Mirae Asset chairman: AI progress at 10–20%, peak uncertain

By Rebecca Stone

Mirae Asset Group chairman Park Hyeon-joo put the AI revolution at perhaps 10–20% of its development on the 27th, arguing that AI growth can keep semiconductor demand rising despite peak concerns.

Park made the assessment at the Kyung Hee University Graduate School of Business’s 60th-anniversary insight forum. “The AI revolution, in my view, has perhaps progressed 10–20% even if it has come a long way,” he said. He framed the figure as his personal estimate; he attached no shipment, capacity or pricing numbers.

How far has AI actually progressed?

Park did not dismiss the semiconductor peak theory. He said it could prove right or wrong. His case was conditional: continued AI expansion could raise semiconductor demand, while the ability to produce memory and high-bandwidth memory, or HBM, stays concentrated among a limited number of companies.

His logic linked two conditions rather than dismissing the peak argument. AI adoption would have to keep expanding, and the supplier base for memory and HBM would have to remain narrow, for that demand case to hold. He offered no separate forecast for AI growth rates, memory pricing or HBM output.

Park linked that concentration to the strategic position of Korean suppliers, asking whether NVIDIA could operate without Samsung Electronics’ memory and SK Hynix’s HBM. He used the question to underline their commercial importance. It was not a disclosed procurement claim, because his remarks supplied no NVIDIA contract, capacity commitment or shipment figure.

That distinction matters for the peak debate. Park did not cite a fab investment, wafer capacity, pricing target or shipment number. His case rested on demand: further AI expansion could increase semiconductor demand even while the ability to make memory and HBM stays concentrated. His 10–20% estimate described his view of remaining runway, not measured industry capacity.

The distinction keeps a chairman’s outlook separate from a capacity claim. His statement supports a view on potential demand; it does not establish how much Samsung Electronics, SK Hynix or other memory suppliers can ship.

Where does Korea fit?

Park extended the argument beyond semiconductors, saying Korea must secure a specific position in robotics. He acknowledged that Korea can lag the United States in international trade but argued that leadership should remain with Korea.

He did not identify the companies or robotics segments that should hold that position, and he offered no production or revenue target. The proposal extended his strategic argument without quantifying a new industrial bet.

Why challenge the strong-won policy?

Park challenged the assumption that a strong won should serve as a benchmark for national competitiveness. He called that mindset outdated and asked whether a strong currency was desirable for strengthening manufacturing. A strong won lowers import prices and appears to support GDP, he said, but those benefits must be weighed against price competitiveness in automobiles, steel and batteries.

He called for a new national strategy, including exchange-rate management, to maintain competitiveness in core industries such as semiconductors and robotics. It remained Park’s proposal; no government action was announced in his remarks. For exporters, Park linked currency policy directly to manufacturing competitiveness rather than treating GDP arithmetic as the sole measure.

For manufacturers, the relevant issue is whether currency policy supports the price competitiveness needed to defend overseas sales, rather than whether a stronger currency simply makes imports cheaper. Park placed exchange-rate management within industrial policy for semiconductors and robotics.

What is Mirae Asset preparing next?

Park also said Mirae Asset plans to expand overseas mergers and acquisitions. “Another M&A announcement for Mirae Asset will likely be made tomorrow,” he said. An announcement was expected; a closing was not.

The wording also matters. A promised announcement is not a completed acquisition, and it establishes neither where Mirae Asset will deploy capital nor what financial returns it expects.

Park named no targets and said specifics would emerge only after the board completed its procedures. The report gave no transaction value, country or sector, leaving scale and strategic effect open until the board process advances.

Two near-term tests remain: the promised overseas M&A announcement and whether AI-linked memory and HBM demand sustains the strategic position Park assigns to Samsung Electronics and SK Hynix despite a possible semiconductor peak.