The AI boom is coming to your pocket, and it is called RAMpocalypse.
The AI boom, which is about to take over your pocket, is called RAMpocalypse. The next time the phone bill comes in higher than expected, it could be because it is hidden in an unlikely place. Not the camera. Not the screen. Not even the processor. Maybe the memory chip.
The smartphone industry is in an odd bind in 2026. AI's demand for memory has grown significantly, and memory makers don't have enough capacity to produce all at once. The outcome is an issue that's been dubbed “RAMpocalypse” and "RAMmageddon". The names are hyperbolic. The figures are more difficult to shrug off.
IDC forecasts that overall smartphone shipments will reach 1.09 billion units in 2026, the industry's largest annual drop ever. Meanwhile, the average price of a smartphone handset is projected to hit a new record of $550, which is about $100 more than in 2025. The switch may be more apparent to those on a budget.
The current shortage is unlike the semiconductor shortage that occurred during the pandemic. Things went awry back then: factories shut down, transport was disrupted, and demands changed. This time, it's not from the back but from the front, as AI data centres are chewing up vast amounts of memory.
To pass around massive quantities of data at high speed, AI accelerators require high-bandwidth memory, or HBM. HBM is not the same as the LPDDR memory that is found in smartphones, but rather part of a larger memory manufacturing ecosystem. The world's top three memory makers, namely Samsung Electronics, SK hynix, and Micron, have a competitive edge in the AI market because of their financial rewards for increased investment and production investments.
That's a strategic shift in capacity toward higher-margin enterprise memory, IDC said. Other big cloud providers like Microsoft, Google, Meta, and Amazon are also making bids to acquire supplies in their pursuit of AI infrastructure.
Economics is hard for phone manufacturers to stomach. An AI data centre can consume a lot of pricey memory, whereas a budget smartphone needs to stay affordable enough for people to purchase it. This is starting to be reflected in that.
Phone companies have long been rivalling each other by increasing the RAM in the specs. Six gigabytes doubled to eight. Eight became 12. There was limited availability of 16GB on some of the flagship models. Now, manufacturers are taking precautions.
In May, TrendForce noted that mobile DRAM contract prices continued to increase sharply. It predicted that LPDDR4X prices could increase by an average of 70% to 75% from the previous quarter in Q2 '26, and that LPDDR5X prices could grow by 78% to 83%. The prices of these components represent memory costs, not the cost of an actual phone, but they show how much pressure is being placed on manufacturers.
Specifications are already being influenced by the response. According to TrendForce, the 12GB configuration is becoming the standard in high-end devices, and the 16GB version is declining. Mid-range phones seem to be converging around 8GB, while entry-level phones are typically around 4GB.
It sounds like a hidden adjustment by the industry. A one per cent increase in the price of each memory module, however, can rapidly turn into a massive cost for a company that is selling millions of phones.
Here's where the lack becomes more real. There's room to breathe on a premium smartphone. Its manufacturer can either increase the price or cover part of the price increase, or maintain the margin. Budget brands offer much fewer options.
As memory and NAND prices stay high, IDC believes the sub-$100 smartphone market, which accounted for over 170 million devices in 2025, will be "challenging to sustain". This is expected to put pressure on smaller Android manufacturers, which are mainly targeting lower-priced segments.
Apple and Samsung are better off. Both companies are projected to benefit from increased market share in 2026, thanks to size and access to memory supplies, as well as a higher proportion of premium products, according to IDC. The impact was cushioned by Apple, which was able to obtain memory supplies early.
It does not mean that they are untouched, though. The price of memory also took a toll on Samsung's mobile business, as the company itself noted in a post. Samsung's own post noted that it was the high cost of memory that was impacting its mobile business, while its semiconductor division was reaping the benefits of the surge in demand for AI. Samsung has been inking multi-year supply deals with big data centres and anticipates the broad chip shortage will continue through 2028, Reuters reported in July.
The company that's in the middle of the AI boom, experiencing the same plight as its phone business, is seeing the potential of its memory business thrive. PC makers are under the same kind of pressure, as DRAM and NAND are critical components of PCs, too. The implications are even more significant in the technology supply chain.
Cars have more and more advanced electronics, such as driver assistance and infotainment systems. In contrast, AI servers need a significant amount of conventional memory and storage in addition to HBM to handle the increasing workload. Manufacturers just can't construct new factories out of thin air.
SK Hynix recently approved a 54.3 trillion won ($38 billion) investment plan for new South Korean semiconductor facilities. It will have planned factories to manufacture advanced DRAM, such as HBM, and NAND flash. The new cleanroom capacity is expected to become available starting in 2028 and is not expected to be 100% available until 2028 or later.
This explains why the end of the shortage is not expected to be anytime soon for everyone. Memory density will remain constrained through 2026 and possibly even into 2027, as smartphone volume drops in that year and then picks up again in 2028, IDC anticipates.
Samsung's forecast is more conservative, projecting shortages to last through 2028. So it's not really a case of running out of RAM; it's about RAMpocalypse.It is a question of who gets the first pick when the memory of the world isn't growing at a rate to please everyone.
In the case of AI businesses, memory is a vital infrastructure. For phone manufacturers, it's just one of the pieces of the puzzle. These firms are the ones developing the next generation of AI systems and have more future contracts with chip makers in their coffers. The impact trickles down the pipe.
It arrives at the laptop manufacturer. Next, the telephone company. Then the retailer. Finally, the consumer who goes to a shop and sees an iPhone and a competing phone goes to the specs and sees that the cheaper phone has fewer specs. The AI revolution could be taking place in large-scale data centres thousands of miles out of the way.
REFERENCES