Capital is moving into AI at a scale that now rivals entire industry cycles. Data centers, GPUs, and energy systems are being built ahead of long-term visibility on demand. This raises a more complex question for founders and investors. If AI is already delivering real value, where does the next layer of risk emerge, and how should ecosystems like Korea position themselves within it?
AI Demand Is No Longer the Question
Artificial intelligence has moved beyond experimental deployment. It is now driving measurable economic activity, infrastructure expansion, and capital allocation at a scale rarely seen in modern technology cycles.
Global investment signals reflect this shift. OpenAI announced its Stargate initiative with plans to deploy up to USD 500 billion in AI infrastructure over four years. Meanwhile, Microsoft has indicated it will invest around USD 80 billion in AI-enabled data centers in a single fiscal year. Market estimates cited by Reuters suggest that major technology firms could collectively spend more than USD 630 billion on AI infrastructure in 2026 alone.
At the same time, demand continues to outpace supply. Semiconductor leaders such as TSMC and ASML have maintained strong outlooks tied to AI workloads. The International Energy Agency projects global data center electricity demand could approach 950 terawatt-hours by 2030, with AI as the primary driver.
So, unlike the earlier cycles where expectations outpaced capability, this time, the technology is working, and it is already embedded in real economic systems.
A Pattern from Previous AI Cycles Still Holds
Despite this progress, one dynamic has remained consistent across decades.
Steve Shwartz, an angel investor, award-winning author, and former AI operator across multiple generations of technology, points to a recurring pattern. Early AI systems in the 1970s and 1980s generated technical breakthroughs but failed to deliver sustained commercial impact. Hype outpaced execution, and the market eventually reset.
Today’s cycle is different in one critical way. According to Shwartz, modern AI has already produced “transformative economic and industrial outcomes,” driven by milestones such as deep learning breakthroughs and large language models.
However, he also highlights a familiar behavioral risk. In correspondence with KoreaTechDesk, Shwartz noted,
“During high-momentum periods, founders and investors alike often fall prey to overconfidence.”
That overconfidence, he explains, can distort expectations and lead to inflated valuations, even when the underlying technology is valid.
This shows clear implication that even when AI works, the surrounding system can still misprice risk.
Where AI Is Actually Creating Value Today
The current AI market is not evenly distributed. Value is concentrating only on specific layers of the stack.
Shwartz describes a clear structure. Companies that control key infrastructure generate durable value. This includes GPU providers, foundational model developers, infrastructure software platforms, cloud service operators, and a growing set of AI-native application companies.
This concentration aligns with observable market behavior. Companies supplying compute capacity and cloud infrastructure are capturing a disproportionate share of revenue, while many application-layer startups are still proving their business models.
And so, this creates a structural imbalance that shapes where capital flows, and more importantly, where risk accumulates.
The Emerging Risk: Infrastructure May Outrun Demand
The central tension in the current cycle is no longer technological uncertainty. It is capital allocation.
Massive investments are being directed toward data centers, compute infrastructure, and energy supply. These investments are justified by strong current demand. However, the scale and speed of expansion introduce a new question: will future demand truly match the capacity being built today?
Shwartz frames this as a structural risk rather than a technological one.
“There is no chance of the type of AI winter we had at the end of the 1980s.
AI is working and it is having a major impact on society.”
But he draws a parallel to a different historical moment. During the dot-com era, companies such as Global Crossing built extensive fiber infrastructure ahead of actual demand. When demand failed to materialize at the expected pace, the result was financial collapse across multiple players.
Indeed, the current situation may not be entirely identical because demand for AI services is already strong, and in many cases exceeds available supply. However, the scale of planned data center expansion introduces the possibility of overbuilding.
The IEA report also suggested that large-scale data center projects are already facing constraints related to power availability, land access, and local regulatory resistance. In some cases, projects have been delayed or canceled due to these factors. At the same time, developers are increasingly planning dedicated power generation to support future capacity.
The risk, then, is not that AI demand disappears. It is that infrastructure supply may eventually exceed economically sustainable utilization levels.
Korea Is Entering the Same Infrastructure Race
At the same time, South Korea is not just observing this trend from a distance. The country has actually been actively participating.
The government has outlined an “AI Computing Infrastructure Development Strategy” and is advancing plans for a National AI Computing Center Korea. Public announcements indicate plans to secure tens of thousands of advanced GPUs, alongside expanded tax incentives for AI-related infrastructure.
These initiatives reflect Korea’s strategic objective to strengthen its position in the global AI ecosystem by building domestic computing capacity and reducing reliance on external platforms. This strategy is known as Korea’s sovereign AI vision.
At the same time, structural constraints remain. Policy research from Korea Information Society Development Institute highlights challenges related to power supply, grid capacity, land availability, and cooling infrastructure. These are not minor issues. They directly determine how quickly and efficiently data centers can be deployed.
Historically, such constraints have also influenced where hyperscale data centers are located across Asia. Korea has been working to improve these conditions, but the system-level complexity remains still.
The Real Question: Catching Up or Building Ahead
Now, South Korea has been sitting between two competing risks.
On one side, insufficient infrastructure could limit competitiveness. AI development increasingly depends on access to large-scale compute resources, and delays in building capacity could weaken domestic innovation and enterprise adoption.
On the other side, aggressive expansion introduces capital risk. If infrastructure is built faster than long-term demand justifies, returns may compress, and resources may be misallocated.
While this may look like a binary outcome, it actually shows a timing issue.
Shwartz’s broader observation about overconfidence shows a clear reflection about this as well. When demand is strong and capital is abundant, systems tend to expand quickly. That is why the challenge now is maintaining the alignment between capacity, utilization, and economic value over time.
What This Means for Global Founders and Investors
Now, for global participants in the startup ecosystem, the implications extend beyond Korea.
AI remains one of the most significant growth areas in technology. However, the structure of the market is becoming increasingly capital-intensive. Infrastructure ownership, energy access, and scale are becoming decisive factors.
For investors, this shifts the focus. Evaluating AI opportunities now requires assessing not only product capability, but also exposure to infrastructure cycles and capital efficiency.
As for founders, the landscape is equally complex. Competing directly in infrastructure layers requires significant resources, while building on top of existing platforms introduces dependency on external providers.
Korea’s trajectory illustrates this tension in real time. It is investing to secure long-term positioning, while navigating the same structural constraints and risks seen globally.
Conclusion: A Different Kind of Risk in the AI Era
In the end, the uncertainty phase has now passed, and AI is no longer facing a collapse in capability.
Today, the next challenge lies in how the ecosystem builds around it. Capital is flowing into infrastructure at an unprecedented scale. That expansion is necessary, but it also introduces the possibility of imbalance.
Hence, the outcome will depend on how closely supply aligns with future demand, and how effectively systems manage physical, financial, and regulatory constraints.
And for Korea and the broader global ecosystem, the question is no longer whether AI will scale. It is how precisely that scaling is executed.
Key Takeaways
- AI demand is real and growing, supported by large-scale investments from companies like OpenAI and Microsoft
- Global AI infrastructure investment could exceed $630 billion annually, increasing capital allocation risk
- Steve Shwartz highlights that overconfidence cycles persist even when technology is viable
- The primary risk has shifted from technological failure to infrastructure overbuild and capital misalignment
- Korea is accelerating AI infrastructure through GPU expansion and the National AI Computing Center
- Structural constraints in Korea include power, land, and grid capacity, as noted by Korea Information Society Development Institute
- The key strategic challenge is balancing infrastructure expansion with long-term demand and economic sustainability
- For investors and founders, AI opportunity now depends on capital efficiency and positioning within the infrastructure stack
🤝 Looking to connect with verified Korean companies building globally?
Explore curated company profiles and request direct introductions through beSUCCESS Connect.
– Stay Ahead in Korea’s Startup Scene –
Get real-time insights, funding updates, and policy shifts shaping Korea’s innovation ecosystem.
➡️ Follow KoreaTechDesk on LinkedIn, X (Twitter), Threads, Bluesky, Telegram, Facebook, and WhatsApp Channel.