Vadim Shpak
For many venture investors, semiconductors were historically a difficult area to invest in. Development cycles are long, technical risk is high, manufacturing is capital-intensive and it can take years before a company reaches meaningful revenue.
AI is changing that equation.
Enourmous amount of capital going into AI is not flowing only into LLMs. Every new model and every increase in AI usage requires more computing power, memory, networking, advanced packaging, cooling and energy.
This makes the entire AI supply chain increasingly attractive for investors.
From a VC perspective, there is another advantage. You do not necessarily need to predict which LLM will ultimately win. Instead, you can invest in the infrastructure that many of them will need.
We can already see the impact on semiconductor valuations.
Positron, an AI inference chip company, is a good example. In February 2026, the company raised $230 million in a Series B at a valuation above $1 billion. Just five months later, it was reportedly discussing another $750 million financing, with a first tranche potentially valuing the company at $3.5 billion and a second at around $5 billion.
Whether those valuations are eventually achieved is less important than what the change tells us. Investor appetite for semiconductor companies positioned directly in the AI infrastructure build-out has increased dramatically.
But the opportunity goes much further than designing AI processors.
Singapore provides an illustration of how a country can position itself around this opportunity.
Singapore is unlikely to compete with the US or China in the premier league of LLMs. Building these models requires enormous amounts of capital, computing infrastructure and a concentration of the world's best AI talent.
“Today, one of the most attractive ways to invest in AI may not be another AI application or another LLM. It may be the technology underneath them.”
Instead, Singapore is playing to its existing strengths: semiconductors, advanced manufacturing, engineering capabilities and its position within Asian supply chains.
The government is putting substantial capital behind this strategy. Under RIE2030, Singapore is committing S$37 billion to research and innovation over five years. Semiconductors have been selected as the first national RIE Flagship, with S$800 million committed. Singapore is also developing a S$500 million national semiconductor R&D fabrication facility and has committed an additional S$1 billion to Startup SG Equity to catalyse private investment into Singapore-based startups.
This type of government support matters because semiconductor startups cannot always be built with venture capital alone. Governments can fund infrastructure, research and early technology development that would otherwise be difficult for a startup or VC fund to finance.
We are already seeing the results.
Silicon Box is perhaps the best example. Founded in Singapore in 2021, the advanced semiconductor packaging company has grown very quickly. By August 2026, its Singapore factory had shipped 500 million units - five times the number reported less than a year earlier.
The company is now aiming to increase production capacity tenfold during 2026 and expects to reach 1.5 billion units shipped in early Q4.
Capital has followed that growth. Silicon Box recently raised a further $156.5 million extension to its Series B, taking its valuation above $1.7 billion. It is also investing $300 million to expand its Singapore factory.
Silicon Box illustrates why the semiconductor opportunity created by AI extends far beyond GPUs. As AI systems become larger and more demanding, the industry needs better ways of combining chips while improving performance, power consumption and manufacturing economics. Advanced packaging and chipset architectures are becoming increasingly important parts of that solution.
For VCs enthusiasm still needs to be balanced.
Semiconductors remain capital-intensive businesses. Technology risk is substantial, customer qualification takes time and scaling manufacturing is very different from scaling software. Investors need to understand not only whether the technology works, but also find a credible path to manufacturing and enough capital to reach commercial scale.
There is also a larger risk.
Much of today's excitement around semiconductors is on expectations that AI infrastructure spending will continue to grow. Today that demand is very real. But semiconductor markets have always been cyclical.
If AI investment continues growing, companies supplying the infrastructure can benefit enormously. If today's AI boom eventually cools, the companies supplying that infrastructure will also be affected.
That does not make semiconductors less investable. It simply means investors need to understand what is driving the demand.
Today, one of the most attractive ways to invest in AI may not be another AI application or another LLM. It may be the technology underneath them.
