Micron's $41.5 Billion Quarter and SK Hynix's $29 Billion US Debut: AI Memory Demand Is Not Peaking

There is a number that captures the scale of what happened in the semiconductor sector on Wednesday evening: 346. That is the percentage by which Micron Technology's revenue grew year over year in its fiscal third quarter - a figure so large it reads less like a financial result and more like a misprint. It was not. Micron reported $41.5 billion in quarterly revenue, up from $9.3 billion in the same period a year ago, and guided its fiscal fourth quarter to $50 billion. The stock surged roughly 16 percent in after-hours trading, snapping a fear-driven selloff that had sent shares down 13 percent just the day before. On the same day, South Korea's SK Hynix announced plans to raise $29.4 billion through a US Nasdaq listing - the second-largest equity offering in history behind SpaceX's record debut earlier this month. Taken together, the two events constitute the clearest signal yet that the AI memory trade is not peaking. It is accelerating.

The Numbers That Silenced the Bears

Heading into Wednesday's report, the mood around AI-exposed semiconductor stocks was genuinely anxious. Micron had fallen 13 percent on Tuesday as part of a broader tech selloff. Analysts polled by FactSet were modeling $35.75 billion in revenue and $20.83 in earnings per share - already elevated expectations by historical standards. Micron beat both by a wide margin. Revenue came in at $41.5 billion. Earnings per share hit $25.11. Gross margin reached 84.6 percent on a GAAP basis, a figure that would have seemed implausible for a memory company just two years ago, when the industry was cycling through one of its worst downturns and Micron's gross margin sat near 22 percent.

The profitability story is worth dwelling on. Memory chips have historically been among the most cyclical businesses in technology - prone to violent swings between oversupply and shortage, with margins that compress and expand in ways that make long-term modeling nearly impossible. What Micron is reporting now is structurally different. The cloud memory business unit, the most direct proxy for AI data center demand in the company's portfolio, generated $13.77 billion in revenue in the quarter, up from $3.39 billion a year ago, with an operating margin of 78 percent. The core data center unit grew even faster on a year-over-year basis. These are not the numbers of a cyclical business at a temporary peak. They are the numbers of a business that has found a new demand floor.

HBM4 and the Multi-Year Agreement Signal

The most strategically significant disclosure in Micron's earnings release was not the revenue figure or the guidance. It was the announcement of what the company called "transformational Strategic Customer Agreements" - 16 multi-year deals that lock in volume and provide pricing visibility for memory supply. HBM4, Micron's latest generation of high-bandwidth memory built on its 1-beta DRAM technology, is already in high-volume shipments to its lead customer, with qualification samples going to additional end customers. The 16 agreements represent approximately 20 percent of Micron's DRAM volume and a third of its NAND volume over the agreement period.

The significance of this disclosure is difficult to overstate. The memory business has always been vulnerable to the accusation that its current strength reflects temporary over-ordering by hyperscalers building out AI infrastructure - that the demand is real but the inventory cycle will eventually turn. Multi-year agreements with locked-in volume and pricing do not eliminate that risk entirely, but they substantially reduce it. Micron CEO Sanjay Mehrotra said the agreements "will significantly enhance the durability and predictability of Micron's strong financial performance." That is not marketing language. It is a structural claim about the business that the market will now spend the next several quarters either validating or disproving.

SK Hynix Brings the AI Memory Trade to Wall Street

If Micron's earnings were the exclamation point on AI memory demand, SK Hynix's US listing announcement was the broader market's response. The South Korean chipmaker - the world's second-largest memory producer and the dominant supplier of high-bandwidth memory to Nvidia - filed with the Securities and Exchange Commission on Tuesday to list American Depositary Receipts on the Nasdaq under the ticker SKHY. The company plans to issue up to 17.79 million new shares valued at approximately $29.4 billion, with bookbuilding beginning July 6, pricing on July 9, and trading commencing July 10.

The scale of the offering is remarkable. At $29.4 billion, it would be the second-largest equity offering in history, surpassing Saudi Aramco's $25.6 billion IPO in 2019 and Alibaba's similarly sized offering in 2014. Only SpaceX's $85.7 billion debut earlier this month ranks higher. SK Hynix is valued at approximately $1.1 to $1.2 trillion, its shares have risen 281 percent year to date, and as of the fourth quarter of 2025 it controlled roughly 57 percent of the global high-bandwidth memory market. Bank of America, Citigroup, Goldman Sachs, and JPMorgan are managing the offering.

The strategic rationale for the US listing is straightforward. SK Hynix trades on the Korea Stock Exchange, where its valuation has historically been discounted relative to US-listed peers. By listing ADRs on the Nasdaq alongside Micron, the company gains access to the deepest pool of technology-focused institutional capital in the world and creates a direct valuation comparison with its primary US competitor. Ryu Young-ho, a senior analyst at NH Investment Securities, put it plainly: "The most attractive benefit for investors is that SK Hynix will trade on Nasdaq alongside rival Micron, giving the company an opportunity to be re-rated in the US market." The proceeds will fund capacity expansion - new chip factories in South Korea and chipmaking equipment including extreme ultraviolet scanners from ASML.

What the Convergence Means for Investors

The simultaneous arrival of Micron's blowout quarter and SK Hynix's US listing announcement is not a coincidence. Both events reflect the same underlying reality: the AI infrastructure buildout has created a demand environment for high-bandwidth memory that is unlike anything the industry has experienced before, and the companies best positioned to supply that demand are now generating financial results and attracting capital at a scale that would have seemed speculative just 18 months ago.

The risk that investors need to hold alongside this optimism is the one that has always haunted the memory sector: the inventory cycle. Hyperscalers are ordering aggressively today because supply is tight and the cost of being caught short of HBM capacity is enormous. If AI infrastructure spending slows - whether because of macroeconomic pressure, a technology plateau, or simply a period of digestion after years of aggressive buildout - the same dynamics that drove Micron's revenue from $9.3 billion to $41.5 billion in a single year can reverse with similar speed. Micron's multi-year agreements are a partial hedge against that scenario, but they are not a guarantee.

What Wednesday's events confirm is that the AI memory trade has moved from speculative to structural in the eyes of the capital markets. Micron is now a $1.2 trillion company guiding to $50 billion in quarterly revenue. SK Hynix is raising $29 billion to build more capacity. The question for investors is no longer whether AI memory demand is real. It is whether the current pace of demand growth is sustainable, and whether the companies supplying it can maintain the pricing discipline that has produced margins previously unimaginable in this industry. The answers to those questions will define the next chapter of the AI trade - and Wednesday gave the bulls considerably more ammunition than the bears.