Custom silicon comes of age

As AI, automotive electrification, industrial automation and personalised healthcare drive unprecedented demand for specialised computing, custom silicon is emerging as one of the industry’s most important growth engines. Ram Vedantham, Head of Semiconductor Delivery at HCLTech, explains why general-purpose processors are increasingly insufficient, how engineering partnerships are becoming critical to success, and why post-silicon validation may be the next competitive battleground.

By HCL

The semiconductor industry is entering a new phase of evolution. For decades, growth was driven largely by increasingly powerful general-purpose processors that could serve a broad range of applications. Today, however, the rapid rise of artificial intelligence, edge computing, advanced automotive architectures and data-intensive workloads is pushing traditional silicon approaches towards their limits.

As a result, enterprises, hyperscalers, OEMs and technology innovators are increasingly turning to custom silicon to achieve the performance, efficiency and differentiation that off-the-shelf devices can no longer deliver.

According to Ram Vedantham, who leads semiconductor delivery at HCLTech, this shift represents far more than a simple technology trend. It signals a fundamental transformation in how products are architected, developed and brought to market.

“We are seeing a fundamental shift between very general-purpose catalogue devices and processors into a more specialised, fine-tuned, end-application-oriented approach,” says Vedantham. “That is what is driving custom silicon growth.”

The perfect storm driving custom silicon

The forces behind custom silicon adoption are numerous, but they share a common theme: modern applications increasingly require domain-specific optimisation.

Artificial intelligence is perhaps the most visible example. AI training and inference workloads demand specialised architectures capable of processing vast quantities of data while maintaining acceptable power consumption. This has created demand for custom CPUs, GPUs, TPUs and AI accelerators designed around specific workload requirements.

The trend extends beyond cloud and hyperscale environments. Edge AI applications require highly efficient processing in constrained power envelopes, while networking infrastructure must evolve to support increasingly complex data flows between distributed intelligent systems.

At the same time, the semiconductor market itself is expanding at a remarkable pace.

“The industry was originally expected to reach the trillion-dollar mark around 2030,” Vedantham observes. “Now it appears that figure may be reached significantly earlier.”

Several sectors are contributing to that acceleration.

In automotive applications, the transition from distributed electronic control units towards zonal architectures is dramatically increasing semiconductor content. The simultaneous rise of electric and hybrid vehicles further amplifies demand for highly specialised silicon.

Healthcare represents another significant growth opportunity. Personalised medicine and advanced diagnostics increasingly require compact systems that combine sophisticated sensing technologies with powerful local processing capabilities.

Industrial automation, factory digitisation and government-related applications are creating similar requirements. Across these markets, organisations seek devices that combine high computational performance with small form factors, robust security and application-specific functionality.

Another powerful driver emerged in the wake of the pandemic.

Supply-chain resilience and design sovereignty have become board-level priorities for many organisations and governments. Rather than relying exclusively on standard solutions sourced through global supply chains, companies increasingly want greater control over critical technology assets.

“We are seeing regions become more autonomous and independent in terms of being self-sufficient for their semiconductor needs,” says Vedantham.

Beyond performance: The real value of custom silicon

While performance remains a key attraction, the value proposition of custom silicon extends well beyond raw computing capability.

Historically, semiconductor design centred on optimising power, performance, area and development time. Those metrics remain important, but new factors are gaining prominence.

Among the most critical is performance-per-watt.

“The AI data centre has changed the discussion,” Vedantham explains. “It’s no longer just about performance. It’s about how much performance you can deliver per watt.”

This is particularly important as AI infrastructure scales. Energy consumption has become both an economic and operational challenge, making efficiency a strategic differentiator.

Custom silicon also enables organisations to create genuine product differentiation. By tailoring architectures around specific workloads and integrating unique capabilities, companies can develop solutions that competitors cannot easily replicate.

This differentiation increasingly relies on close integration between hardware, software and security. Rather than treating these elements separately, many organisations now view them as components of a unified architecture.

“There’s a need to integrate hardware, software and security into a joint ownership model,” says Vedantham.

The result is greater control over intellectual property, stronger competitive positioning and the ability to optimise complete systems rather than individual components.

Supply-chain advantages further strengthen the case. Custom silicon programmes allow organisations to develop deeper relationships across their supplier ecosystems while building greater resilience into future product roadmaps.

Ultimately, however, success still begins with delivering the right power-performance profile for a specific application.

“Power, performance and area remain the starting point,” Vedantham notes. “Everything else builds on that foundation.”

The capability gap

Despite widespread recognition of custom silicon’s strategic importance, many organisations struggle to execute successful programmes.

The challenge is not one of ambition. Rather, it is a combination of economics, expertise and complexity.

Leading-edge semiconductor development is extraordinarily expensive. Advanced AI data-centre-class devices can require investments exceeding hundreds of millions of dollars before reaching production.

Beyond financial commitments, organisations must also address architecture development, verification, physical design, software integration, manufacturing coordination and post-silicon validation.

Verification complexity in particular has become a significant challenge.

As SoCs integrate more functionality, larger numbers of IP blocks and increasingly sophisticated interfaces, validation requirements continue to expand. Modern verification environments must simulate and analyse behaviour across a huge range of operating conditions long before first silicon arrives.

Attracting and retaining skilled engineering talent presents another major obstacle.

“Talent availability is critical,” says Vedantham. “You need strong teams capable of implementing extremely complex architectures.”

The challenge extends across the entire development lifecycle, from front-end architecture and RTL design through physical implementation, firmware integration and post-silicon validation.

Infrastructure requirements represent yet another barrier. Advanced verification platforms, characterisation laboratories and silicon bring-up facilities require substantial investment that many organisations find difficult to justify independently.

These realities are increasingly driving companies towards specialised engineering partners capable of providing expertise, reusable IP, development frameworks and ecosystem relationships.

The rise of strategic engineering partnerships

As custom silicon becomes more mainstream, the role of engineering service providers is changing fundamentally.

According to Vedantham, successful partners must be able to support customers throughout the entire lifecycle, from architecture definition to deployment.

That means far more than simply supplying engineering resources.

An effective partner must be able to contribute system-level expertise, design capabilities, verification methodologies, physical implementation skills, post-silicon validation and software integration.

Equally important is the ability to bridge traditionally separate disciplines.

“We should be there at each phase of deployment,” Vedantham says.

HCLTech’s semiconductor organisation currently combines thousands of hardware engineers with embedded software and firmware specialists, enabling integrated hardware-software development approaches.

The company has also focused heavily on reusable IP blocks, automation frameworks and AI-assisted development methodologies.

AI itself is becoming an increasingly important engineering tool.

“We are developing AI-based frameworks that can significantly reduce development cycles,” Vedantham explains.

Reducing design schedules by even a modest percentage can create substantial competitive advantages in markets where first-mover positioning often translates directly into market share.

Another differentiator is domain expertise.

Semiconductor requirements vary dramatically across automotive, industrial, telecoms and medical applications. Understanding the system-level context surrounding a chip often proves as important as understanding the silicon itself.

As custom silicon adoption expands into new sectors, that combination of semiconductor and application knowledge becomes increasingly valuable.

Building repeatable silicon platforms

Although custom silicon adoption continues to rise, many organisations remain focused on individual projects rather than scalable platform strategies.

Vedantham believes the next phase of industry evolution will centre on transforming custom development into more repeatable and reusable frameworks.

Several technology trends are helping enable that transition.

Chiplet architectures are becoming increasingly important, particularly in high-performance AI applications. By combining multiple specialised dies within a single package, chiplets offer opportunities to improve scalability, reduce development costs and accelerate product development.

Standards such as UCIe are also enhancing interoperability between chiplets and heterogeneous system components.

At the same time, advanced packaging technologies are enabling more sophisticated integration approaches, creating architectures that would have been impractical only a few years ago.

However, these opportunities introduce new complexities.

“The architecture becomes more modular and reusable,” Vedantham explains, “but it also creates significant packaging and design challenges.”

To address those challenges, organisations increasingly seek design-as-a-service models that allow them to focus on application differentiation rather than building complete semiconductor competencies internally.

For many companies, semiconductor development is not a historic core strength. Yet it has rapidly become essential for competitive differentiation.

That creates a growing opportunity for solution partners capable of bridging the gap between system requirements, silicon development and manufacturing ecosystems.

Why post-silicon matters more than ever

As silicon complexity continues to grow, the industry’s attention is increasingly shifting beyond tape-out towards post-silicon engineering.

Historically, much of the industry’s focus centred on design and verification. Today, validation, characterisation, qualification and failure analysis have become equally critical.

This growing importance reflects the scale and complexity of modern devices.

Advanced AI processors can incorporate extraordinarily large die sizes while operating at cutting-edge process nodes. Multiple die stacks, heterogeneous architectures and sophisticated packaging technologies introduce additional validation challenges.

Many issues do not reveal themselves until silicon arrives from the foundry.

RF performance characteristics, analogue behaviours, package interactions and thermal effects frequently require extensive post-silicon investigation.

“The complexities involved in chip development do not always appear during the pre-silicon stage,” says Vedantham. “They often emerge after silicon arrives.”

As a result, bottlenecks in validation and qualification increasingly determine time-to-market performance.

Inside HCLTech’s advanced semiconductor lab

To address these challenges, HCLTech has invested heavily in post-silicon infrastructure through its Advanced Semiconductor Lab (ASL).

The facility represents part of a broader semiconductor infrastructure investment programme exceeding $60 million, including laboratories focused on semiconductor engineering, qualification and equipment.

At the centre sits the ASL facility itself, a 40,000-square-foot engineering environment featuring approximately 25,000 square feet of Class 1K and Class 10K cleanroom space.

Vedantham describes the laboratory as one of the company’s most significant strategic investments.

“It is a critical cog in the entire silicon development flow,” he says.

The facility supports the complete post-silicon lifecycle, including functional validation, performance validation, automated test engineering (ATE), characterisation, reliability testing, qualification and failure analysis.

Capabilities extend to board design and manufacturing, allowing programmes to progress from first silicon through to high-volume production readiness within a single environment.

Equipment includes RF and electrical characterisation platforms, advanced test systems from suppliers such as Advantest and Teradyne, electron microscopy tools, failure-analysis systems and focused ion beam (FIB) technology.

An SMT production line further enables rapid prototyping and manufacturing support.

Accelerating time to market

The strategic value of integrating these capabilities lies in eliminating traditional hand-off delays.

In many development programmes, pre-silicon and post-silicon activities occur across different organisations, facilities and geographical regions. Knowledge transfer challenges frequently create delays and increase risk.

HCLTech’s integrated approach aims to remove those barriers.

By combining design expertise with post-silicon validation resources, engineering teams maintain continuity throughout the development process.

This creates opportunities to identify critical risks earlier, prioritise validation efforts more effectively and accelerate root-cause analysis when issues arise.

Perhaps most importantly, advanced failure-analysis tools enable rapid investigation and correction of silicon issues.

“When anomalies occur, we can identify them, validate potential fixes and proceed with greater confidence,” says Vedantham.

The result is shorter validation cycles, improved yields and faster progression towards high-volume manufacturing.

As semiconductor complexity continues to increase, such capabilities are likely to become increasingly important.

The heart of the system

For Vedantham, the future of semiconductor innovation extends far beyond individual products or technologies.

He believes the industry is undergoing a broader transformation in how semiconductors are perceived and valued.

“We are entering uncharted territory,” he says. “The semiconductor is no longer just a component in the system. It is the heart of the system.”

As AI continues to reshape computing, as vehicles become increasingly software-defined, and as industries embrace intelligent automation, custom silicon will play a central role in enabling the next generation of innovation.

The opportunities are enormous, but so too are the challenges. Success will require deeper collaboration across the semiconductor ecosystem, greater investment in design and validation infrastructure, and new approaches to engineering partnerships.

For companies willing to embrace that shift, the rewards may be substantial. As the trillion-dollar semiconductor era approaches, custom silicon is rapidly moving from competitive advantage to competitive necessity.