How the AI boom is making phone prices soar

How the AI boom is forcing tech prices higher Beneath the surface, smartphones contain thousands of individual parts. Lurking behind the screen, the battery, the speakers and the camera, a single chip is the main thing keeping your phone running. This is the dynamic random-access memory chip (DRAM), the short-term working memory inside your phone. It's what lets you switch quickly between apps while your phone processes information. And if you're looking to buy a new phone, this chip could be the reason it's about to cost you a whole lot more. The rise of AI data centres is making memory chips a scarce commodity and consumers are paying the price. The memory problem does not stop with your phone. Those same DRAM and NAND chips also live inside your laptop, desktop computers and gaming consoles. And right now, there are not enough of them to go around. Last year, NAND and RAM memory were both incredibly cheap. At one point, a 16-gigabyte RAM chip cost as little as $4.76. But that began skyrocketing in late 2025 and that same chip could cost you almost $60 as of June, according to the WA Data Science Innovation Hub (WADSIH). The budget smartphone market is the latest victim of the global shortage, with multiple tech bosses saying memory chips are now their most expensive component. Samsung has already hiked the price of its phones and its tablet line-up, making devices with higher storage up to $115 more expensive. Microsoft, Google and Apple have also raised the price of several different phones, computers and tablets over the past few months. In February, video game developer Valve announced it was pushing back the launch of its newest hardware due to uncertainty about memory. Then in June, it revealed its Steam Machine would cost $US1,049 ($1,511), hundreds of dollars more than the pre-shortage price estimated by several industry outlets. Smart TVs, digital cameras and smart home devices could be the next to face price hikes. Data centres eat the chips business There's also a third type of memory involved in the market: high bandwidth memory, or HBM. This is the specialised, high-speed type of memory used to power artificial intelligence (AI) servers. Inside an AI data centre, HBM allows an AI system to process large amounts of data very quickly, training it even faster. But AI systems need a lot of HBM chips to keep running and memory manufacturers are trying to keep up with skyrocketing demand. Alex Jenkins is the director of the WA Data Science Innovation Hub and says the memory supply chain has been deliberately redirected. "There's only three companies in the world that really make DRAM, they control over 90 per cent of the market," Mr Jenkins says, referencing Micron, SK Hynix and Samsung, who also make high bandwidth memory. As well as needing a different kind of memory, AI also burns through it at a much faster rate, requiring thousands of servers to train and keep running. "When we train an AI, we don't use one AI server. We might use 10,000 AI servers," Mr Jenkins says. "Services like ChatGPT, in order to train that AI, we're looking at entire data centres, 200,000 AI chips all working together." The sheer amount of chips needed, Mr Jenkins adds, makes it much more profitable for manufacturers to shift to producing HBM RAM for AI companies. "Essentially all these memory manufacturers decided, 'Well, why am I even bothering making DRAM, which I'm selling at near cost price, when I can make HBM RAM, and make a very healthy margin and make lots of money that way'." Priced out of digital worlds Laura Alice Bracken is a Geelong-based artist and creative technologist, who uses programming and software in her work. Sometimes that means transforming energy generated by moss into an interactive animation. Other times, she's creating a "cyberdeck" — a custom-built, portable computer — set into a vintage decorative seashell. Access to the internet means people can teach themselves tech and animation, but the memory shortage is changing things in a way that's "frustrating", she says. "About two years ago, I was thinking about upgrading my GPU [graphics processing unit]. I was starting to get bigger renders, bigger jobs and I wanted to have something that could push it. "I started to save for it, and then … I just started watching the prices go up and up and up and up. "And I was like, 'Oh no, I've missed the boat.' And now [that upgrade] is just so far outside of my price range, there's just no way." And it's not just hobbyists who are feeling the pinch. Independent researchers and smaller AI companies are also facing shortages, according to Shuai Wang, a research fellow at the University of Queensland's ielab. "More and more in the computer science field, when we go to conferences, we see that companies [are starting] to make more and more impact," Dr Wang says. "Ten years ago, it was the research institutes making the impact. "But it's not because the research institutes are becoming less smart, it's because they don't have enough resources." Bracken says the soaring prices mean new artists may be priced out of the field. And others are shifting to less high-tech artwork and design, something she thinks is connected to price hikes. "People are starting to realise that these price rises are not necessarily going anywhere [and] they're starting to futureproof their understanding of tech," she says. Some industry think tanks have predicted the memory crunch could last about two years, until more fabrication plants or "fabs" come online. SK Hynix's chairman, Chey Tae-won, took a more dire view, predicting in March the shortage could persist into 2030. "HBM requires a large number of wafers … it is physically impossible to suddenly increase wafers in the short term," he told reporters in California. "Securing more wafers takes at least four to five years." In the meantime, phones and other technology will not only cost more — cheaper, entry-level electronics could go off the market entirely. Low-end products are already "becoming unprofitable", global technology research firm Omdia stated in recent analysis. Many companies, according to Omdia, have returned to cheaper, older screens and cameras to accommodate the price of RAM. And at least one company, gaming console brand Ayaneo, suspended pre-orders for an upcoming release, citing rising memory costs. And there's no incentive for memory manufacturers to change, according to Dr Wang. "It becomes a competition," he says. "All of [the AI companies] want to acquire and reserve those resources, and they actually tend to buy more at higher and higher prices." It will be at least another two years of shortages and price hikes before more capacity comes online, Mr Jenkins says. "It's enormously complex and it takes a long time," he says. Credits - Reporting: Brianna Morris-Grant - Design: Teresa Tan - Development: Ashley Kyd - Share options - Copy link - X (formerly Twitter)