AI Boom Strains Even Google, Which Turns Down Meta's Request

*Editor's Note: 'AI PRISM' (Personalized Report & Insight Summarizing Media) is an "AI-based customized news recommendation and summary service" developed with support from the Korea Press Foundation. It selects and provides six customized news items for each reader type.

1. Google Hits AI Infrastructure Bottleneck Too: Will Inability to Meet Contracted Capacity Hold Back Growth?

- Key Summary: Google informed Meta, a major customer, in March this year that it could not provide as much of its AI model "Gemini" capacity as requested, the Financial Times (FT) reported on the 28th. Alphabet's unfulfilled backlog stands at $460 billion, nearly double the previous quarter, and the company cannot meet demand that reaches 23 times the capacity it can handle. As a result, Meta has not even been able to set a launch schedule for the developer tools of its own AI model "Muse Spark," and is advising employees to use tokens, the unit of AI usage, more efficiently. Meanwhile, the BIS forecast in its annual report that AI optimism could lead to a prolonged investment downturn, shaking financial markets and dealing a blow to the global economy.

2. Government to Embed AI in Manufacturing Including Semiconductors: "Creating 100 Trillion Won in Added Value"

- Key Summary: The National AI Strategy Committee, the Ministry of Trade, Industry and Energy, the Ministry of Science and ICT, and the Ministry of SMEs and Startups announced "The Path to Korea's Manufacturing Transformation: Manufacturing AI 2030 Strategy" on the 29th, deciding to invest 20 trillion won in joint public-private funding through 2030. The three core tasks are building a "national manufacturing data library" that links manufacturing data scattered across ministries, developing manufacturing-specialized AI models, and spreading regional manufacturing AI. In addition, the government will prioritize allocating 48 billion won in next year's supplementary budget to convert into data even the tacit knowledge (manufacturing know-how) of master manufacturers nearing retirement. The plan also includes fostering "full-stack AI factories" that condense manufacturing AI capabilities as export products, entering advanced countries and the Middle East on a turnkey basis and the Association of Southeast Asian Nations (ASEAN) and Latin America through platform construction and consulting.

3. AI Growth Engine Urgently Needed, but Kakao Tripped Up by Performance Pay

- Key Summary: The Kakao union staged a "logout day" on the 29th, escalating its struggle following the breakdown of wage agreement negotiations, with unions of five affiliates including Kakao Pay and Kakao Enterprise joining in. Kakao designated this year as the first year of company-wide "AI innovation," presenting goals of transforming into a KakaoTalk-based "agentic AI (AI that sets and executes its own goals)" platform, unveiling its large language model (LLM) "Kanana 2.5," and strengthening "AI orchestration" collaboration with global Big Tech. However, with internal turmoil continuing, including the resignation of former Chief Product Officer (CPO) Hong Min-taek, core tasks are losing momentum. Meanwhile, Kakao's stock, which traded in the 60,000 won range early this year, recently fell to the 30,000 won range, and the company said it plans to concentrate funds secured through measures such as the sale of its Dunamu stake on investment in AI growth engines.

4. 600 AI Apps Built by Employees: SKB Accelerates Company-Wide AX

- Key Summary: SK Broadband built its in-house AI development platform "Playground" in February this year, cutting the development environment setup time from more than two months to five minutes and creating an environment where any member can directly develop AI agents. Through this, about 600 AI applications are being developed and operated, and some 30 AI agents that judge and act on their own have been applied in the field. Notably, the AI monitoring and diagnostic agent "C-One" automatically detects abnormal signs in wired networks, immediately identifies causes and inspection priorities, and even suggests immediate measures such as remote equipment resets or on-site inspections. "As members create AI agents themselves and apply them in the field, the way we work is fundamentally changing," said Sung Jin-soo, head of SK Broadband's Network Center.

5. Former Game Developers Join Forces: Pet Commerce Turns Profitable on the Power of Data

- Key Summary: EFIL, a startup founded in 2018 by four developers from Netmarble (251270), operates customized feed recommendation, health information, and pet food commerce businesses centered on its pet care platform "Meongnyang Bogam," recording revenue of 8.3 billion won and operating profit of 1.4 billion won last year. EFIL's core competitiveness is its high-density consultation data, having secured 12,000 meaningful consultation records last year containing pet conditions and owner concerns. It also aims for a structure that achieves more than 100 billion won in revenue with a workforce of around 25 people, by automating much of the product sourcing, recommendation, and consultation process. "Building proprietary data assets that commercial AI services do not have is our survival strategy," said Lee Kyung-jun, CEO of EFIL.

6. AI Self-Driving Robots Deployed: Aircraft Inspection Cut from 12 Hours to 1 Hour

- Key Summary: The Ministry of Trade, Industry and Energy held the "3rd Industrial Convergence Regulatory Special Exemption Review Committee" on the 29th and granted a regulatory special exemption allowing AI self-driving robots to enter the aprons of Incheon and Gimpo airports. As a result, self-driving robots can photograph the underside of aircraft and use AI to identify defects and damage, and aircraft inspection time, which took up to 12 hours when performed by people, is expected to be cut to within one hour. Regulatory easing was also carried out in the energy sector, with a demonstration special exemption taking effect that allows power consumers to deduct electricity supplied from a shared energy storage system (ESS) from their total power usage. "We will continue to rationalize regulations so that the public can feel the difference in everyday life," a ministry official said.