Silicon Resurgence: The Real Growth Opportunity

While the decade leading up to 2020 seemed to indicate a slowdown of semiconductor innovation with the alleged “death” of Moore’s Law, the semiconductor industry has roared back as a center of innovation and is set up for an era of unprecedented growth. The primary driver of this growth is the exponentially increasing compute intensity of artificial intelligence (AI, in particular, Generative AI) workloads across cloud and enterprise datacenters, autonomous vehicles, robotics, as well as financial, healthcare, industrial, and consumer segments. The era of scaling out compute with general purpose silicon is now behind us as we transition to a new era of accelerated computing with energy-efficient domain specific architectures that can meet the compute intensity of AI workloads. This new computing era mandates cost and power efficiency through optimizations across hardware and software stacks, and co-design of software and hardware. The semiconductor industry is embracing this challenge head on, and spawning a new generation of exciting semiconductor technology companies.

The Focus on Silicon

While venture capital (VC) firms often cast a wide net, focusing on a broad spectrum of industries, Maverick Silicon takes a more targeted approach by honing in on the semiconductor sector, recognizing the criticality of its role in the build out of the AI infrastructure that will power our future. This strategic focus allows us to dive deep into one of the most critical industries of our time, providing an opportunity to identify transformative opportunities early. Semiconductor technology has a long development timeline; with economics and supply chain that are vastly different from other technology segments, and an ecosystem that is highly intertwined. Deep understanding of the intricacies is critical to recognizing true potential and driving attractive return on investment (ROI). Maverick Silicon’s expertise in this space provides us with the insight needed to identify companies that go well beyond surface-level trends, and partnering with companies that are poised for delivering exceptional value to customers while driving explosive growth.

Breaking down Semiconductor Growth

The rise of accelerated computing with specialized domain specific architectures is first disrupting the cloud datacenter infrastructure. Nvidia, with its investments in high-end GPUs and CUDA, is driving the massive shift towards accelerated computing, which in turn is driving the growth for interconnects, switches, memory, storage, power delivery, cooling, etc., effectively resulting in a complete transformation of the cloud datacenter. The hyperscalers have responded with their own investments in custom ASICs (termed XPUs), for AI training and inference, further accelerating this transformation. Google TPU, AWS Trainium and Inferentia, Meta MTIA, and Microsoft MAIA are all examples of such custom ASICs. We expect continued strong growth in both GPUs and XPUs, and the connectivity/memory/storage/power/cooling infrastructure around these compute elements, as cloud operators continue to invest in building out their AI training and inference infrastructure and services in coming years.

The next frontier for specialized silicon, beyond cloud datacenters, is the Edge infrastructure – that spans across the cloud/carrier edge and device IOT (internet of things) edge within broader segments such as automotive, industrial (Smart cities/factories, etc.), consumer, healthcare, and defense. -As AI models, including large language models (LLMs), get more efficient over time, deploying technologies like MOE (Mixture of Experts), MLA (Multi-Head Latent Attention) and Reinforcement Learning (as demonstrated recently by DeepSeek), we expect the adoption of AI to accelerate across this Edge infrastructure, underpinning the growth of specialized silicon that is purpose built for a multitude of Edge use cases such as autonomous vehicles, robotics, etc.

The exponentially increasing compute intensity across Cloud and Edge infrastructure, in the post Moore’s law era, is also necessitating the shift from large monolithic die to a system of chips (SoC) paradigm, where multiple smaller dies (chiplets) are interconnected within a package substrate to deliver better overall costs. This shift is another key driver of growth within the semiconductor industry, as innovative packaging solutions, interconnect standards, IP/chiplet marketplace, platform architecture and performance modelling solutions, and hardware/software co-design solutions will all be required to enable future chiplet based SoCs.

Finally, geo-political considerations driven global supply-chain shifts, push for semiconductor independence, sovereign AI and compute infrastructure are expected to further act as growth catalysts for the worldwide semiconductor industry, which has been projected to cross $1T by 2030 (Source: PWC State of the Semiconductor Industry, Nov 2024).

Why Baya Systems?

The silicon resurgence outlined above, and the unprecedented growth of semiconductors that is projected going forward has been attracting significant private capital in recent years, after almost 15 years of scarce private capital for semiconductor start-ups. A large majority of this recent private capital is flowing into AI silicon and infrastructure focused start-ups. One critical shared architectural challenge for many of these AI silicon start-ups is that they need cost and energy efficient data movement (for both monolithic and multi-die chiplet based platforms) – as AI processing involves significant data movement within the platforms and hence is a primary cost and power driver of the overall solution. For chiplet based platforms, an additional critical requirement is the need to do upfront architectural what-if analysis and performance modelling when integrating chiplets from multiple vendors, and this has emerged as a key headwind for the industry transition to chiplet based platforms.

At Maverick Silicon, we are laser-focused on identifying breakthrough technologies that have the potential to disrupt and redefine industries. Baya Systems stands out to us for several reasons. First, their breakthrough compilable network-on-chip technology can deliver hyper-efficient data-movement solutions thus addressing a key bottleneck in high-performance computing solutions such as GPUs, XPUs, CPUs, switches, etc. Additionally, their software-driven design approach for single-die and multi-chiplet solutions, allowing upfront architectural tradeoff analysis and performance modelling, can substantially simplify architecting scalable solutions while radically reducing execution risk and time to market. Finally, while there are competing vendors that have fulfilled the market needs so far, the AI era going forward is demanding more stringent KPIs – across wire efficiency, bandwidth, latency, correct-by-construction, future-proof-design, and fast design closure – that need a paradigm shift in how interconnect fabrics get architected and implemented. And we believe that Baya technology is unmatched by competition in this regard, giving them a strong competitive advantage.

But we fundamentally believe that technology differentiation is not enough. We place equal emphasis on the team behind the innovation. A company may have a great product, but without a strong, visionary leadership team capable of executing at scale, success remains uncertain. That’s why we prioritize a combination of deep technical expertise, a bold vision, and a track record of successful execution when evaluating opportunities. This is where Baya Systems shines. Their founding team of successful serial entrepreneurs, and dynamic industry veterans combine the startup mindset with a strong focus on customer orientation, delivery and quality. With a combination of cutting-edge technology, a clear market gap to fill, and a world-class team with a track record of delivery, Baya Systems is well-positioned to thrive in the semiconductor industry, and we’re proud to support their journey.

Written by Manish Muthal, Senior Managing Director at Maverick Silicon.

The views expressed herein are solely the views of the author(s) and are not necessarily the views of Maverick Capital, Ltd. or any of its affiliates. They are not intended to provide, and should not be relied upon for, investment advice.