A Actually Delivers
The transition now underway with the Intel 18A process node continues this tradition, but at a moment when constraints are sharper and margins for inefficiency smaller. At the transistor level, 18A introduces RibbonFET, Intel’s implementation of gate-all-around (GAA) transistors. By fully surrounding the channel, RibbonFET improves electrostatic control, reduces leakage, and enables higher drive current at lower operating voltages. These characteristics directly target the power-density challenges facing modern compute workloads.Equally important is PowerVia, Intel’s backside power delivery architecture. By moving power routing to the backside of the wafer, PowerVia reduces routing congestion in the signal layers, improves voltage stability, and enables tighter transistor packing. This is not an incremental enhancement; it is a fundamental rethinking of how power is delivered to address scaling constraints in advanced nodes. And this is a technology that only Intel is delivering today.
Intel 18A is not simply a continuation of manufacturing node efficiency. It represents a deliberate architectural step that combines two major transitions, and that combination sets it apart in the current foundry landscape. Each technology individually addresses a known scaling limit. Together, they reflect a coordinated architectural response to the power, density, and efficiency demands of AI-class workloads. And with 18A, Intel is the only major chip manufacturer bringing both gate-all-around transistors and backside power delivery into production concurrently.
Other leading foundries are moving in this direction, but on a more staged timeline. TSMC, for example, has outlined plans to introduce backside power delivery (referred to as Super Power Rail) in its future 1.6nm-class process. This approach reflects a measured progression: first introduce GAA and assure stable production, then layer in backside power once the ecosystem and manufacturing readiness align.
Intel’s decision to integrate both technologies in 18A is not about aggressiveness for its own sake. Rather, it reflects an assessment that the power and routing challenges associated with AI workloads are already significant. Addressing them earlier allows Intel to improve performance per watt and density without delaying these benefits until a later generation. In practical terms, Intel continues to position 18A as delivering meaningful gains in performance per watt relative to the Intel 3 node, driven by transistor efficiency, improved power delivery, and tighter integration between design and process technology.
Equally important, 18A is not positioned solely for Intel’s internal products. It is a cornerstone of Intel Foundry, intended to support third-party customers facing the same physical constraints and seeking leading-edge manufacturing without relinquishing architectural control.
Development and manufacturing for these technologies is anchored in the United States, including facilities in Oregon and Arizona. While this is important for maintaining a continuous supply chain amid geopolitical events, there is another reason it is critical. By onshoring development and manufacturing, Intel controls the full feedback loop from device research through process integration and high-volume manufacturing.
Looking Ahead to Intel 14A: Continuity, Not Reinvention
If 18A represents Intel’s decision to address multiple scaling constraints simultaneously, the Intel 14A node being developed is best understood as emphasizing continuity rather than reinvention. While Intel has shared fewer public specifics around 14A, the direction is clear. The node is expected to build directly on the architectural foundations established with RibbonFET and PowerVia, extending gains in performance, power efficiency, and manufacturability rather than introducing another disruptive transition.This sequencing matters. By absorbing the complexity of GAA transistors and backside power delivery in 18A, Intel positions 14A to focus on refinement, yield learning, and broader ecosystem enablement. For customers, particularly third-party foundry customers, this reduces adoption risk while preserving access to leading-edge capabilities.
In effect, 14A signals that Intel is not treating advanced process technology as a one-off leap, but as a sustained, multi-generation platform. That continuity is critical in an AI-driven market where infrastructure investments are measured in years, not quarters.
Enabling AI and Future Technologies Through Transistor and Manufacturing Advances
AI did not suddenly appear in the last two years. Many of today’s models and algorithms have existed conceptually for more than a decade. What changed is the availability of compute capable of executing them efficiently at scale. Advanced transistor design and manufacturing are central to that shift. Higher density, improved power efficiency, and better signal integrity directly enable larger models, faster inference, and more sustainable deployment economics.Intel’s process roadmap is closely aligned with these requirements. Both 18A and 14A are designed to support the workloads that increasingly dominate modern datacenters — AI training and inference, high-throughput analytics, and heterogeneous compute architectures operating under strict power constraints.
Here, Intel’s vertical integration becomes an advantage. Insights from chip design inform process development, and manufacturing realities feed back into architecture decisions. That closed loop is difficult to replicate in a fabless model. This advantage is further extended by an often overlooked dimension of Intel’s work: its comprehensive research and development in fundamental technologies. This includes continued work on transistor materials, interconnects, advanced packaging, and quantum computing. Intel’s quantum research, in particular, reflects a long-term view of compute that extends beyond classical scaling. While quantum remains an emerging field, Intel’s engagement underscores the depth of its device-level expertise and its willingness to invest ahead of clear commercial timelines.
The point here is not that quantum computing will replace classical silicon in the near term. (It won’t.) It is that Intel maintains research capabilities spanning current production nodes and future models — an increasingly rare combination in the industry.
Cradle to Grave in a Fabless World
Most of today’s semiconductor ecosystem is built around specialization. Fabless companies focus on architecture and software. Foundries focus on manufacturing execution. Advanced research is often distributed across universities and consortia.What sets Intel apart is that it operates across these three tightly coupled domains:
- Chip design, delivering CPUs, accelerators, and platforms for client, datacenter, and AI workloads
- Transistor and process technology, inventing and industrializing new device architectures
- Manufacturing at scale, operating leading-edge fabs capable of high-volume production
Why Transistor Leadership Now Directly Shapes AI Economics
The connection between advanced transistor technology and AI economics is often implied but rarely stated plainly. AI deployments at scale are governed by cost per unit of useful work, whether that’s measured as tokens generated, inferences served, or models trained within a fixed power budget. As datacenters become power-constrained, the dominant variable shifts from raw performance to performance per watt per square foot.This is where transistor and manufacturing leadership becomes decisive. Technologies such as gate-all-around transistors and backside power delivery are not abstract innovations. Rather, they are mechanisms for extracting more compute from the same electrical and physical envelope. Higher density, improved voltage stability, and reduced interconnect congestion translate directly into lower operating costs and higher sustainable utilization. These characteristics are especially relevant as the market shifts its focus to activating inference.
Foundry relevance follows from this reality. Customers building AI platforms are not simply buying smaller transistors. They are buying the ability to deploy more capability without exceeding power, cooling, or footprint limits. Process technology that materially improves efficiency therefore reshapes the economics of AI infrastructure.
Intel’s approach links these layers together. By advancing transistor architecture, power delivery, and manufacturing in concert, and by offering those capabilities through Intel Foundry, the company positions itself not just as a manufacturer, but as an enabler of viable AI economics at scale.
Why This Matters Now
The ramp of 18A and the emergence of 14A should be viewed less as isolated roadmap milestones and more as evidence that Intel’s innovation engine remains aligned with the realities of modern computing.Acknowledging the strengths of TSMC and Samsung does not diminish Intel’s role. Instead, it clarifies it. Intel is competing in a domain where technical depth, manufacturing execution, and long-term continuity increasingly matter as much as scale.
For CIOs, architects, and technologists, the takeaway is straightforward: Intel’s foundry strategy is grounded in sustained transistor innovation, disciplined manufacturing execution, and a cradle-to-grave model that remains rare in a predominantly fabless industry.
ANALYST INSIGHT: Transistor Leadership and Manufacturing Excellence in the Sub-2nm World
The defining constraint in modern computing is no longer demand for compute. It is the ability to deliver exponentially more compute within a fixed power, thermal, and physical envelope. AI — particularly generative and emerging agentic workloads — accelerates this tension. Datacenters are...
moorinsightsstrategy.com