How memory, artificial intelligence and geopolitics created one of the great investment opportunities of 2025–2026
October 2026
The artificial intelligence revolution has created extraordinary wealth for investors over the past several years. Companies producing the computational infrastructure behind AI have experienced remarkable increases in revenue, profitability and market valuation. Yet while investors concentrated their attention on graphics processing units (GPUs), another critical component of the AI revolution remained comparatively overlooked: memory.
Micron Technology, one of the world’s largest manufacturers of semiconductor memory, became one of the standout investments of 2025–2026. Between May 2025 and October 2026, its share price increased more than elevenfold, transforming a $10,000 investment into over $110,000.
What made this opportunity particularly interesting was that the underlying investment thesis was relatively straightforward. As computational power increases, the demand for memory must increase alongside it. Combined with historically attractive valuations, government subsidies, geopolitical considerations and expanding AI infrastructure investment, Micron found itself at the intersection of several powerful economic forces.
The Relationship Between Memory and Compute
To understand Micron’s investment opportunity, we must first understand the relationship between computation and memory.
GPUs perform the mathematical operations necessary to train and operate artificial intelligence models. However, these processors cannot function efficiently without immediate access to enormous quantities of data.
Think of a GPU as a highly skilled worker and memory as the workspace containing the information required to perform a task. Increasing the number of workers without expanding their workspace eventually creates congestion.
The same principle applies to artificial intelligence.
As computational capacity increases, systems generally require greater memory capacity and bandwidth to prevent processors from sitting idle while waiting for information. Memory bandwidth describes how quickly information can move between memory and a processor. Capacity describes how much information can be held close to that processor.
Both are essential.
Modern AI accelerators increasingly rely on High Bandwidth Memory (HBM), a specialized form of DRAM constructed by vertically stacking memory chips and connecting them through extremely dense electrical pathways. This architecture enables extraordinary data-transfer speeds while improving energy efficiency. Micron’s HBM4 technology, for example, delivers more than 2.8 terabytes of bandwidth per second per stack, illustrating how rapidly memory technology is advancing.
Importantly, the relationship between memory and compute is not perfectly linear. Software improvements, compression and architectural innovations can reduce memory requirements for particular workloads. Nevertheless, expanding AI infrastructure creates substantial demand for increasingly sophisticated memory systems.
The central insight is simple: computational power becomes less valuable when memory cannot keep pace.
The Investment Opportunity Hidden Behind NVIDIA
During the early stages of the AI investment boom, investors understandably concentrated on NVIDIA. Its GPUs became essential infrastructure for training large language models, and its extraordinary revenue growth attracted enormous amounts of investment capital. However, the enthusiasm surrounding GPU manufacturers created an interesting valuation discrepancy.
Memory manufacturers such as Micron continued to trade under the shadow of their historical reputation as cyclical commodity businesses. Traditionally, memory manufacturing experienced dramatic fluctuations in profitability. When demand increased, manufacturers expanded production capacity. Eventually, excess supply caused prices to collapse, creating painful downturns.
Consequently, investors frequently assigned memory manufacturers lower earnings multiples than companies producing specialized processors. But the AI revolution was changing the economics of memory. HBM required advanced manufacturing processes, sophisticated packaging and close collaboration with accelerator designers. These characteristics created barriers to entry and made certain memory products considerably more valuable than conventional consumer DRAM.
Investors who recognized this transformation early could purchase exposure to AI infrastructure without paying the premium valuations associated with the most celebrated GPU manufacturers. The opportunity was not simply that Micron was inexpensive. It was that the market may have been valuing a changing business using assumptions inherited from an older industry structure.
Higher-Value Memory and the Expansion of Profitability
Not all memory is economically equivalent. A gigabyte of conventional memory installed in a consumer laptop serves a very different purpose from a gigabyte of HBM integrated into an advanced AI accelerator. Specialized memory requires more complicated manufacturing processes, greater reliability and significantly more technical integration. Consequently, these products can command higher prices per bit.
For Micron, this created an opportunity to improve profitability without necessarily increasing its share of the overall memory market. The company could prioritize higher-value products, including HBM, enterprise server memory and advanced storage solutions. This transition became increasingly visible in its financial performance. Micron reported fiscal 2026 revenue of approximately $133.2 billion, compared with $37.4 billion in fiscal 2025.
Such extraordinary growth demonstrated how rapidly memory economics could change when surging demand encountered constrained manufacturing capacity. The investment opportunity was therefore driven by two forces: expanding memory consumption and substantially improved profitability on the memory being sold.
Geopolitics: Bringing Semiconductor Manufacturing Home
Artificial intelligence is not merely a commercial technological revolution. It is increasingly considered a matter of national economic and security importance. Semiconductor manufacturing remains heavily concentrated in East Asia, creating vulnerabilities for countries dependent on uninterrupted access to advanced chips. Growing geopolitical tensions between the United States and China, particularly surrounding Taiwan and semiconductor technology, have intensified concerns about supply-chain resilience.
Washington has consequently prioritized bringing semiconductor manufacturing back onto American soil. The CHIPS and Science Act represents one of the most significant industrial policy initiatives supporting this objective and Micron became a major beneficiary.
In December 2024, the US government finalized an award of up to $6.165 billion in direct funding to support advanced memory manufacturing projects in Idaho and New York. In June 2025, Micron announced an expanded American investment vision totaling approximately $200 billion, encompassing manufacturing facilities, research and advanced packaging capabilities. These investments are intended to reduce American dependence on overseas semiconductor production while strengthening domestic technological capabilities.
For Micron, government support helps offset the enormous capital expenditure required to construct advanced manufacturing facilities. However, subsidies do not automatically guarantee higher profits. New factories remain expensive, and their economic success ultimately depends on demand, manufacturing efficiency and industry pricing. Nevertheless, Micron’s position as the only major US-headquartered manufacturer of DRAM gives it considerable strategic importance.
India and the Diversification of Global Supply Chains
Micron’s geographic expansion extends beyond the United States. In February 2026, the company inaugurated a semiconductor assembly and testing facility in Sanand, Gujarat, India. The project represents approximately $2.75 billion in combined investment from Micron and its government partners. Importantly, this facility does not manufacture advanced memory wafers from scratch. Instead, it assembles and tests wafers produced elsewhere in Micron’s global manufacturing network. This distinction matters because semiconductor supply chains involve multiple stages, including fabrication, packaging, assembly and testing.
Diversifying these operations reduces dependence on any single geographic location. India also offers access to an expanding technical workforce and a government actively encouraging semiconductor investment. For Micron, the expansion provides additional manufacturing flexibility while strengthening its position within an increasingly diversified global technology ecosystem.
The Beginning of a Memory Supercycle?
Perhaps the most compelling element of Micron’s investment thesis concerns the next five years. Historically, memory markets have experienced relatively short periods of prosperity followed by painful oversupply.
However, artificial intelligence introduces the possibility of a longer-lasting demand expansion. The initial wave of AI investment concentrated on training increasingly capable models.
The next phase involves deploying those models across businesses, governments, consumer applications and physical machines. This transition from training to inference could substantially expand the installed base of AI infrastructure.
Every additional AI server requires memory. Larger context windows increase memory requirements. More simultaneous users require greater working capacity. Autonomous vehicles, industrial robots and intelligent devices introduce further opportunities for specialized memory.
Micron is also establishing closer relationships with strategic customers. Its June 2026 agreement with Anthropic, for example, includes collaboration on AI infrastructure architecture and memory supply.
Such arrangements potentially provide greater demand visibility than the traditional commodity memory business. The resulting investment thesis is that sustained AI adoption, combined with disciplined capacity expansion and increasing technical complexity, could extend the duration of the memory industry’s expansion.
A genuine supercycle would mean more than rapidly rising demand. It would require demand to remain strong enough to absorb new manufacturing capacity while supporting attractive profitability.
Competitors including Samsung, SK hynix and emerging Chinese manufacturers continue investing in production capabilities. Excessive industry expansion could eventually recreate the oversupply conditions that characterized previous cycles.
The Investment Lesson
Micron’s extraordinary performance between 2025 and 2026 illustrates an important principle of investing.
The most obvious beneficiary of a technological revolution is not necessarily the most attractively priced investment.
While markets focused on the companies producing computational power, memory manufacturers offered exposure to an equally essential component of the emerging AI economy. Investors who understood the relationship between compute and memory could recognize that the expansion of one industry would create substantial demand for the other.
When this technological relationship was combined with attractive starting valuations, improving product economics, government subsidies and geopolitical incentives, the opportunity became particularly compelling.
Whether the coming five years deliver a genuine memory supercycle remains uncertain. Semiconductor manufacturing will always involve substantial capital requirements, technological competition and fluctuations in supply and demand. Nevertheless, Micron’s transformation demonstrates how technological change can alter the economics of an established industry. The broader lesson extends far beyond semiconductors.
When a new technology transforms the world, investors should examine not only the companies creating that technology, but also the indispensable infrastructure required to make it work. Sometimes the greatest opportunities are found in the components everyone needs, but comparatively few investors initially appreciate.
