Insights
Semiconductor engineering is moving to the center of the AI economy
For much of the AI boom, attention has stayed near the top of the technology stack: models, copilots, agents, and applications. Gilroy Mathew, COO of UST, sees an equally consequential shift happening underneath them. As AI moves into data centers, factories, vehicles, and edge devices, the engineering required to support it is becoming more tightly connected.
Watch interview: UST COO Gilroy Mathew On AI, Semiconductors & Why Factories Won't Need Human
“2026...is a year of convergence.” — Gilroy Mathew, COO, UST
Mathew calls 2026 a “year of convergence.” It is a useful description of what engineering leaders are confronting. Semiconductor architectures are evolving alongside AI workloads. Manufacturing is becoming more sensor driven. Vehicles are becoming software-defined, while AI infrastructure is pulling data centers back into strategic conversations. People often treat these as separate technology trends. Mathew argues they increasingly belong to the same engineering story.
DIVIDER
The AI conversation is moving closer to the silicon
Semiconductor engineering has become strategically important because AI keeps changing what computing infrastructure needs to do. Mathew points to an interesting example: CPUs. During the GPU-led expansion of AI infrastructure, it was tempting to assume their strategic importance was diminishing. He argues that agentic AI is bringing CPUs back into focus because increasingly autonomous systems require orchestration.
The larger point is more useful than predicting which processor architecture wins the next cycle. AI workloads are changing quickly, and the underlying compute architecture changes with them. For CTOs making long-lived infrastructure and product decisions, that makes flexibility across silicon, software, and systems increasingly important.
Mathew has watched that shift over a long period. UST began developing semiconductor engineering capabilities 17 years ago, building experience across verification, validation, VLSI design, hardware-software integration, and automation. According to Mathew, that work now spans nine of the world’s leading semiconductor companies. What began as a specialized engineering capability increasingly sits inside much larger conversations about AI infrastructure and intelligent products.
DIVIDER
A chip is only one part of the system
The industry’s center of gravity is also expanding beyond chip design. AI has renewed demand for physical infrastructure, particularly data centers, after years in which cloud migration sometimes encouraged the opposite narrative. Mathew says UST is helping customers establish AI data centers in the US, India, and Taiwan, while also supporting hyperscalers developing their own AI chipsets through design, verification, and validation.
That combination reveals something important about where semiconductor engineering is heading. The chip cannot be considered independently of the environment in which it operates. Compute requirements, energy, software, data residency, system architecture, and workload behavior increasingly influence one another.
Sovereignty adds another consideration. Mathew expects countries to place greater emphasis on where data and AI infrastructure reside, increasing the relevance of domestic AI data-center capacity. He also flags the other side of that expansion: energy availability and environmental cost. The demand for compute may be accelerating, but infrastructure economics still have to work.
For CXOs, that makes semiconductor strategy broader than procurement. Compute decisions are becoming architecture decisions, infrastructure decisions, and increasingly business-continuity decisions.
DIVIDER
The ecosystem around the chip may matter as much as the fab
Mathew makes a similar argument about semiconductor manufacturing. Fabrication attracts attention because of its scale and strategic importance, but a functioning semiconductor industry requires considerably more around it. Chips still need testing, packaging, design expertise, components, engineering talent, suppliers, and the institutional capability to connect those activities. That is why he repeatedly returns to one word: ecosystem.
“Semiconductor has to be looked at as an ecosystem.” — Gilroy Mathew
His discussion of Outsourced Semiconductor Assembly and Test (OSAT )illustrates the point. After fabrication, semiconductors still need assembly, testing, and packaging before moving into products. In India, UST and Kaynes Technology have established an OSAT operation in Sanand, Gujarat, combining semiconductor engineering with electronics manufacturing. The first phase begins with 11 production lines on a 46-acre site, with further expansion planned as the operation develops.
The strategic lesson extends beyond one country or facility. Semiconductor competitiveness depends on how effectively engineering, manufacturing, suppliers, academia, and specialist capabilities reinforce one another. Adding capacity without developing those connections risks creating impressive assets surrounded by capability gaps.
DIVIDER
Talent could become the harder constraint
Mathew describes the shortage of semiconductor design resources in India as “very, very high,” making talent a critical constraint as the sector scales. He points to work underway on VLSI design, custom ASICs and chiplet-based customization, alongside collaboration with academia to expand the pool of design talent.
Companies can purchase infrastructure relatively quickly; building experienced engineering teams takes years. For engineering leaders, talent strategy becomes part of technology strategy. The ability to move between silicon, embedded systems, software, AI, and physical products may become particularly valuable as those disciplines converge.
DIVIDER
Physical AI takes the argument beyond the data center
The same convergence is already reaching the factory floor. Mathew describes a future in which factories become “darker” for humans but “lit for the sensors.” His point is not simply that factories will automate more tasks. Connected systems, sensors, AI agents, robotics, and physical AI are beginning to change how industrial environments are designed and operated.
Vehicles offer another view of the same transition. As cars and trucks become more software-defined, the boundary between mechanical engineering and digital engineering becomes harder to maintain. Compute, embedded software, connectivity, telemetry, and AI increasingly influence the behavior of the finished product.
Mathew points to work with a major Indian truck manufacturer as an example. A connected system covering roughly 30,000 trucks was expanded to more than 250,000 vehicles, while an IoT platform collected more than 80 telematics parameters. AI-based tools could then turn that operational data into uses including dealer monetization.
This is where the semiconductor conversation becomes much more tangible. The value of silicon ultimately appears in what the physical system can sense, decide, communicate, and do.
DIVIDER
The next competitive advantage may sit between disciplines
Mathew expects the next several years to bring greater emphasis on platforms and solution-based approaches rather than isolated projects. The bigger implication for business leaders is that competitive advantage will increasingly depend on how effectively technologies that were once managed separately work together.
Semiconductor engineering is therefore moving closer to business strategy. AI is connecting decisions about compute and edge infrastructure with product development, manufacturing, and the physical environments where intelligence ultimately has to perform. For engineering-focused CXOs, scaling AI means understanding these dependencies early enough to avoid unnecessary cost, complexity, and operational risk.
You can no longer treat AI, semiconductor engineering, infrastructure, and physical products as separate investment decisions. Mathew’s idea of convergence is ultimately about connecting those choices so that technology investment translates into better products, more efficient operations, and scalable growth.
Watch the full CNBC-TV18 Voices from the Valley interview with UST COO Gilroy Mathew →
DIVIDER
FAQ
1. Why is semiconductor engineering important to AI strategy?
Semiconductor engineering shapes the compute, performance, power, and system capabilities AI depends on. As AI moves into data centers, edge devices, vehicles, and factories, semiconductor decisions increasingly affect business outcomes.
2. How is AI changing semiconductor engineering?
AI is influencing verification, compute architecture, custom chip design, and edge systems. Gilroy Mathew also highlights renewed CPU relevance as agentic AI increases orchestration demands.
3. What does semiconductor ecosystem mean?
A semiconductor ecosystem includes more than fabrication. It also depends on design, validation, packaging, testing, component manufacturing, engineering talent, academia, and production capability working together.
4. Why does semiconductor talent matter for AI growth?
AI infrastructure and intelligent products require specialized skills across VLSI design, validation, ASICs, embedded systems, and hardware-software integration. Building that engineering depth takes time.
5. What should CXOs consider as AI and semiconductor engineering converge?
CXOs should look beyond individual AI applications and consider the underlying compute, infrastructure, edge systems, and product architecture needed to scale AI reliably without creating unnecessary cost or complexity.