Skip to content 
Let’s start with a number that would have sounded absurd eighteen months ago.
Micron Technology, a chipmaker from Boise, Idaho, best known for making the memory chips inside your laptop and phone, is now worth over $1 trillion.
Its stock has gained nearly 860% in the past twelve months. It is up more than 70% in May alone, putting it on track for its biggest single month since December 1987.
And it took the company just 48 trading days to go from a $500 billion valuation to a $1 trillion one. That is the fastest doubling in market cap history, beating Samsung’s record of 82 days, which was set just weeks earlier. Prior to that, Tesla had held the record since 2021 with a 230-day run between the two milestones.
The meteoric rise, which includes a gain of more than 186% from its March 30 trough, comes as investors pile into stocks that stand to benefit from soaring demand for memory semiconductors, a crucial element for building AI infrastructure.
So the obvious question is: what on earth is happening?
Memory is having its moment
To understand Micron’s rise, you need to understand what memory chips actually do in the age of AI.
When you use ChatGPT or any large language model, the AI system needs to store and retrieve enormous amounts of data very quickly while it processes your request. That requires a special kind of chip called High Bandwidth Memory, or HBM. Think of HBM as the short-term working memory of an AI system, the layer that keeps all the context alive while the AI is thinking.
Nvidia’s powerful AI chips, which power most of the world’s AI data centers, need HBM to function. And there are only three companies in the world that can make HBM at scale: SK Hynix, Samsung, and Micron. That’s it. No one else is close.
This scarcity is not theoretical. As of early 2026, Micron had officially sold out its entire HBM3E and HBM4 inventory for the year. Every chip it can make has already been spoken for. The waiting list is real. Analysts at Mizuho noted that Micron might be able to double prices of its latest HBM4 chips next year, such is the imbalance between supply and demand.
The numbers are not a drill
Micron’s recent financials look less like a semiconductor company and more like a software business in hypergrowth mode.
In Q2 FY2026, Micron reported total revenue of $23.9 billion, up 196% year on year. DRAM revenue hit $18.8 billion, up 207%.
Gross margins expanded to 75%, up 18 percentage points in a single quarter. Non-GAAP EPS came in at $12.20, up 682% year on year.
Free cash flow hit a quarterly record of $6.9 billion. The Cloud Memory division, which is essentially the HBM unit powering Nvidia’s AI GPUs, saw revenue nearly double to $5.28 billion with gross margins of 66%.
Those are not typos.
And then came the forward guidance, which is where analysts genuinely lost their minds.
Micron guided Q3 FY2026 revenue at $33.5 billion, with gross margins of approximately 81% and EPS of $19.15.
All of that was well above the prior consensus estimate of $24.29 billion. An 81% gross margin. For a company that makes physical chips. In factories. Most software companies would envy that number.
The UBS moment
On May 26, UBS analyst Timothy Arcuri published what Bloomberg described as the most aggressive single analyst revision in recent semiconductor history. He raised his price target on Micron from $535 to $1,625, a 204% increase in one note, implying a market value of roughly $1.8 trillion. That would make Micron larger than Meta Platforms, Tesla, or Berkshire Hathaway at their current valuations.
The stock surged 19% that day, its biggest single-day gain since November 2011, crossing the $1 trillion market cap threshold for the first time. A day later it pushed to another all-time high, gaining a further 6.7%.
The core of the UBS argument was this: the market has been applying a cyclical discount to Micron for decades, assuming that every boom would be followed by a brutal bust.
What AI has done is structurally changed the demand picture, and once the market starts applying a more normal, durable-growth multiple to the stock, the re-rating has much further to run.
The average price target sits around $685, which, remarkably, implies some downside from current levels even as every single analyst is bullish, because the stock has simply run past the targets faster than the upgrades can keep up.
Why this time might actually be different
Memory chips have always been cyclical. Prices boom when supply is tight. Companies invest heavily to build new fabs. New supply floods the market. Prices crash. Companies lose money. Repeat every four to six years. This cycle has destroyed investor returns many times over.
So why should anyone believe this time is different?
The argument goes like this. AI infrastructure is not a one-time spending spree. It is a permanent, structural shift. The major cloud providers are committing hundreds of billions of dollars every year to data center expansion, and that spending is not going away. Each new generation of AI models requires more memory, not less. As AI usage moves beyond chatbots into coding agents and longer-running tasks, systems need to retain more information, which means more memory per server.
Here is the structural kicker that most people miss. Each gigabyte of HBM consumes three times the wafer capacity of standard DDR5 memory. So every time a fab shifts production toward HBM for AI, it simultaneously tightens supply of regular DRAM for consumers. The result is a price squeeze across the entire memory market, not just the premium AI segment.
DRAM prices are up 171% year on year. DDR5 spot prices have quadrupled since late 2025. NAND flash prices are climbing 70 to 75%. The AI infrastructure boom is lifting all of memory, not just the high-end slice.
Three and a half years after ChatGPT kicked off an unprecedented trillion-dollar spending spree on data center infrastructure, Big Tech companies and the top AI startups are still unable to find enough computing capacity to serve their customers. Demand for memory is widely expected to continue outstripping supply until at least next year, with some analysts projecting shortages could continue until late in the decade.
The scale of what is being built
The numbers around Micron’s expansion plans give you a sense of just how seriously the company is betting on this structural shift.
Micron raised its full fiscal year 2026 capex outlook to over $25 billion, up from prior guidance of $20 billion. It broke ground in January on a massive $100 billion campus in New York, with wafer output expected by the second half of 2028. Volume production of HBM4 for Nvidia’s Vera Rubin platform has already started. Next-generation HBM4E products are on track to ramp in 2027.
Meanwhile Nvidia’s latest quarterly filing revealed it had increased its own manufacturing, supply, and capacity commitments by 27% since January to $119 billion, as it tries to secure memory and production pipelines amid intense competition. When your biggest customer is signing $119 billion in supply commitments, the demand picture looks rather durable.
The geopolitical angle adds another layer. Micron is expanding US manufacturing as part of a broader domestic investment programme worth $200 billion. That has won the company vocal support from the White House, and sparked speculation that the US government may seek a stake in the company, similar to deals it has struck with Intel and a group of quantum computing firms.
What could go wrong
None of this means the stock is risk-free. Far from it.
Supply expansions by all three memory giants simultaneously are the central bear case. If SK Hynix, Samsung, and Micron are all building new fabs at the same time, the supply glut that has historically crushed memory prices could return within two to three years. The history of this industry is littered with exactly this pattern. Every boom has eventually ended the same way.
Then there is the capital expenditure burden. Micron is now spending over $25 billion a year building factories. If AI demand slows, or if a next-generation chip architecture reduces the amount of HBM needed per system, those investments become a serious liability sitting on the balance sheet.
There is also the broader productivity question worth sitting with. The equity market is currently pricing Micron as though the AI-driven productivity boom is not just real but durable and large. The San Francisco Fed, however, estimates only a 21% probability that the economy is currently in a genuine high total-factor-productivity regime, even as labor productivity has risen sharply. The gap between those two numbers is essentially the uncertainty premium embedded in every AI-related stock at current valuations. If the productivity payoff from AI takes longer to materialize, or proves smaller than hoped, memory demand can still be decent, but the valuation multiples of companies like Micron are almost certainly too high.
So, bubble or breakthrough?
The honest answer is: it depends entirely on your time horizon.
In the near term, the fundamentals are extraordinary and the momentum is real. HBM supply is sold out. Margins are at historic highs. Guidance is staggering. The AI build-out shows no signs of slowing. And Micron is not alone in this rally. SK Hynix is up 1,007% over the past year. Samsung is up 469%. Intel is up 500%. The entire memory complex is being repriced, not just one lucky company.
In the medium term, the key risk is not a sudden collapse in AI demand. It is a supply response. When it arrives, margins will compress. The only real question is whether, by the time that happens, the earnings base is so much larger that even a normalized multiple still justifies a high stock price. Analysts expect roughly $105 in adjusted EPS by the end of fiscal 2027, a more than 1,200% increase from the $8.07 Micron reported for fiscal 2025. When earnings grow that fast, even a tripling of the stock price can leave it looking statistically cheap.
What is certain is this: memory chips, long dismissed as a dull commodity business prone to painful cycles, are now at the center of the most important infrastructure build in a generation. Micron is no longer just a memory company. It is, whatever the right price turns out to be, an AI infrastructure company. And that kind of re-rating, once the market accepts it, tends to be very hard to reverse.
