The artificial-intelligence boom has opened a new front in Washington’s semiconductor debate. As data centers absorb increasing quantities of high-speed memory chips, companies that rely on conventional memory for cars, medical equipment, broadband networks, smartphones and computers are pressing the U.S. government for help.
This is no longer a problem for technology firms or semiconductor producers. This is an economic and geopolitical dispute, related to prices and industrial policy, the rivalry between the U.S. and China, and the future of America’s semiconductors industry chain. In the heart of the dispute, there is the lack of dynamic random-access memory, which is known as DRAM, and high-bandwidth memory, which is known as HBM.
The latter is necessary for artificial-intelligence chips utilized by large data centers, whereas DRAM is essential for PC, smartphone, automobile, medical equipment, and telecommunication applications. This shortage has sparked a rather extensive lobbying effort in Washington. Telecommunication industry associations, makers of medical devices, automobile firms, and retailers asked the Trump administration to act. Meanwhile, Apple is trying to obtain approval to work with ChangXin Memory Technologies, or CXMT, a Chinese memory chip maker.
AI demand reshapes the memory market
The semiconductor industry has traditionally experienced cyclical shortages. Companies expand production when demand rises, prices increase, new capacity enters the market and supply eventually catches up. The current crisis is different because AI data centers are consuming memory at a scale that is altering manufacturers’ investment and production priorities.
AI systems need large amounts of memory to train models, process data and deliver responses. As models become larger and more complex, their servers require more HBM alongside conventional DRAM and solid-state storage. The result is not simply higher demand for one specialist product. It is pressure across the entire memory ecosystem.
The focus on HBM technology is increasing since the margins are higher and the technology is related to the fast-growing infrastructure of AI. This focus is limiting the supply of memory products like DDR5 and LPDDR5X, which can be used in computers and smartphones. S&P Global Market Intelligence noted that due to the switch to the production of HBM, the market had become tight concerning the supplies of conventional memory and prices rose. Therefore, large chip producers like Samsung Electronics, SK hynix and Micron Technology have conflicting priorities.
AI companies and cloud services would be happy to secure supplies via long-term contracts. However, traditional companies need access to components of their devices, which have longer development cycles. According to the reports, there is a shortage of HBM until at least 2027 for AI firms. It poses problems for car manufacturers and medical device producers, who cannot change their products quickly.
Prices spread beyond the technology sector
The shortage is increasing costs throughout the economy. J.P. Morgan Global Research estimates that DRAM prices could rise by more than 400% between the beginning of 2024 and the end of 2026. It also estimates that prices for computers, peripherals and related parts had risen 37% since the end of 2024.
The New York Times reported that memory prices had increased approximately fourfold over the previous year, citing 18 people familiar with the market and private negotiations. Many of those sources spoke anonymously because they were not authorized to discuss commercial arrangements publicly.
The impact is already visible in consumer electronics. Apple has raised prices on some products as memory costs have increased. Tim Cook, Apple’s outgoing chief executive, described the market shock as “a 100-year flood on memory pricing.” Cook also said Apple had reluctantly raised prices and warned that the company had not experienced anything comparable during his more than four decades in the industry.
Higher memory prices affect more than the final cost of a smartphone or laptop. They can increase the expense of servers, broadband upgrades, cybersecurity equipment, vehicle electronics and medical systems. Companies may pass those costs to consumers, reduce product specifications or delay production.
J.P. Morgan estimates that every 10% increase in hardware costs could add about 0.1 percentage point to core consumer inflation and personal-consumption-expenditure inflation. It estimates that the memory shock could eventually contribute between 0.2 and 0.4 percentage point to inflation, although those figures are analytical estimates rather than official government forecasts.
Trade groups demand federal action
On June 3, the lobby group gained more traction as nine trade organizations from the United States sent a letter to U.S. Treasury Secretary Scott Bessent and Commerce Secretary Howard Lutnick. These associations comprised representatives of broadband firms, telecommunication providers, producers of medical devices, automotive companies, and retailers.
According to the trade organizations, the growth of memory chip availability was being threatened by the rising demand for the AI data centers, which posed an “urgent imbalance” between demand and supply, resulting in potentially long-term price hikes and disruption of important supply chains in the country. Trade associations suggested that the U.S. government should guarantee that policies related to semiconductors do not favor AI data centers at the cost of other industries.
Their concerns are practical. A broadband provider unable to obtain memory chips may delay network upgrades. A medical-device company may face longer production cycles. An automaker may struggle to secure components for infotainment systems, driver-assistance technology and electronic control units. Retailers, meanwhile, could face higher prices and reduced availability of consumer devices.
The coalition has also called for closer monitoring of supply, demand and prices. Some participants have discussed requiring companies that receive government support to serve a broader range of customers instead of allocating most production to the highest-paying AI buyers.
The proposal would represent a significant expansion of the traditional debate over semiconductor subsidies. The CHIPS Act was designed primarily to encourage domestic production and reduce dependence on vulnerable overseas supply chains. The emerging question is whether government support should also determine how private companies distribute scarce memory capacity.
Industry resists allocation controls
Memory manufacturers and their industry representatives oppose direct government intervention in prices or production allocation. Their argument is that the shortage is best addressed by expanding capacity, not by directing companies to reserve fixed quantities for selected industries.
Semi, which is an association of leading semiconductor manufacturers, has also cautioned that any efforts aimed at controlling prices or production will make the shortage even worse. The industry would rather see tax breaks, long-term contracts, and quicker construction of new factories. This is due to the heavy reliance of semiconductors on investments in capital. Building a new factory is a process that takes years from start to completion. Furthermore, HBM is also harder to produce than regular DRAM because it relies on advanced packaging technologies and strict quality requirements. From the point of view of the producers, assured allocation by the government will decrease the motive to increase production capacities or make customers wait for government action instead of signing the contract.
The downstream industries see the matter differently. They argue that market forces are already favoring AI companies because data centers can pay more and sign long-term agreements. Without federal action, traditional manufacturers may be forced to compete for residual supply at much higher prices.
This is the fundamental policy conflict: chipmakers want freedom to allocate production according to commercial demand, while other industries want Washington to prevent AI companies from effectively monopolizing the supply chain.
Apple’s China dilemma
Apple’s reported effort to source memory from CXMT has added a geopolitical dimension to the shortage. The company has reportedly tested CXMT chips for possible use in iPhones, iPads and Macs, particularly products intended for the Chinese market. Apple has also sought clarity or approval from the U.S. administration before proceeding with broader commercial arrangements.
For Apple, the calculation is straightforward. A new supplier could create bargaining leverage against Samsung, SK hynix and Micron, reduce dependence on a small group of producers and help limit the impact of rising memory costs.
But CXMT is politically sensitive in Washington. The Pentagon has designated it as a Chinese military company under Section 1260H, while lawmakers have argued that Chinese memory manufacturers could benefit from state support and eventually weaken U.S. and allied semiconductor industries.
Apple’s potential use of CXMT therefore raises several questions. Would commercial access to a Chinese memory producer help relieve the immediate shortage? Could it expose U.S. companies to intellectual-property or data-security risks? Would it undermine the objectives of U.S. export controls? And could subsidized Chinese memory eventually displace Micron and other non-Chinese producers?
A bipartisan group of U.S. senators has warned Apple against purchasing memory from CXMT and Yangtze Memory Technologies, or YMTC. The lawmakers have demanded that Apple exclude the companies from its supply chain, arguing that short-term price relief should not come at the expense of national security.
Lawmakers push tougher China restrictions
The legislative reaction is taking a turn in an entirely different direction from what Apple seeks. On July 14, in a letter to Commerce Secretary Lutnick, Rep. John Moolenaar, chairman of the House Select Committee on China, along with Rep. George Whitesides, have encouraged the administration to reinforce limitations on Chinese memory manufacturers. The lawmakers ask to add CXMT to the Commerce Department’s Entity List and impose stricter controls on YMTC. Moreover, the representatives of Congress urge the imposition of restrictions on procurement of Chinese DRAM and HBM for use in AI, data centers, federal IT, and critical infrastructure. The legislators believe that the solution to the supply deficit should not increase the US reliance on a strategic rival.
The lawmakers wish Washington to work together with South Korea, Japan and European states so that Chinese memory makers do not have an opportunity to take advantage of the gaps in the export control system used by allied countries. This approach can be viewed as a reflection of the change in the general U.S. semiconductor strategy. Washington is no longer only concerned about the level of sophistication of the chip.
The restrictions, however, could intensify short-term supply pressure. Blocking Chinese memory would remove a potential source of DRAM and NAND at a time when companies are already struggling to secure components. That could increase prices for U.S. consumers and manufacturers, even if policymakers believe the long-term security benefits justify the cost.
A test for U.S. industrial policy
The memory shortage is becoming a test of whether the United States can pursue two goals at once: accelerate AI development and protect the wider economy from AI’s demand for scarce resources.
Tax breaks or subsidies can also be distributed by the administration in order to promote chip fabrication in the US. The administration may offer more help to allies. It may use its own procurement powers to support vital industries. Restrictions may be imposed on Chinese companies, or conditions may be set for companies enjoying grant support under the CHIPS Act. Each of the above options carries certain risks.
It may take several years for the effect of subsidization to be seen. Regulation of allocations may distort the market. Restrictions on export may aggravate the problem of chip shortage even more. Permission for Chinese producers will offer a fast solution, but at the same time, it will negate the power Washington has while implementing the measures. The Defense Production Act is another possible tool. There is, however, no indication of how the Act has ever been applied to memory chips. The administration should decide whether it deals with an emergency situation or just a market one.
The shortage may last beyond the immediate crisis
The most important feature of the current crunch is its potential duration. New memory factories cannot be built quickly, and HBM production requires specialized equipment and packaging capacity. Meanwhile, AI companies continue to expand data-center construction and compete aggressively for supply.
Samsung and SK hynix have warned that AI-related memory shortages could persist through 2027 as demand for HBM continues to absorb manufacturing capacity.
For consumers, the likely result is higher prices and fewer low-cost devices. For manufacturers, it means longer procurement cycles, more expensive components and pressure to redesign products. For Washington, the shortage presents a political dilemma in which every solution produces a competing economic or security problem.
The lobbying rush demonstrates that AI is no longer merely an emerging technology sector. Its demand for chips is reshaping industrial priorities across the global economy. Whether the United States responds with subsidies, supply guarantees, tougher China restrictions or limited emergency controls will help determine who receives the next generation of memory—and who is left waiting.
