Samsung and SK Hynix are as vital to AI as NVIDIA

The shift to hardware infrastructure

The U.S. stock market is currently experiencing a tug-of-war between optimism for artificial intelligence and concerns over excessive investment costs. While some companies have seen stock price plunges due to high spending, venture capital firms are doubling down on the physical components of the AI era. Andreessen Horowitz (a16z) recently formed a $1.1 billion Machine Age Fund to target these specific hardware needs.

The fund focuses on the foundational layers of the AI ecosystem, including semiconductors, memory, networks, and data centers. It also looks toward robotics as a primary area for growth. This strategy acknowledges that while model developers are currently struggling with a shortage of computational resources, the next wave of growth will come from technologies that solve these resource constraints.

Raghu Raghuram, the fund's co-head, explained that the fund does not pre-allocate money to specific sectors. Instead, it follows the bottlenecks of the industry. If data volume grows, the fund will invest in memory; if networks become the limiting factor, it will pivot toward data supply technologies.

AI as the next electricity

Raghu Raghuram compares the ripple effects of artificial intelligence to the historical impact of electricity. He argues that while the cloud transformed enterprise IT, AI is poised to drive changes across nearly every industry that contributes to the gross domestic product (GDP). He believes that many people have not yet benefited from AI because most of the demand has yet to emerge.

The comparison to electricity is rooted in how technology moves from industrial use to everyday life. Just as electric motors eventually moved from factories into home appliances for cooking and cleaning, AI is expected to find unforeseen applications in various consumer and industrial settings. The fund's name, Machine Age, reflects this belief that AI will reshape human and machine interaction similarly to how steam engines and the internet transformed civilization.

The goal of these investments is to improve computing performance and power efficiency. Raghuram notes that simply expanding existing factories is not enough to satisfy the growing demand. The industry requires breakthroughs that can handle the massive scale of AI development and operations while remaining efficient enough to be commercially viable.

The importance of South Korean memory

During a recent visit to South Korea, Raghuram highlighted the country's dominance in memory and manufacturing as a cornerstone of the AI supply chain. He stated that Samsung Electronics and SK Hynix are globally important companies, as crucial as NVIDIA, because memory has become vital in AI. This underscores a shift in perspective where memory is no longer a secondary component but a primary driver of performance.

The growth of SK Hynix, specifically driven by high-bandwidth memory (HBM), serves as a primary example of this trend. Raghuram argues that the long-term value created by these companies is enormous despite daily stock price fluctuations. Because Western companies using NVIDIA chips require reliable memory suppliers, the ties between South Korean manufacturers and Western supply chains are becoming increasingly significant.

The concentration of technical talent and large-scale production in South Korea makes it a hub for next-generation breakthroughs. Raghuram believes that because of this industrial foundation, Korean startups are well-positioned to pioneer new technologies. He is specifically looking for opportunities in both next-generation memory and groundbreaking robotics for the electronics manufacturing sector.

Criteria for hardware breakthroughs

When evaluating potential investments, Raghuram emphasizes the capabilities of the founder over the specific technology area. He looks for leaders who possess deep technical expertise and the ability to navigate complex global supply chains. A critical requirement for these founders is a concrete plan to secure their first customers in a crowded and competitive market.

The fund maintains high standards for what constitutes a meaningful technological advancement. Raghuram noted that improving by 10% to 20% over existing solutions is not enough to convince customers to switch their systems. To be successful, a new technology must offer a significant edge, specifically a technological advantage of 3x, 5x, or 10x.

This high bar for innovation is intended to filter out incremental updates and focus on true breakthroughs. By seeking out these massive leaps in performance, the fund aims to identify the technologies that will actually solve the current computational shortages. This approach ensures that the investments contribute to the actual productivity gains that the AI industry promises.