Micron's Q4 Revenue Nearly Quadruples as $61.5B Guidance Tops Estimates

Micron reported fiscal fourth-quarter results on Wednesday that beat estimates, with revenue almost quadrupling year on year, and guided fiscal first-quarter revenue to about $61.5 billion, above the $57 billion analysts polled by LSEG had expected, according to CNBC.

The memory maker, the only U.S.-based producer of high-bandwidth memory used in AI systems, said adjusted earnings per share for the coming quarter should come in at about $38.15, against the $35.40 analysts had modelled. Net income in the quarter just ended climbed to $37.7 billion, or $32.87 per share, from $3.2 billion, or $2.83 per share, a year earlier, the company's statement says.

Fourth-quarter DRAM revenue rose 343% from a year ago to $39.8 billion, or 73% of total sales.

Micron's stock has gained more than 500% over the past year, lifted by a worldwide supply crunch in memory chips, and its market capitalisation has topped $1.2 trillion even though it holds the smallest HBM market share of the three leading suppliers, alongside SK Hynix and Samsung. That combination, a stock that has already multiplied and a company that is not the largest player in the fastest-growing part of the market, shapes how the quarter's numbers landed with investors.

Guidance and the capital bill

Micron says it is investing $250 billion to build two new HBM campuses. The largest broke ground in Clay, New York, in January, and its first new fab in Boise, Idaho, is scheduled to come online next year, CNBC reports.

CEO Sanjay Mehrotra told the earnings call the company has a "strong roadmap for future HBM products" and is working with Nvidia on what he described as the industry's first custom HBM implementation. He also said Micron raised compensation for every employee in fiscal 2026, a detail that lands against a background of labour tension: hundreds of Micron staffers threatened to strike at its factories in Taiwan over pay and negotiations, while similar action at SK Hynix and Samsung produced bonuses above $500,000.

Hendi Susanto, a portfolio manager at Gabelli Funds, wrote in an email after the release that it was "another strong beat and raise for Micron", adding: "At this point, I have not heard any negative data points pointing to the memory cycle reversing toward a decline anytime soon for the foreseeable future."

Analysts remained constructive into the print. JPMorgan reaffirmed a $1,540 price target and described a "constructive" setup ahead of the report, according to finance.biggo.com. Yahoo! Finance Canada reported that analysts stayed bullish after the beat-and-raise, with the memory demand outlook extending into 2027.

What the market actually did

The share reaction was far from euphoric. CNBC noted the stock rose only slightly in extended trading, and reported on Thursday that it "barely budged" on the results, with Wall Street still expecting gains to come later.

Other outlets read the tape differently. inkl.com headlined that the stock was "plunging anyway" despite the beat, while TradingView reported the stock dropped after hours as a heightened capex forecast "overpowered" the Q4 beat and the above-consensus Q1 guidance. That is a genuine divergence in framing rather than a difference in the numbers, and it reflects how much of the AI memory trade is already priced in.

The quarter's headline revenue figure also needs care. qz.com and Seoul Economic Daily both reported revenue of $54.2 billion, while tokenpost.com described a revenue jump of 379%. CNBC frames the same quarter as revenue "almost quadrupling" from $11.32 billion a year earlier. The company's own statement is the cleanest reference point for the year-on-year comparison.

Read-through to Asia and the wider supply chain

Micron's numbers moved chip stocks across Asia. Investing.com reported that Japanese chipmakers rallied on the earnings, calling it a boost to the AI trade, and Yahoo Finance reported Asian stocks rising on chipmaker gains and soft U.S. inflation, with the Nikkei outperforming. The Korea Herald described a deepening memory shortage extending the boom for Korean chipmakers, and Korea JoongAng Daily reported that Micron's record earnings signalled upside for Samsung and SK hynix.

Attention now turns to TSMC, which has set its earnings call for 15 October, with AI-chip demand in focus, per tokenpost.com. Nvidia, meanwhile, has been spending its own cash: its board increased the chipmaker's share buyback plan by $150 billion, the Boston Herald reported on 28 September.

Not only memory

Elsewhere in the supply chain, GlobalFoundries sees "strong" demand for Chinese optical modules amid the data centre construction boom, Shankaran Janardhanan, senior vice-president of the firm's photonics and radio frequency division, told the South China Morning Post in an interview published on 27 September. Janardhanan said the appetite for Chinese optical modules, used to transmit data in high-performance data centres, had partly driven the company's decision to bet heavily on photonics. He was speaking at the company's annual tech summit in Shanghai.

On the battery side, Transport & Environment published analysis on 30 September finding that by 2030 sufficient Made-in-EU cells will be available for the scope of the Industrial Accelerator Act, including corporate and private EVs under a subsidy or tax scheme, provided all projects materialise, including medium-confidence ones. CleanTechnica, which carried the analysis, reports that local content requirements for cells and cathode active materials could mean an additional 34% in EU battery demand compared with the current baseline in 2027.

Xavier Sol, director of sustainable investments and batteries at T&E, said: "European industrial competitiveness hinges on a simple question: will high value clean tech jobs come to Europe or remain in Asia?" T&E's analysis directly contradicts recent claims by the car industry that local content criteria are too ambitious, CleanTechnica says, and it notes that China controls up to 90% of global pCAM and CAM production capacity depending on chemistry, and as much as 95% for LFP.

Further up the stack, a paper submitted to arXiv on 28 September by Johnny Jingze Li, Abdulla Kuleib, Kalyan Basu and Gabriel A. Silva examined how 17 open-weight language models redistribute activity across attention heads as a task demands more serial steps. The authors report that in most models, layers just before mid-depth concentrate activity while later layers spread it, and that Qwen2.5 base models from 0.5B to 7B spread less than the average model in every task, concentrating activity in parts of their second half where Llama models from 1B to 8B and OLMo-2 spread. The contrast largely holds between Llama-3.1-70B and Qwen2.5-72B, which have the same number of layers and heads, and post-trained models keep much of their base model's pattern.

In semiconductor manufacturing, Emergence AI said it is moving its neuroformal AI technology into active deployments with fabless chipmakers, while integrated device manufacturers among its customers are asking it to extend the work into the fab, EE Times reported on 30 September. Satya Nitta, co-founder and executive chairman, told EE Times at SEMICON India 2026: "You cannot make chips any faster, but what you can do is definitely get more chips per wafer yielding with AI." Nitta said neuroformal AI pairs large language models with symbolic AI, with algorithms proposing and symbolic verification checking the result. He took the executive chairman and chief scientist roles about a month ago when Emergence AI appointed Ian Eslick as CEO, EE Times reports.