My work places me simultaneously in three domains being reshaped by technology: AI, semiconductor supply chains, and the circular economy.

▶ Listen to summary
AI-synthesized voice, cloned from the author's own voice

This position has one advantage: I can see technology’s acceleration, yet I touch its physical limits every day. When you’re running a debate engine on four AI models with one hand while handling utility bills for a metal recycling production line with the other, your feelings about the phrase “technology will change everything” become more complicated.

So my view of 2030 starts not from technological potential, but from constraints. Behind every beautiful-looking technology curve lies a physical, economic, or ethical ceiling. Only by seeing the ceiling clearly can you know where the real opportunities are.

AGI: Deepening Will Come, Generality Won’t Yet

AGI has been the hottest narrative of recent years. But if you actually use AI to get things done—not for play, but for daily decision-making—you’ll know we’re still far from general intelligence.

In The US AI Three-Year Countdown, I analyzed how the current bottleneck of AI isn’t just algorithms, but also the physical limits of computing resources (energy consumption, heat dissipation) and the ceiling of high-quality training data. The AI of 2030 is more likely to manifest as “deepening within specialized domains”—medical diagnosis, materials science, manufacturing optimization—rather than a comprehensive AGI breakthrough.

But the impact of “deepening” itself is already significant enough. When AI reaches a level surpassing human experts in specific domains—as I discussed in AI Understands Human Nuance Better Than People, this is already happening in the realm of social intelligence—the cost structures and talent demands of industries will be thoroughly rewritten. Human-machine collaboration isn’t an option, it’s an inevitability.

Quantum Computing: The Threat Arrives Before the Application

Quantum computing holds enormous potential in cryptography and materials science, and it is also a key geopolitical variable. But it faces unstable quantum coherence and extremely high error-correction costs. By 2030, hybrid systems will still dominate, and general-purpose quantum computers will remain experimental.

Interestingly, the threat of quantum computing runs faster than its application. Nations are already preparing for the possibility that “quantum supremacy” could break existing encryption systems, which directly affects the strategic positioning of the semiconductor industry. In Taiwan Semiconductor’s Tenfold Leap, I discussed how Taiwan’s next round of competition lies not only in process technology but in system integration capability. The quantum computing timeline will accelerate this shift.

Climate Resilience: Not a Slogan, but a Cost Structure

Doing metal recycling at the company, climate resilience isn’t an abstract concept to me—it’s the monthly utility bill.

The development of renewable energy is constrained by materials science bottlenecks and the long cycles of infrastructure renewal. But the more fundamental issue is this: climate risk isn’t a single event, but a trigger for compound crises—when extreme weather simultaneously strikes supply chains, food systems, and population movement, no single-point solution is enough.

This is precisely why I keep saying “the circular economy isn’t idealism.” When the costs of water, electricity, and chemicals all rise, the efficiency of resource reallocation directly determines your competitiveness. This is especially true for the semiconductor industry—if recycling efficiency improves by 5%, the impact on the profit structure far exceeds cutting personnel costs by 1%.

The Silent Restructuring of Social Structure

Technological change draws all the attention, but changes in social structure are often more far-reaching.

Aging, the gig economy, and digital transformation are happening simultaneously. The education system is still teaching skills from twenty years ago, while the labor market already demands entirely different capabilities. In The Canary in the AI Employment Coal Mine, I discussed how what gets eliminated isn’t specific jobs, but the entire logic of division of labor. UBI (Universal Basic Income) has moved from an academic concept to a policy experiment—not out of idealism, but because the old social contract can no longer hold.

Don’t Predict, Build Resilience

My most honest attitude toward 2030 is this: I don’t know. No one knows.

Behind every trend that appears certain lies a variable that could be overturned by a black swan. AGI might arrive five years early because of an algorithmic breakthrough, or it might stall for ten years because energy costs soar.

So the strategy I choose isn’t precise prediction, but building resilience that can survive across multiple scenarios. Deepening human-machine collaboration capability in the AI domain. Strengthening system integration rather than merely defending process technology in the semiconductor domain. Making resource reallocation a core competitiveness rather than just an ESG label in the circular economy domain.

True foresight isn’t predicting the future. It’s understanding which variables you can influence, then concentrating your resources there.