In this video, Eli, the computer guy, discusses the ongoing and worsening shortage of DRAM (Dynamic Random-Access Memory) that is expected to continue at least through 2028. He explains that while much attention has been given to AI’s demand for GPUs, the real hardware strain extends beyond that to include power supplies, storage, and especially RAM. SK Hynix, one of the world’s largest memory manufacturers, has warned that DRAM supply growth will lag behind demand for several years due to the massive requirements of AI servers. This shortage is not just a short-term issue but a prolonged imbalance affecting the mainstream PC market as well.

Eli highlights that SK Hynix’s internal analysis projects limited growth in commodity DRAM production, with the focus shifting heavily toward AI server needs rather than consumer electronics. The company and other memory manufacturers are adopting conservative capacity expansion strategies, prioritizing profitability over flooding the market with new supply. This cautious approach stems from the belief that the current AI-driven demand is a bubble that will eventually burst, making large investments in new factories risky. As a result, existing inventories are at historically low levels, intensifying allocation pressures and making it harder for consumers to access RAM.

The video also critiques the AI boom, describing it as fundamentally destructive and driven by inflated valuations and hype. Eli argues that while AI has some value and potential, the enormous capital expenditures—such as the $1.4 trillion planned by OpenAI’s Sam Altman—are disproportionate to the actual functional benefits most users experience. He compares the current AI hype to the infamous Clippy assistant from Microsoft Office, which was more of a nuisance than a help, suggesting that many users are overwhelmed by AI features they don’t want and struggle to disable. This analogy underscores his skepticism about the practical usefulness of much of today’s AI technology.

Eli further discusses the implications of the DRAM shortage for both consumers and businesses. With server demand for RAM expected to grow from 38% in 2025 to over 53% by 2030, the consumer market will face increasing scarcity. Some manufacturers have already sold out their RAM production slots for 2026, signaling tight supply for years to come. This situation may force companies and individuals to rethink hardware refresh cycles, potentially extending device lifespans to six or seven years instead of the usual three to five, due to difficulty in obtaining replacement parts like RAM.

Finally, Eli encourages viewers to consider the broader impact of these supply constraints on technology infrastructure, such as running local servers or databases that require significant memory. He invites discussion on the topic and promotes his Silicon Dojo educational platform, which offers hands-on technology classes focused on AI and programming. Despite the challenges posed by the DRAM shortage and AI hype, Eli remains committed to empowering learners with practical skills to navigate the evolving tech landscape.