Laptop prices won't come down until 2028 - cover image

Why Laptop Prices Won’t Come Down Until 2028

by

I’ve been watching laptop prices climb since late last year. I’ve written a few warnings about it along the way, and most of them turned out to be too conservative.

Depending on the model, laptop prices in Australia have gone up anywhere from 20% to 100% this year. Machines that sold for $1,500 twelve months ago are now easily $2,200. Some have done worse than that.

So this article is different to the others. This isn’t a warning about the next few months. This is me putting a stake in the ground on where I think this whole thing lands.

My call: laptop prices are not going to meaningfully improve before the second half of 2028. And when they do improve, “improve” won’t mean going back to 2024 pricing. It’ll mean the market finally stops climbing.

I’m not trying to scare anyone. If you need a laptop, you should buy one. But I’d rather you went in with your eyes open than got blindsided at the counter.

Let me walk you through why I’ve landed there.

Table of contents

Quick recap: how we got here

If you’ve read my earlier pieces on how AI is pushing up the cost of computers and parts, you know the shape of this.

The short version:

  • AI data centres need enormous amounts of memory
  • The memory that goes into AI servers earns manufacturers far more per wafer than the memory that goes into your laptop
  • Samsung, SK Hynix and Micron have shifted production toward the profitable stuff
  • Everything else, the DRAM in your laptop and the NAND in your SSD, is fighting over what’s left

That’s it. That’s the whole story. It’s not a factory fire or a pandemic. It’s a business decision, and it’s been repeated every quarter for two years.

What’s actually happened to laptop prices

Most coverage of this talks about memory chip prices. That’s the cause, but it’s not what you experience at the counter.

So here’s what I’m actually seeing on the ground in Australia this year.

Laptop prices are up anywhere from 20% to 100%, depending on the model.

I can’t name specific models or retailers, for reasons that’ll be obvious. But I can show you the shape of it.

Type of machineRoughly a year agoRoughly nowIncrease
Budget consumer$599$1,09983%
Entry mainstream$899$1,39956%
Business$1,800$2,70050%
Mid-range consumer$1,500$2,20047%
Gaming$2,499$3,29932%
Premium$2,999$3,89930%

Two numbers matter more than the table itself:

  • Nothing has gone up less than 20%. There is no corner of the market that got off lightly. Not one.
  • Around 50% is the most common increase I’m seeing. If you want a single number in your head, that’s the one. Not the 20% floor, and not the 100% worst case. Half again on what it used to cost.

The dollar increase and the percentage increase point in different directions

The premium machine went up $900, which is nearly twice the budget machine’s $500 in actual money. But in percentage terms the budget buyer copped nearly three times the hit.

Whichever way you cut it, someone is hurting.

Why the budget end got hammered hardest

On a $600 laptop, the RAM and SSD might have been a quarter of the build cost. When memory prices roughly triple, there’s nowhere for that increase to hide. There’s no premium chassis to cheapen, no fancy display to downgrade, no fat in the margin. It’s already bare bones.

So the price has to move, and it moves a long way in percentage terms even though the dollar figure isn’t enormous.

The other driver at that end is stock turnover. Plenty of models were sitting in Australian warehouses bought under old, cheap contracts. When that stock ran out and got replaced at current pricing, the shelf price didn’t creep up. It jumped. Same model number, same spec sheet, near enough to double the price.

That’s the situation where a customer walks in with a printout from three months ago and genuinely doesn’t believe me. I don’t blame them.

Business laptops have copped it badly too

This one surprised a lot of people, including some of the resellers I talk to.

Business machines carry more memory and more storage than consumer equivalents, and buyers of business gear expect 16GB or 32GB as standard rather than the bare minimum. That means more of the expensive component in every unit.

Some business lines have gone up more than others, depending on the brand and how their supply contracts were structured. But 50% is common, and I’ve seen worse on specific configurations.

If you’re a small business owner who budgets a fleet refresh every three or four years, that number is going to sting.

Premium and gaming held up relatively better, but not well

Twenty to forty per cent is still a serious increase. A gaming laptop that was $2,499 is now well past $3,000, and that’s before you account for GPU pricing pushing in the same direction.

The reason these held up better is that memory is a smaller share of an expensive machine, so a tripling in memory cost moves the total less in percentage terms. Those brands also tend to have deeper supply contracts locked in further ahead, and frankly some of those machines had margin in them to absorb part of the hit.

But “better” here means 30% instead of 80%. Nobody’s celebrating.

What this really means

Every buyer has been hit. Where you sit in the market changes the shape of the pain, not whether you feel it.

  • Budget buyers copped the biggest percentage increase, and they’re the least able to absorb it
  • Business buyers copped a big percentage increase on machines they buy in quantity
  • Premium and gaming buyers copped the biggest increase in actual dollars

The ugliest part is the first one. The people with the least money to spend have absorbed the biggest proportional increase.

The component numbers behind it

If you want the evidence behind those retail increases, here it is.

TrendForce originally forecast conventional DRAM contract prices to rise 55 to 60% in the first quarter of 2026. They then revised that upward to 90 to 95%, and NAND Flash from 33 to 38% up to 55 to 60%.When the actual first quarter numbers came in, conventional DRAM contract prices had risen roughly 93 to 98% quarter on quarter.

That’s contract pricing, which is what the big brands pay. It reaches retail with a lag of a few months, which is exactly the pattern you’ve been watching all year.

You can see the same thing in parts pricing. The PC Doctor tracked Australian retail and found the cheapest 32GB DDR5 kits at Centrecom, PCCG and Scorptec jumped 38% in a single month back in January, going from around $499 to $689.

GPUs copped it too. The average price of the cheapest RTX 5090 cards in Australia went from $4,832 in November 2025 to $5,566 in January 2026, and mid-range cards have climbed steadily. That’s why gaming laptops are getting squeezed from two directions at once.

To put the memory situation in perspective, DRAM and NAND used to make up somewhere between 10 and 18% of a notebook’s total build cost. Based on what I’m seeing through distribution, it’s well north of that now. On some machines the RAM and SSD are the single biggest line items in the build.

That’s a fundamental change in how a laptop gets priced, and it’s why the machines carrying the most memory copped the biggest percentage rises.

The rest of 2026

There is one encouraging sign, and it’s worth being honest about.

TrendForce’s third quarter outlook has DRAM contract price increases moderating to 13 to 18%, down from the brutal first quarter numbers, with NAND slowing to 10 to 15%. The reason given is blunt: consumer buyers in PCs and smartphones have reached their affordability limit.

Read that again. Prices aren’t slowing because supply improved. They’re slowing because people stopped buying.

So the rate of increase is flattening. But it’s flattening off a base that’s already roughly triple where it was, and the direction is still up.

Here’s what I expect through the back half of this year:

Further price rises on mainstream, business and gaming laptops. The mid-range is where the volume is, and it’s where the increases keep landing. Business lines and gaming machines both carry more memory and storage than an entry-level unit, so they carry more of the cost.

Fewer genuinely cheap laptops. Not “cheap laptops get worse”. Fewer of them existing at all. If a brand can’t make money on a sub-$700 machine, they simply won’t build it. IDC put it plainly: (cite index=”5-1″>”The era of bargain-priced PCs and tablets is behind us for now,”</cite> according to research manager Jitesh Ubrani.

Fewer sales and less discounting. This is the change most people will actually feel. When stock is tight and margins are thin, nobody discounts. The end of financial year and Black Friday runs are going to look very ordinary compared to a few years back. Retailers will still run “save $X” tickets, but they’ll increasingly be measured against inflated reference prices rather than real savings. I’ve covered that pattern in my piece on laptop price gouging, and it’s about to get worse.

Corners cut on some models. Selective, not universal. You’ll see it in the places buyers don’t check: a dimmer panel, a smaller battery, plastic where there used to be metal, single channel RAM instead of dual channel. Consumer Reports spoke to TrendForce about this, and their analysts noted it wouldn’t only be memory that gets trimmed.

Some 8GB machines at the very bottom. I don’t think this becomes a mainstream trend. Most manufacturers know 8GB is a bad experience in 2026 and it’ll come back on them in reviews and returns. But on the cheapest units, where the choice is 8GB or no product at all, you’ll see it.

That single channel point deserves its own mention, because it’s the cheapest corner to cut and the hardest to spot. A machine with 16GB in a single stick performs noticeably worse than the same machine with 2x8GB. I’ve written up why dual channel RAM matters if you want the detail. Check the spec sheet before you buy.

The market is shrinking, and that’s part of the story

This is something I don’t see covered much in consumer coverage, but it matters.

2026 is one of the toughest years the PC industry has had in a long time. Not just in pricing. In actual volume.

IDC now forecasts global PC shipments to decline 11.3% for the full year, with conditions worsening through the fourth quarter, when shipments could fall 20% year on year.

For context, they were predicting a 2.4% decline back in November 2025. The revisions have gone one direction all year.

Unit shipments are down 32 million units, from 284.7 million in 2025 to a forecast 252.5 million in 2026. But here’s the kicker: the total market value still goes up, expanding 1.6% to $274 billion, because prices have risen so much. IDC has PC average selling prices rising 18.3% in 2026.

Fewer machines, sold for more money. That’s the market right now.

What I’m seeing in Australia specifically

Global forecasts are one thing. Here’s what my contacts at the major distributors are telling me, and what I can see in our sales data.

Australian laptop unit sales are down somewhere in the range of 10 to 20% so far in 2026.

That’s not a published figure and I can’t link you to a report on it. It’s what I am seeing in my sales, and what I’m hearing from people who move the stock, and it’s consistent across more than one distributor. Treat it as a well-informed estimate rather than gospel.

What I find interesting is how closely it tracks IDC’s global number. Two completely different sources, one a global research firm and one a bloke speaking to his mates in the channel, landing in the same place. That gives me a fair bit of confidence in both.

Why should you care as a buyer? Because a shrinking market changes retailer behaviour. Less competition for your dollar, less reason to discount, less range on the shelf, and more pressure on staff to move whatever’s in stock rather than what actually suits you. Plan accordingly.

Where MacBooks fit

I’ve been a Windows bloke my whole career, so take this as a genuinely reluctant observation.

MacBooks are holding up as good value right now, and I think they’ll keep doing so through this crunch.

That’s not because Apple has been spared. They haven’t. As I covered in my piece on the July 2026 laptop price rises, Apple’s Australian distribution pricing went up 17 to 19% on MacBooks. That’s a real increase and I’m not going to pretend otherwise.

But go back to the table above. Nothing in the Windows market went up less than 20%, and 50% was the typical figure.

Apple came in under the floor of what everyone else copped.

Two things work in their favour:

  • They buy memory at enormous scale on long contracts. Apple locks in supply years ahead in ways smaller Windows brands can’t. That smooths the shocks out.
  • Their baseline configs haven’t been trimmed. While Windows brands shave specs to hold price points, the entry MacBook Air still ships with the same unified memory it did. In a market where everyone else is quietly giving you less, standing still is a competitive advantage.

The result is that the gap has closed a long way. A MacBook Air was never the value pick against a mid-range Windows laptop. Now, against a Windows machine that’s gone from $1,500 to $2,200 and had its display or memory configuration trimmed, the comparison looks completely different.

I’m working on a proper head to head on this, because it deserves more than a few paragraphs. For now: if you were on the fence about switching, the economics have moved in Apple’s direction big time, and they’ll probably stay there until this crunch resolves.

2027: the gap gets wider before it gets better

This is the year most people are quietly hoping for relief. I don’t think they’re going to get it.

TrendForce’s July outlook has DRAM staying structurally tight right through 2028, with the supply and demand gap actually widening further in 2027 before any reversal.

The reason is simple. New factories take years, and the ones being built now don’t produce anything useful in 2027.

New fab capacity arriving in 2027 won’t produce meaningful output until the second half of that year, and substantial contributions aren’t expected until 2028.

Take Nanya, one of the smaller Taiwanese players. TrendForce reported their Fab 5A plan in August. Wafer starts begin in the second half of 2027, monthly output reaches 30,000 wafers in 2028, then 35,900 in 2029.

Sounds promising until you look at the history. That fab was announced back in April 2021 with mass production planned for 2024. Five years later it still hasn’t produced a thing.

That’s the pattern across the industry. Announcements are fast. Silicon is slow.

The other half of the equation: demand isn’t slowing

Everything above is about supply. Factories, wafers, timelines. That’s the half of this story everyone writes about.

But supply only matters relative to demand, and this is where I think most of the coverage misses the point.

New capacity arriving in 2028 doesn’t help you if demand has grown faster than the capacity being added. And right now, that’s exactly what’s happening.

Look at what the big players are spending

According to figures compiled by the Financial Times and reported by Tom’s Hardware, Google, Amazon, Microsoft and Meta collectively plan to spend $725 billion on capital expenditure in 2026, up 77% from last year’s record $410 billion.

Sit with that for a second. Not $725 billion in total. $725 billion in a single year, up more than three quarters from the year before.

Google alone went from around $91 billion in 2025 to guidance of $175 to $205 billion for 2026. That’s roughly a hundred billion dollar increase from one company in one year.

And it’s still climbing. Nvidia’s CFO cited third party forecasts of hyperscale capex topping $1 trillion in 2027 and total AI infrastructure spending reaching $3 to $4 trillion a year by 2030.

Here’s the stat that tells you everything

Microsoft’s CFO attributed $25 billion of its record capex budget to rising memory chip prices.

One company. One year. Twenty-five billion dollars of extra spending, purely because memory got dearer. And they paid it. They didn’t reduce their order, they didn’t wait for prices to come down, they didn’t downgrade the spec.

Now think about a laptop manufacturer trying to buy memory in that market. You’re bidding against a customer who will absorb a $25 billion cost increase without blinking and keep buying.

You’re not going to win that. Neither is the brand that builds the laptop you’re looking at.

They’re already planning for 2028

This is the bit that shifted my thinking on the timing.

Amazon’s CEO said capacity constraints are likely to persist through 2027, with projected demand in 2028 already informing infrastructure planning.

So the companies eating all the memory supply aren’t treating this as a temporary spike they’ll ride out. They’re building 2028 into their forecasts and buying accordingly.

What that means for the fab timeline

Go back to those new factories coming online in 2028. They add capacity, which is good. But they add it into a market where the biggest buyers have already grown their spending 77% in a year and are forecasting more.

New supply arrives. Demand has already grown to meet it. Consumer memory gets whatever’s left over, same as now.

That’s why I keep saying prices will ease rather than crash. The supply side improves in 2028. The demand side improves right along with it.

2028: relief, but not until the second half

Here’s where I land on the timing, and I want to be specific, because “2028” on its own is too vague to be useful.

First, when the capacity actually arrives. IEEE Spectrum’s analysis, summarised by PC Gamer, lays out the timeline. Micron gets a Singapore fab into production in 2027, but it’s making HBM for AI, not DRAM for PCs. Micron’s actual new DRAM factory in New York isn’t in full production until 2030. SK Hynix has facilities slated for late 2028. Samsung’s Pyeongtaek fab is due in 2028.

So the earliest the physical supply picture improves is sometime in 2028.

IDC’s Jitesh Ubrani has said the same from a different angle: a decrease in memory prices is possible no earlier than 2028, and the market is unlikely to return to 2025 price levels.

That’s an analyst at one of the two biggest research firms in the industry landing on the same year I did, from a completely different direction. Good sign.

But “no earlier than 2028” is the floor, not the answer. Here’s why I think it’s the back half of that year specifically.

There are four lags stacked on top of each other:

  1. Fab ramp. A fab that starts producing in early 2028 doesn’t hit meaningful volume for months. Nanya’s own schedule has them reaching 30,000 wafers a month across the whole of 2028, not on day one.
  2. Demand absorption. As I covered above, the hyperscalers are still growing their spending and already planning around 2028 demand. A good chunk of that new capacity is spoken for before it exists.
  3. Contract pricing. Brands buy memory on quarterly contracts negotiated ahead of time. Even once spot prices soften, contract pricing takes a quarter or two to follow. We’ve watched this work in the other direction all year.
  4. Retail. What the brands pay takes another three to six months to reach a shelf in Australia. Add freight, local margin, and the fact that nobody drops a retail price the day their cost drops.

And plenty of people think even that’s optimistic. Some assessments put a genuine return to normal pricing at 2028 to 2029, and only if AI infrastructure demand grows at forecast rates instead of accelerating. SK Hynix has publicly floated the shortage running past 2030.

So when I say “second half of 2028 at the earliest”, I’m sitting at the optimistic end of the range.

Why this isn’t like past shortages

This is the bit I want you to take away, because it’s the difference between this crunch and every other one I’ve lived through.

In a normal shortage, demand outstrips supply, manufacturers build more capacity, supply overshoots, prices crash. That cycle has run four or five times in my career. It’s why people say “just wait, memory always gets cheap again.”

This one is different, for two reasons.

The bottleneck isn’t capacity, it’s choice. Manufacturers could make more consumer memory. They’ve chosen to make a different kind, because that kind earns them a lot more money. New factories don’t change that decision. If AI demand is still strong in 2028, a good chunk of that new capacity gets pointed at AI too.

The demand isn’t finite. In past shortages, demand came from a market that eventually got its fill. This time the buyers are spending three quarters more each year and telling investors they can’t build fast enough. There’s no natural ceiling in sight.

That’s why I don’t buy the argument that prices will crash back to pre-AI levels. They’ll ease. They won’t reset.

There’s a third, quieter reason too. The floor has moved. Windows and modern software expect more memory than they did in 2023. The industry wanted 16GB to become the standard baseline, and this crunch has stalled that shift rather than reversed it. Once supply loosens, that shift resumes, and it costs money.

One bit of good news

Here’s something almost nobody is talking about, and it actually matters for buyers.

DRAM and NAND are about to split apart.

The same TrendForce outlook that has DRAM tight through 2028 has NAND Flash flipping from undersupply to surplus by the second half of 2027.

NAND is the flash memory in your SSD. DRAM is your system RAM.

If that holds, it means:

  • Storage gets cheaper before RAM does
  • A laptop with a big SSD but modest RAM becomes relatively better value
  • Machines with a spare or upgradeable M.2 slot get more attractive, because you can add cheap storage later
  • Soldered LPDDR5X memory stays expensive the longest

That last point is worth thinking hard about. Almost every thin and light laptop now has soldered RAM. Whatever you buy is what you’re stuck with for the life of the machine, and you’re paying peak pricing for it.

What this means if you’re buying

If you need a laptop now

Buy it. Don’t wait for a correction that’s the better part of two years away. A laptop you needed in 2026 that you finally buy in late 2028 has cost you two years of being unproductive.

Just buy smarter:

  • Prioritise RAM over storage. RAM is the thing you can’t fix later and the thing that stays expensive longest.
  • 16GB minimum. Don’t let a $200 saving talk you into 8GB.
  • Check dual channel versus single channel before you buy.
  • Prefer a machine with a spare or upgradeable M.2 slot.
  • Compare specs, not price brackets. The model number carried over. The spec sheet may not have.
  • Don’t hold out for a sale. Genuine discounting is drying up. If the price is fair today, that may well be the best price you see.
  • Be especially careful at the budget end, because that’s where the increases have been steepest and the corner-cutting heaviest.

If your current laptop still works

Keep it. A clean out, a fresh install and a bit of care will buy you another year or two. If your machine takes an upgradeable SSD, you might get more life out of it cheaply once NAND loosens up in late 2027.

If you’re a student

Buy at the start of the academic year and buy for four years, not one. My student laptop buying guide covers the specs that actually matter.

If you’re buying for a business

This bracket has been hit harder than most people realise, with 50% increases common and some lines worse. Lock in pricing where you can, stretch your refresh cycle, and get quotes early rather than assuming last year’s budget still works. My business laptop buying guide covers the warranty and support side, which matters more than usual when you’re keeping machines longer.

If you’re a gamer

Toughest bracket right now, because you’re getting hit by memory pricing and GPU pricing at once. Expect 20 to 40% on the machine itself with GPU costs pushing in the same direction. My gaming laptop buying guide goes through where to spend and where to save.

If you’re open to a Mac

Have a proper look. I don’t say that lightly. The value equation has genuinely shifted and it’s likely to stay shifted for a couple of years.

What I’d actually do

If it were my money, right now:

  1. Buy the machine you need, not the machine you want. Premium features are where the price rises hurt most in dollar terms.
  2. Spend on RAM, save on storage. Storage relief is coming sooner.
  3. Check the real market price before you believe a sale ticket. Genuine discounts are getting rarer, which means the inflated “save $400” tickets are getting more misleading, not less. StaticIce and Google Shopping take thirty seconds.
  4. Ask specialist retailers for a price match. Centrecom, Scorptec, PC Case Gear and JW Computers will usually match a legitimate advertised price, and you get better advice into the bargain.
  5. Don’t wait it out. By the time relief arrives in late 2028, the machine you’d have bought today will be three generations old.

Final word

I could be wrong. If AI infrastructure spending slows sharply, supply frees up faster than anyone expects and I’ll happily write the correction.

But look at what that would take. It’d mean Google, Amazon, Microsoft and Meta collectively pulling back from three quarters of a trillion dollars a year in spending, and doing it soon enough to matter. Nothing in their guidance suggests that’s coming. If anything they keep revising upward in my opinion.

I’ve been reading the fab timelines and the contract pricing for two years now, and I’ve been standing behind a counter watching customers react to the tickets. The factories that would fix this aren’t finished. The ones being finished are aimed at AI. The buyers eating the supply are spending more every quarter. And the companies making the decisions are having the most profitable run in the history of the memory business.

That doesn’t turn around in twelve months. It probably doesn’t turn around in twenty-four.

Buy what you need, buy it well specced, and look after it. That’s the best play available between now and the back half of 2028.

I’ll keep tracking this and update as the quarterly numbers land.

Leave a comment