The AI industry is growing rapidly as related markets are activated by the advancement of AI technology. The AI trend for 2025, which is about to begin, is definitely 'AI agent'. Experts are paying attention to the prospects of the AI industry scattered around, such as edge computing and power infrastructure, as customized AI industry-specific solutions are expected to improve productivity in various industries.

■ Active launch of industry-tailored AI tools
Deepel CEO Yarek Kutilovsky sees customized solutions as the future of AI models. His outlook is that customized AI solutions tailored to each field will dominate the market by solving tasks and providing real ROI.

Currently, the industry is using AI tools to improve work efficiency. Intel is strengthening the ISV software ecosystem by linking over 300 AI functions with AI PCs. Through the AI image creation and correction functions of Adobe products such as Photoshop and Premiere Pro, various designers and editors are utilizing AI functions in practical work areas, and AI translation, AI assistant, and document summarization are being utilized in work.

Recently, global knowledge information content company Thomson Reuters fired a signal flare for its full-fledged entry into Korea by launching 'CoCounsel', an AI assistant for legal professionals. CoCounsel is the latest LLM-based AI legal secretary service that supports lawyers in the legal market to utilize AI technology.

As AI technology launches in specialized fields become more active, adoption by companies is also on the rise. Cloudflare CSO Grant Bougica even made a somewhat pessimistic prediction, saying, “In the next 5 to 10 years, only two types of companies will remain: those that innovate using AI and those that do not and disappear.”

“AI for AI’s sake will gradually disappear, and the question of whether a solution solves a real problem will become more important,” said David Parry-Jones, CRO of DeepEL. “This is where specialized models for specific domains will benefit the most.”

■ Edge VS Server, On-Device VS Data Center

With the development of the AI market, 'on-device AI' has emerged as a hot topic. On-device, which was a major keyword in 2024, is expected to be a major challenge in AI development this year as well. With experts having differing views on edge versus server AI, adoption of compute locations tailored to each company’s target applications will need to be a key consideration.

John Engates, CTO of Cloudflare, emphasized that “to realize the true potential of AI, we need to move computing power closer to where it’s actually needed through edge computing.” Citing the instantaneous decision-making of self-driving cars, low-latency interactive games, and real-time image processing technology, he pointed out that “it depends on distributed computing to place intelligence where it’s most effective.”

On the other hand, Stephan Mesken, VP of DeepL Research, asserts that edge computing is not yet the future of language AI. He said, “The language AI market will grow along with data center infrastructure,” emphasizing that data center computing will inevitably take the lead due to its massive performance compared to limited local resources.

“Despite the high possibility of a shift to edge computing in the future, I believe that language AI models will continue to rely on supercomputers and data centers for the present and for the foreseeable future,” he said, adding that “AI accelerator hardware that can only be implemented in data centers is needed.”

■ AI Data Centers 40% Power Shortage, Power Supply and Infrastructure Companies’ Values Jump
According to market research firm Gartner, data center power consumption is expected to increase by 160% over the next two years due to the rapid increase in power demand caused by AI and generative AI. As a result, it is expected that 40% of existing AI data centers will have power availability issues by 2027.

Data centers are expected to need 500 terawatt-hours (TWh) of power per year to run AI servers in 2027, up 2.6 times from 2023. The increase in electricity demand from data center operators is expected to lead to higher electricity costs, which will act as a burden on potential service supply.

Bob Johnson, VP analyst at Gartner, predicted that “power shortages will persist as it can take years to build new infrastructure such as transmission and distribution and secure power generation capacity,” and “Increased data center power usage could lead to a short-term increase in carbon emissions,” suggesting the introduction of new clean energy technologies such as small nuclear reactors and review of alternative options such as edge computing and small-scale language models.

As the power shortage is expected due to the expansion of the AI market, the value of power infrastructure companies and energy production companies is also rising. The company with the highest growth rate in the domestic stock market in 2024 is HD Hyundai Electric, which rose 376% last year, followed by Gaon Cable at 165% and Hyosung Heavy Industries at 158%.

Constellation Energy, the largest nuclear power generation company in the U.S., has also seen its stock price rise 106% from its low point over the past year as of today. As the U.S. expands its AI infrastructure, the benefits for power infrastructure companies are spreading domestically and internationally. Attention is being paid to whether the expansion of the AI server market will continue globally in the future.