As Meta Muse and OpenAI Astra drive the global expansion of AI applications toward continuously executing, multi-step collaborative agent workflows, expectations for a new round of explosive growth in global AI computing power demand are surging. This is also a key reason why global stock markets have recently been translating the faster-than-expected adoption of AI agent applications into growth expectations for a host of AI computing power industry leaders and core computing power suppliers, including SK Hynix, Samsung, Micron, Nvidia, AMD, and TSMC.

The market is currently reassessing demand for AI core hardware such as CPUs, memory, and storage that support these cutting-edge AI applications. Micron (MU.US), one of the three major storage chip manufacturers, is set to take the stage on Wednesday U.S. Eastern Time and release its highly anticipated earnings report, making it an important window for testing the storage super-cycle and the concrete progress of the unprecedented AI infrastructure boom. Earlier, from mid-July to August, storage leaders including SK Hynix, Samsung, Western Digital, Seagate, and NetApp had already reported stronger-than-expected results, and a series of AI computing power industry leaders including TSMC, AMD, Nvidia, and Broadcom had also announced strong performance data and optimistic outlooks for AI computing power demand.

Wall Street analysts' consensus expectations show that Micron is expected to report revenue of approximately $51.47 billion for the fourth quarter of fiscal 2026, with earnings per share of $31.82. The consensus revenue expectation implies Micron's revenue could achieve a staggering year-over-year increase of about 354.9%. However, such a high threshold also means Micron's results and management's quarterly outlook need to substantially beat expectations, and Micron may also need to provide strong shareholder returns—such as stock buybacks—for the share price to rise after the earnings disclosure. The options market is pricing in a sharp two-way move of about 7.7% after earnings, with call option trading and volatility skew indicating investors are more actively seeking upside gains. The key point jointly favored by JPMorgan and UBS is that tight storage supply and demand may persist, while long-term agreements are expected to enhance the visibility of Micron's future earnings. The 7.7% figure is the two-way implied move corresponding to the cost of a straddle option, not an upward forecast or a guaranteed trading range.

On analyst price targets, JPMorgan's latest bullish research report on Micron shows it maintains an "Overweight" rating, its most optimistic target, with a 12-month price target of $1,540; UBS maintains a "Buy" rating and a 12-month price target of $1,625. As of Tuesday's U.S. market close, Micron shares rose 1.05% to close at $1,065.08, bringing their year-to-date gain to about 275%. In the view of AI application and AI large-model leaders Anthropic and OpenAI, storage chip components for AI data center server clusters, as well as AI GPUs, remain the clearest supply bottlenecks in the AI computing power industry chain. Market research firm TrendForce estimates that server DRAM contract prices will rise by a cumulative 270% in 2026, while enterprise SSD prices will rise by a cumulative 235%; in 2027, HBM contract prices may still rise by 70%–140%, continuing to show doubling growth. These data reflect the combined effect of continuously expanding AI computing power demand and rising storage chip prices. TrendForce calculations also show that shipments of NVL72 racks covering the Blackwell and Vera Rubin platforms are expected to grow by more than 50% year over year in 2027; its research chart shows that the output value of related systems is expected to rise from about $226 billion in 2026 to $711 billion in 2027, a sharp year-over-year increase of 214%.

The More Capable AI Agents Become, the Stronger Storage Demand Gets—How Demand Spreads from HBM to DRAM and SSD

For U.S.-based storage chip giant Micron, its advantages include covering HBM, server DRAM, and NAND storage simultaneously, allowing it to participate in upgrades at different layers of agent infrastructure. Compared with the two major storage giants SK Hynix and Samsung Electronics, its biggest advantage is undoubtedly that it does not have to worry about operating pressure that could be triggered by the huge political pressure from the U.S. government to "bring chip manufacturing back to the United States," nor the uncertainty of tariff policies that the Trump administration could change at any time.

The latest product developments are expanding the scope of work AI can perform. Meta expanded Muse to small-business scenarios on September 29, connecting to business tools such as QuickBooks, Shopify, Stripe, and Slack; OpenAI launched Dots powered by GPT-6 Astra the same day, equipping agents with independent cloud computers and supporting continuous task execution and connections to a large number of applications. The potential incremental value of such products lies in the fact that a single user instruction can trigger multiple steps including planning, retrieval, reading files, calling tools, executing code, and verifying results. It is worth noting that these AI agent technology developments in September directly affect Micron's subsequent demand and business guidance, and cannot be equated with new growth sources for the already-completed August fiscal quarter.

From the perspective of inference systems, the prefill stage needs to process input context, the decoding stage continuously generates output, and the KV cache retains reusable attention states. As context becomes longer, concurrent tasks increase, and task duration extends, the system needs to simultaneously improve computing throughput, memory capacity, and data movement efficiency: HBM handles high-bandwidth active computing and caching, server DRAM provides greater capacity, and enterprise SSDs store reusable context, files, and long-term data. High-performance server CPUs in data centers execute tool calls, retrieval processing, code execution, and task orchestration, while high-speed networking handles data exchange across nodes. These also actively form the engineering logic for demand spreading to AMD, Intel, and Arm-architecture server CPUs, optical interconnect, and storage systems, and also make data center power supply and cooling important conditions for large-scale deployment. For Micron, the opportunity therefore spans multiple product lines including HBM, server DRAM, and enterprise SSDs.

Cache reuse and algorithm optimization can reduce the resource consumption of individual tasks, but if lower costs attract more users, more tasks, and higher concurrency, total infrastructure demand may still expand. UBS expects DRAM and NAND bit shipments in the current fiscal quarter to grow only about 3% and 4% quarter over quarter, respectively, while average selling prices rise about 23% and 22.5%, respectively, largely indicating that its short-term profit forecast is mainly driven by pricing; whether long-term valuation can rise further requires support from demand sustainability, supply discipline, and the realization of long-term storage chip contract value.

Ahead of Micron's Fiscal Fourth-Quarter Earnings, the Options Chain Clearly Shows a Bullish Bias

Hedge fund traders in the options market are betting almost unanimously that U.S.-based storage chip maker Micron Technology (MU.US) could see its stock price swing nearly 8% up or down after it releases fiscal fourth-quarter results after Wednesday's close. During Tuesday afternoon U.S. trading, the semiconductor giant's shares rose as much as 1.8%, trading near $1,072 per share, as market participants positioned themselves ahead of the earnings release. Options chain data for contracts expiring October 2 show that an at-the-money straddle costs about $82.28, implying an expected post-earnings stock price move of about 7.7%. The potential post-earnings trading range corresponding to this expected move is roughly $990 to $1,155 per share before the end of this week. Trading activity shows a clear bullish bias, with a large volume of call options concentrated at key upside strike prices. This optimism is also supported by volatility skew on the call side: compared with put options equally far from the current stock price, out-of-the-money call options carry a clear premium, indicating that market participants are more willing to pay to participate in a bull-market rally than to buy protection against a panic decline.

Call options expiring October 2 with strikes as high as $1,100 rank far ahead of the entire options chain in volume, with more than 12,400 contracts changing hands; call options with strikes of $1,050 and $1,200 also attracted significant demand. On the put side, trading was mainly concentrated at $1,000 and $1,050 strikes, indicating that some cautious traders are positioning around key technical support levels to guard against a disappointing earnings report. Meanwhile, extreme tail-risk positioning shows speculative bets at both extremes: thousands of new bought contracts appeared in deep out-of-the-money calls with strikes of $1,400 and $1,500, corresponding to large-scale hedging positions concentrated in puts with strikes of $600 and $650. Wall Street analysts consensus expects Boise, Idaho-based Micron to report earnings per share of $31.82 and revenue of $51.47 billion. Over the past two years, the company's revenue and earnings per share have exceeded market expectations 100% of the time.

AI Computing Power Frenzy Faces a Major Test! On Micron's Earnings Day, the Market Focuses on Price Increases, Long-Term Contracts, and Shareholder Returns

As Meta Muse and OpenAI Astra push agents toward more complex office, programming, and business tasks, the market is reassessing AI inference demand for storage capacity, bandwidth, and data processing capability. Micron's earnings report, scheduled to be released after the U.S. market close on September 30, will become an important window for observing how this round of demand translates into revenue for core chip components related to AI infrastructure and into profit and cash flow trends for leaders in the AI computing power industry chain. Among them, the key reason international financial giant UBS and Wall Street's JPMorgan both favor Micron is that tight storage supply and demand may persist, while long-term agreements are expected to enhance the visibility of Micron's future earnings. What JPMorgan values is the mutual reinforcement between supply-demand tightness and long-term agreements.

The bank expects Micron's August-quarter DRAM average selling prices to rise more than 20% quarter over quarter, and NAND average selling prices to rise about 20% quarter over quarter. Regarding management's earlier mention of slowing price increases, JPMorgan prefers to understand it as a commercial choice to maintain customer relationships and promote the signing of long-term agreements. Even after factoring in HBM specification downgrades, the bank still expects HBM supply-demand gaps of 20%, 19%, and 16% for calendar years 2026–2028, respectively. At the same time, among the 16 SCA agreements Micron previously disclosed, 14 have remaining performance obligations totaling about $100 billion based on floor prices; JPMorgan expects the proportion of future bit output covered by long-term agreements may have risen above 35%, and could even reach above 50%. The $100 billion here is the future performance amount, not revenue already recognized in the current year, and the increase in coverage still needs to be verified by the earnings report.

UBS, meanwhile, pushes the discussion toward "how long can high profitability last." The international bank's latest channel checks show that the 2027 DRAM demand fulfillment rate may still be only about 60%, server DDR bit demand is expected to grow about 80% year over year, and server and storage system SSD bit demand may grow more than 100%; the bank expects DRAM undersupply to continue at least until the second quarter of 2028. UBS believes long-term agreements will limit some short-term price increase gains, but can support the earnings foundation for future downturns. Institutions are not fully aligned on the price inflection point: Citi, in its latest bullish research report on Micron, expects DRAM and NAND prices to peak in the second quarter of 2027, while UBS places the NAND price peak in the third to fourth quarter of 2027, so this guidance's judgment on the duration of the boom is especially critical.

The significance of long-term agreements among these major storage chip makers for AI computing power thematic investing is that customers exchange long-term procurement commitments for supply guarantees, while Micron gains clearer capacity planning and revenue visibility, helping reduce the sensitivity of some businesses to short-term market prices. Micron management also recently expected these agreements to bring about $22 billion in cash deposits and related financial commitments, of which about $18 billion is cash deposits. They reflect customers' willingness to lock in long-term supply and also help support capacity expansion arrangements; in accounting terms, this portion of cash deposits is recorded in financing activities and will be gradually returned in the latter half of the agreements. Therefore, what deserves more attention in the earnings report is new contract signings, coverage, product mix, and actual delivery progress, as well as how these changes improve the predictability of future operating cash flow.

In addition, after Nvidia重磅 announced a record $150 billion single authorization buyback plan, the largest in U.S. stock market history, and Micron competitors SK Hynix and Samsung both rolled out incremental shareholder return plans, capital returns form another important main line. JPMorgan expects August-quarter free cash flow to exceed $24 billion and is watching arrangements to gradually return excess cash flow after December 9, 2026; UBS expects Micron may begin repurchasing about $20 billion per quarter starting in the second quarter of fiscal 2027, the February 2027 quarter, and then expand to $40 billion–$50 billion. The latter's Micron buyback outlook is an analyst forecast model, not a buyback plan the company has announced. From this, it can be inferred that Micron's earnings report may be widely viewed by investors as a "cash realization test" for the AI computing power investment frenzy—whether demand expansion can form sustained orders, whether long-term agreements can stabilize profitability, and whether cash accumulation can translate into per-share value enhancement will jointly affect what valuation the market is willing to give Micron.

Micron's stock has risen more than 275% year to date, outperforming the S&P 500's 12% gain, making it one of the biggest winners in the AI-driven global tech stock rally. The company produces high-bandwidth memory chips used with Nvidia's AI processors, and demand has surged; severe global supply shortages have prompted it to raise prices for DRAM and NAND storage chips. Citi analyst Atif Malik said: "We expect the DRAM and NAND markets to remain undersupplied, supported by strong AI-driven storage demand. We maintain our previous view that DRAM and NAND price increases will slow over the next four quarters, with prices likely peaking in the second quarter of 2027." Quantitative ratings jointly given by Wall Street analysts and Seeking Alpha analysts are bullish on the stock, assigning it a "Buy" or higher rating. Over the past three months, Wall Street analysts' earnings per share estimates for Micron have been raised 11 times and lowered twice; revenue estimates have been raised 12 times and lowered twice. RBC analyst Srini Pajjuri said: "For the first fiscal quarter, we expect management's guidance to be 3%–5% above the consensus of $56 billion in revenue and $34.85 in earnings per share. Despite some impact from SCA agreement price caps, we still expect first-quarter blended DRAM average selling prices to rise 5%–10%." He added that the firm's model expects Micron's calendar 2027 free cash flow to exceed $100 billion. Wedbush analyst Matt Bryson said Micron's earnings report could, and should, have a positive synergistic effect on the share prices of U.S.-listed storage chip peers such as SK Hynix and SanDisk. He added: "Although we remain bullish on the storage sector based on the above factors, it should be noted that, because it includes June, we believe Micron's fiscal fourth-quarter price performance will be stronger than the price performance reflected in peers' third calendar-quarter results."