SK Hynix wants Wall Street to pay a premium for the AI memory boom. The awkward part is that its Seoul stock has started arguing back.
Two days before its Nasdaq trading debut, SK Hynix is dealing with the kind of share-price swing that makes a clean $28 billion sale much harder to sell. The stock has fallen about 12% so far in July, according to The Wall Street Journal, after a run that had more than tripled its value this year. If you're a US investor being asked to buy the AI memory leader at a fresh Wall Street price, you now have to decide whether you're buying strength or arriving after the easy money has already been made.
The deal is still huge. SK Hynix plans to issue 17.8 million new common shares through American depositary receipts that are expected to begin trading on Nasdaq on July 10, with 10 ADRs representing one common share. Its revised filing priced the shares at 2,425,000 won each, equal to about 242,500 won per ADR, and cut the fundraising target to roughly $28.21 billion from the earlier 45.453 trillion won plan, according to the Journal. Bank of America, Citigroup, Goldman Sachs and JPMorgan are managing the offering.
There is demand. MarketWatch reported, citing a Monday SEC filing, that Baillie Gifford, Coatue Management and Situational Awareness, the hedge fund run by former OpenAI researcher Leopold Aschenbrenner, are set to invest up to $7 billion in the listing. That isn't casual money. It is exactly the kind of cornerstone support bankers want when they take a Korean chipmaker into the deepest US equity market.
What isn't clean is the signal coming from the stock.
Bloomberg tied part of the recent pressure to Meta Platforms, which has been exploring ways to resell excess AI computing capacity through a cloud service instead of only hoarding infrastructure for its own models. Investors read that as a possible crack in the story that AI server demand keeps moving up in a straight line. SK Hynix sits right in the middle of that trade because it supplies the high-bandwidth memory used with Nvidia's AI accelerators and controls roughly 55% to 60% of the global HBM market, depending on the estimate you use.
Frankly, that is why the selloff stings. SK Hynix isn't a weak side character in the AI buildout. It is one of the companies the whole thing depends on. When its shares drop before a US listing, the question isn't whether HBM matters. It does. The question is how much perfection investors have already priced into a company that has become a proxy for the entire AI infrastructure boom.
James Reilly, a senior markets economist at Capital Economics, gave the sharper version of that warning. Bloomberg quoted him as saying that moves of this scale have usually appeared around real bear markets, including the Asian financial crisis, the dot-com bust and the 2008 financial crisis. His conclusion was blunt: the volatility points to excessive froth and raises doubts about the durability of the rally.
Part of the move may also be mechanical rather than fundamental. Bloomberg reported that a roughly $13 billion leveraged ETF tied to SK Hynix has become an amplifier, forcing buying and selling as the underlying shares move. That kind of structure doesn't wait for a sober read on Meta's cloud plans, Nvidia demand or DRAM pricing. It just follows the tape, which is exactly how a large stock can start behaving like a smaller, jumpier one.
None of this changes why SK Hynix wants the money. The company has said the proceeds will go toward manufacturing capacity in South Korea and extreme ultraviolet lithography equipment from ASML for next-generation chips. Tom's Hardware reported last week that the plan includes the Yongin Semiconductor Cluster fab, advanced packaging work in Cheongju and EUV tools, the unglamorous equipment list behind the AI boom. You don't get more HBM by talking about demand. You build fabs, buy scanners and wait.
The offering would still rank among the largest share sales ever. The Journal placed it behind SpaceX's $75 billion offering and ahead of Saudi Aramco's $25.6 billion IPO in 2019 and Alibaba's 2014 listing. That comparison is useful because it shows the size of the bet, but it also shows the risk. Deals this large don't happen in a vacuum. They become market events.
For SK Hynix, the Wall Street pitch is supposed to be simple: give US institutions direct access to the memory company most closely tied to Nvidia, broaden the shareholder base, and narrow the valuation gap with US chip names such as Micron. Meritz Securities analyst Kim Sun-woo recently argued, according to the Journal, that the listing could also help SK Hynix enter the Philadelphia Semiconductor Index, which would pull in passive money from funds tracking US chip benchmarks.
That is a real prize. But you don't get it for free. If the reference price keeps sliding before the first Nasdaq trade, underwriters have to explain not just what SK Hynix is worth, but what kind of AI cycle investors are actually buying. A normal chip listing is about earnings, capacity and market share. This one is about whether the AI infrastructure trade can absorb its own success.
SK Hynix picked a strange week to test that. Its HBM lead is real, its ASML shopping list is real, and the cornerstone interest is real. So is the market's sudden nervousness. July 10 may settle the opening price, but it won't settle the larger question of whether investors still believe every new dollar of AI spending turns into another dollar of chip demand.
Also read: SpaceX becomes the fastest company ever added to the Nasdaq-100, US Investors Can Finally Buy Into SK Hynix Ahead of Its Nasdaq Debut, Samsung's Record Quarter Just Undercut the AI Spending Skeptics
