Aura Kasih
As the tech world anticipates the unveiling of Apple’s latest iPhone iteration, a significant, albeit unwelcome, shift is on the horizon: an anticipated price increase. This potential adjustment by the industry titan signals a broader trend driven by escalating memory costs, a phenomenon that appears to have no immediate resolution. This surge, often dubbed "chipflation" or "RAMageddon," marks a departure from decades of declining memory prices that fueled the affordability and advancement of consumer electronics. The pervasive nature of this issue is evident in the dramatic increase in mentions of "memory prices" and "memory shortage" across corporate earnings calls, underscoring a widespread industry challenge that will likely persist for years.
While the meteoric rise of Artificial Intelligence is frequently cited as the primary catalyst for increased costs across various sectors, including technology, the roots of the current memory crunch predate the widespread adoption of generative AI. The current market dynamic is the result of a complex and highly profitable restructuring within the memory manufacturing industry, occurring within a system that was already struggling to meet burgeoning demand. Manish Bhatia, President and COO of Micron, a leading memory manufacturer, emphasized the fundamental challenge: "We need to build more wafer capacity… [It’s] a very different challenge for the industry than it had been for many years before, where technology alone was able to keep up with the demand." This underscores a shift from incremental technological improvements to a requirement for substantial capital investment in physical infrastructure.
For an extended period, memory manufacturers were able to expand production by refining their fabrication processes to yield more chips from each silicon wafer. However, these incremental gains began to diminish, and the time required to achieve them grew longer. By 2021, even amidst a surge in demand for electronics during the pandemic, Micron recognized that technological advancements alone would no longer suffice to meet long-term demand. A fundamental change was necessary, requiring not only increased wafer processing but also the construction of massive new manufacturing facilities.
This strategic imperative was complicated by a subsequent downturn in the memory market. The surge in consumer electronics purchases during the pandemic had effectively pulled forward future demand. As consumer spending contracted, manufacturers found themselves with substantial excess inventory. This led to financial losses and a temporary deceleration of expansion plans. It was only as the market began to recover that the unprecedented demand generated by generative AI emerged, far exceeding the industry’s previous projections and its capacity to supply.
The global memory market is characterized by a high degree of consolidation, with three primary manufacturers—Samsung, SK Hynix, and Micron—accounting for approximately 90% of the market share. This concentrated structure leaves the global technology ecosystem heavily reliant on the production decisions of a few key entities, forcing them to allocate limited manufacturing capacity between the rapidly expanding AI infrastructure and the needs of consumer devices. In the second quarter of 2026, Counterpoint Research data indicated that Samsung held an estimated 39% of the memory market, followed by SK Hynix with 26%, and Micron with 25%.
The memory components within consumer electronics generally fall into two main categories: Dynamic Random-Access Memory (DRAM) and NAND flash memory. DRAM serves as a device’s short-term working memory, temporarily storing data required for active applications, web browsing, and software execution. NAND flash memory, conversely, provides long-term storage for files, photos, and other data.
Within the data centers powering artificial intelligence, specialized processors heavily utilize a high-performance variant of DRAM known as High-Bandwidth Memory (HBM). David Naranjo, Associate Director at Counterpoint, clarified, "It’s not as simple as saying data centers are consuming RAM. The RAM is not the same." HBM achieves its enhanced capabilities by stacking multiple memory chips vertically and employing advanced interconnectivity and packaging technologies. This architecture allows for significantly faster and more efficient data transfer between HBM and AI processors.
The production of HBM is considerably more complex than that of conventional DRAM, but its higher performance commands a premium price. Leading AI chip manufacturers such as Nvidia and AMD, alongside technology giants like Meta and Microsoft, exhibit an insatiable demand for HBM to fuel their advanced AI systems, and they are willing to invest substantial capital to secure this critical component. Furthermore, HBM consumes a disproportionately larger amount of manufacturing capacity compared to standard DRAM. The fabrication of memory chips occurs on large, circular silicon wafers, with multiple chips produced simultaneously. A finished HBM product, comprising multiple stacked and larger chips, necessitates a greater quantity of silicon. Micron estimates that producing an equivalent amount of HBM requires approximately three times the number of wafers compared to producing the same volume of conventional DRAM.
This economic reality creates a powerful incentive for memory manufacturers to prioritize HBM production. Rather than navigating the uncertainties of forecasting future demand for smartphones and laptops, these companies can secure multi-year supply agreements with some of the world’s most financially robust technology corporations. Jaejune Kim, Samsung’s Executive Vice President of Memory, stated on a recent earnings call, "We have been engaging in discussions with customers, prioritizing those who can guarantee committed future captive demand."
Simultaneously, the proliferation of AI is also driving increased demand for conventional DRAM. Smartphone and personal computer manufacturers are increasingly seeking to implement smaller AI models directly on their devices, which necessitates more sophisticated and higher-capacity conventional memory. Future device iterations are projected to require substantially more memory than their predecessors to effectively support these burgeoning on-device AI capabilities. Song Hyun-jong, President of SK Hynix, observed, "As AI spreads across various services such as search, coding, and productivity tools, the scope of demand is widening from a memory perspective. We are witnessing a structural shift in demand where both AI memory and conventional memory are growing together."
While Samsung and Micron have reiterated their commitment to conventional DRAM, with Bhatia noting it still represents the majority of Micron’s wafer capacity, the confluence of limited supply, escalating demand, and the financial advantages of long-term HBM commitments have made prioritizing AI-related memory production an increasingly lucrative strategy. This shift is reflected in the financial performance of the leading memory manufacturers. SK Hynix reported a record 76% operating margin in the last quarter, a significant increase from 41% the previous year. Micron’s adjusted gross margin reached an impressive 85%, and Samsung’s semiconductor profits experienced a remarkable surge of approximately 250-fold compared to the prior year. Naranjo commented, "It’s more profitable. The three big memory guys, they’re just basically allocating the capacity that they have to these companies."
The impact on consumer pricing is stark. Counterpoint estimates that DRAM prices for smartphones experienced a roughly 56% increase in the first quarter of 2026 compared to the previous quarter, followed by an approximate 83% rise in the second quarter. For a 16GB DRAM module intended for smartphones, Counterpoint estimates the cost escalated from approximately $42 in the second quarter of 2025 to about $181 a year later—a more than 300% increase. While major players like Apple, with their considerable purchasing power, may not face these exact figures, these data points illustrate the dramatic escalation in component costs for high-end mobile devices.
The apparent solution to such supply constraints is to increase production capacity. However, the construction of new semiconductor fabrication facilities is an undertaking that requires years of meticulous planning and execution. Micron’s ongoing endeavors provide a tangible illustration of the scale and complexity involved. In July, the company initiated the foundational work for its planned manufacturing complex near Syracuse, New York. Upon completion, this facility is slated to feature 2.4 million square feet of cleanroom space, positioning it as the largest semiconductor manufacturing site in U.S. history by this metric and Micron’s most expansive facility globally. The cleanroom areas, crucial for chip fabrication, will constitute only a portion of the overall construction, which spans an area equivalent to approximately 350 football fields. Bhatia elaborated, "By the time we’re done, we’ll have built probably 15 to 20 million actual square feet of building space to support the 2.4 million square feet of cleanroom." This extensive non-cleanroom space is dedicated to housing the complex infrastructure required to support chip manufacturing, including extensive water treatment, power generation, air handling, and utility distribution systems.
The timeline for such a project is substantial. Even with accelerated progress, the initial stages, from permitting to breaking ground and pouring the first concrete, can span several years. The physical construction itself is typically divided into three broad phases: site preparation (six to nine months), building the structural shell and exterior (approximately 18 months), and the most intricate phase, the installation of the vast network of mechanical, electrical, plumbing, and piping systems. These systems are essential for transforming a large industrial building into an environment of unparalleled precision, where even a single microscopic particle can compromise the integrity of a microchip. Bhatia aptly described the required environment: "Imagine trying to build all of New York state and not have any ants anywhere in the entire state. That’s what we’re trying to do on every one of these wafers, and we do it over and over again."
The completion of the physical structure is merely a prelude to the complex process of equipping and activating the facility. Micron anticipates activating and testing critical systems such as power, water, and ventilation by the end of 2028, followed by a period in 2029 dedicated to installing manufacturing equipment and commencing pilot production. Meaningful output is not expected until 2030, signifying a total project duration of roughly six years from initial permitting to full production. While Micron’s Idaho facility is further along and projected to begin wafer output in mid-2027, the company is making substantial investments, more than doubling its capital expenditures to exceed $25 billion this year, to expedite these crucial expansion efforts.
Despite these significant investments, pinpointing a definitive end to the current supply-demand imbalance remains challenging, as demand continues to grow at a pace that rivals manufacturers’ capacity expansion. Bhatia stated, "We’ve said that right now it’ll be beyond 2027, and we don’t see when it closes just because demand continues to grow so fast." Industry analysts concur, with Counterpoint projecting that capacity will not align with demand until late 2027 or early 2028 at the earliest. IDC similarly forecasts that the shortage will persist through 2027 and well into early 2028 before any significant relief is realized.
The escalating component costs present a formidable challenge for personal computer and smartphone manufacturers. The sustained multiple-fold increases in memory costs cannot be indefinitely absorbed. Companies face difficult choices: pass these costs directly to consumers through price hikes, reduce production volumes, or integrate less memory into their devices. Alternatively, they may shift their focus towards higher-priced, premium devices that offer sufficient profit margins to offset the increased component expenses. This strategic pivot is already evident across the industry. Microsoft, citing soaring memory costs, recently increased the prices of its Xbox consoles and some Surface Pro models. Meta also raised the price of its Quest 3 VR headset.
Apple, often considered a benchmark for its ability to navigate market fluctuations, is a particularly revealing case. Its considerable scale grants it significant leverage with suppliers, and its customer base has historically demonstrated a willingness to pay premium prices. Nevertheless, Apple has already implemented price adjustments across its Mac and iPad lineups. The focus now shifts to the iPhone.
Current market intelligence suggests Apple may strategically adjust its product release cadence, potentially introducing the iPhone 18 Pro and Pro Max alongside its inaugural foldable iPhone this fall, while deferring the launch of the standard iPhone 18 and a refreshed Air model until spring 2027. This approach concentrates the initial fall release on the higher-end, more profitable segment of the market, allowing Apple to better protect its margins. Supply chain tracking indicates Apple is preparing approximately 10% more units of the iPhone 18 Pro and Pro Max for the September-December period compared to the iPhone 17 Pro and Pro Max during the same timeframe last year.
During Apple’s most recent earnings call, the company cautioned of slower growth, partly attributed to intensifying supply constraints, including those affecting memory chips. Higher memory costs were directly cited as a factor impacting gross margins. Former CEO Tim Cook described the memory pricing situation as a "100-year flood," characterized by "exponential increases" that necessitated price adjustments. Projections from The Wall Street Journal, based on component cost estimates, suggest the iPhone 18 Pro could launch with a starting price of $1,299, a $200 increase over the iPhone 17 Pro. While Samsung and Google have also raised prices on some new phone models, these often include increased storage configurations.
Apple’s robust ecosystem, loyal customer base, and tightly integrated product offerings position its iPhones as more akin to luxury goods than interchangeable electronics. This differentiation has allowed Apple to gain market share by maintaining stable pricing while competitors have raised theirs. In markets like China, the world’s largest smartphone market, premium iPhones have become a near-default choice as lower-cost alternatives approach comparable price points.
For other manufacturers, the market pressures are more acute. Industry analyses indicate that many smartphone companies are already reducing projected shipment volumes while pivoting towards higher-priced models. This strategy aims to maximize revenue per unit, even if overall sales volumes decline. Consequently, the market may see an increase in devices featuring advanced OLED displays, integrated AI functionalities, and larger memory configurations, all designed to justify elevated price points. This trend could lead to flat or even increased revenue for the PC and smartphone sectors, despite a potential decline in unit sales. In essence, the memory shortage, exacerbated by the demands of AI, is not merely increasing the cost of electronics but is fundamentally reshaping which types of devices and companies can economically thrive in the market.
The long-term outlook suggests that increased memory supply will eventually emerge. Samsung, SK Hynix, and Micron are all making substantial investments in expanding their manufacturing capabilities. However, the multi-year timelines required for new facilities mean that immediate relief from the current shortage is unlikely. SK Hynix, for example, is investing heavily in its Yongin Semiconductor Cluster, accelerating its completion timeline. Similarly, Samsung and SK Hynix plan to establish multiple new memory chip factories in South Korea. Micron’s significant capital expenditure in the U.S., encompassing both New York and Idaho facilities, underscores the global scale of these expansion efforts.
A potential fourth major player in China, CXMT, is also developing new fabrication plants. While its technological capabilities are not yet on par with the leading global manufacturers, Apple is reportedly evaluating CXMT chips, although regulatory approvals would be necessary for such a partnership. An alternative scenario for supply-demand normalization could involve a contraction in demand, particularly if the current AI investment bubble were to burst. Such a development would fundamentally alter the economic landscape for memory manufacturers.
However, for the foreseeable future, the market dynamics appear more predictable. Long-term contracts secured by AI companies with memory suppliers provide a degree of certainty for production planning. Industry projections indicate that while the rapid pace of price increases may decelerate to single or low double digits, prices are unlikely to revert to pre-shortage levels. Instead, a "new normal" is expected, with prices stabilizing at levels significantly higher than those seen in 2025. The ongoing construction of new fabs, the sustained demand from AI data centers and increasingly sophisticated consumer devices, and the economic incentives favoring high-margin AI applications over lower-margin consumer electronics all contribute to this outlook. As Naranjo succinctly put it, "It’s not a happy ending anytime soon."
Safa Marwah
Ayu Aulia
