‘The era of the cheap smartphone has ended’: Customers are now paying for the global AI fever

One of technology’s great bargains is quickly turning on its head as the seemingly unstoppable AI tidal wave rolls on.

Over the past 20 years of handheld tech, consumers have benefited from runaway developments in memory (RAM). Our smartphones saw their permanent storage and working memory limits double, largely without the price tag following suit.

Ever since the inception of the smartphone, major competitors like Apple and Samsung have fought tooth and nail to entice buyers with increasingly impressive features.

But the latest tech boom is starting to break that paradigm.

The world’s biggest technology companies are now reserving enormous quantities of high-performance memory for data centres, while the handful of manufacturers capable of supplying it divert factories and investment away.

Tech giants like Microsoft, Amazon, Google and Meta can sign long-term agreements and reserve production before it even exists.

In short, the global AI fever has created a bottleneck forcing prices to explode.

The biggest players are able tosecure supply years ahead and pay almost whatever is necessary, meaning manufacturers of cheaper phones and laptops are being priced out.

Industry tracker TrendForce expects conventional DRAM contract prices to rise another 13 to 18 per cent in the third quarter of 2026. NAND flash — the permanent storage that lives inside our phones and laptops, is set to climb another 10 to 15 per cent.

Those increases are actually slowing compared with the extraordinary rises recorded in the first half of 2026. But it’s not necessarily because factories have caught up. Quite the opposite.

TrendForce says consumer manufacturers have genuinely reached the limit of what they can afford and are now reducing orders.

How AI gobbled up the market

It’s no surprise that building talking computers capable of immediately repackaging the breadth of humanity’s intelligence in seconds need a bit more power.

Training large language models like ChatGPT is a monolithic task that keeps evolving by the day, and companies are accounting for the massive requirements needed to stay in the “AI arms race”.

High-bandwidth memory, or HBM, is built by stacking DRAM beside an AI processor. It moves data quickly enough to keep advanced accelerators fed with information.

Training AI models takes thousands of those accelerators, and then serving it to millions of users requires even more infrastructure, along with conventional server DRAM and vast amounts of storage.

TrendForce estimates HBM and server-grade RDIMM products will consume a combined 51 per cent of global DRAM bit supply in 2026. DRAM and NAND will account for an estimated 47 per cent of major cloud companies’ capital expenditure this year, potentially rising to 68 per cent in 2027.

The numbers show how decisively the market has swung towards data centres, as corporations and governments go all-in on what is being described as a landmark moment for humanity.

Analysis from IDC describes it as a “zero-sum game”, with experts warning this could become a lasting strategic reallocation of the world’s RAM capacity, rather than a normal short-term shortage.

Three companies hold the keys

Almost all advanced DRAM comes from just three businesses. South Korea’s Samsung Electronics, SK Hynix, and America’s Micron.

In the first quarter of 2026, Samsung controlled 38.5 per cent of global DRAM revenue, SK Hynix 28.8 per cent and Micron 22.4 per cent. Together, they held almost 90 per cent of the market, according to TrendForce.

DRAM factories cost billions, take years to build and must achieve extraordinary yields while repeatedly moving to smaller manufacturing processes. One failed transition can wipe out years of investment.

The industry’s brutal boom-and-bust cycles have helped remove most of the competition. German manufacturer Qimonda collapsed in 2009, while Japan’s Elpida went bankrupt in 2012 before being bought by Micron.

A report from McKinsey found that surviving required enormous intellectual-property libraries, sophisticated engineering and enough money to withstand savage downturns.

Manufacturers now have four tough choices: raise prices, reduce memory, downgrade other components or abandon low-margin products.

Omdia says memory represented almost 60 per cent of the manufacturing cost of smartphones priced below $400 ($A558) during the first quarter.

In devices below $US99, it exceeded 64 per cent.

Manufacturers are responding with cheaper displays, smaller camera sensors and older processors. Omdia expects shipments below $400 (A$138) to fall more than 22 per cent.

Gartner forecasts higher memory costs will lift PC prices by 17 per cent and smartphone prices by 13 per cent during 2026 alone. It expects buyers to keep devices longer and predicts the sub-$US500 ($A698) entry-level PC category could disappear by 2028.

“This is the steepest contraction in device shipments witnessed in over a decade,” Gartner senior director analyst Ranjit Atwal said.

“Higher prices will narrow the range of devices available, prompting buyers to hold on to devices for longer, fundamentally altering upgrade cycles.”

Gartner simultaneously expects global PC shipments to fall 10.4 per cent this year.

Meanwhile, IDC’s smartphone outlook has just become even darker. It forecast shipments would plunge 16.7 per cent while the average selling price rose 27.6 per cent to $US581 ($A811).

That does not mean every existing phone will become 27.6 per cent dearer, because much of the increase reflects cheap models disappearing and premium devices taking a larger share of a shrinking market.

Nevertheless, experts fear “the era of the cheap smartphone has ended.”

“The memory tsunami that we warned about is now hitting the market in full, and consumers are starting to pay the AI bill,” IDC vice-president Francisco Jeronimo said.

IDC expects the shortage to keep distorting the market deep into 2028.