The AI boom isn’t just happening in data centers. It’s showing up on the price tag of the laptop in your shopping cart. Here’s the full chain of cause and effect, with the numbers to back it up.
On June 25, 2026, Apple did something it almost never does: it raised prices on products mid-cycle. The MacBook Air jumped from $1,099 to $1,299. The MacBook Pro went from $1,699 to $1,999. The entry-level MacBook Neo climbed from $599 to $699. Tim Cook called the situation a “hundred-year flood” and said he’d never seen anything like it in over 40 years. Apple’s stock fell more than 6% that day — its worst session since April 2025.
Apple wasn’t first. It was late. Dell raised prices 15–20% back in December 2025. Lenovo voided all existing customer quotes on January 1, 2026. HP warned that the second half of 2026 would be its hardest stretch. Nintendo added $50 to the Switch 2 and blamed memory costs by name. Microsoft tacked $100–150 onto Xbox consoles.
Behind every one of these announcements is the same culprit: artificial intelligence — or more precisely, the data centers being built to run it. Here’s exactly how a server farm in Virginia ends up costing you money at checkout.
Why Laptop are getting more Expensive: The short version
AI companies are buying so much memory that there isn’t enough left for laptops. Memory (RAM and SSD storage) used to be about 16–20% of a laptop’s manufacturing cost. In 2026, it’s roughly 35%. DRAM contract prices nearly doubled in Q1 2026 alone — up 90–95% quarter over quarter, a record — and rose another 43–63% in Q2 depending on the product type. NAND flash, the stuff inside your SSD, jumped 70–75% in Q2, actually outpacing DRAM.
When the two components that make up a third of a laptop’s bill of materials double or triple in price within a year, the sticker price follows. Analysts at TrendForce calculated that at a $900 average retail price, laptop brands would need to raise prices at least 30% just to maintain existing margins.
AI data centers need staggering amounts of memory
Every AI chip — an Nvidia GPU, a Google TPU, an AWS Trainium — is bolted to stacks of high-bandwidth memory (HBM). HBM is what lets these chips read and write data fast enough to train and serve large language models. No memory, no AI.
Large US tech companies are expected to spend around $650 billion on AI infrastructure in 2026. That capital expenditure translates directly into memory orders. According to TrendForce, data centers will consume over 70% of top-tier memory chips in 2026 — and would take more if they could get it.
Cloud providers aren’t just buying more; they’re buying differently. Hyperscalers have locked in multi-quarter and multi-year purchase agreements, paying premium prices to guarantee allocation. Micron announced $100 billion in contracted revenue across 16 long-term customer agreements, and its entire 2026 HBM output is already sold out. When the biggest buyers on Earth pre-purchase supply years ahead, everyone else — including the companies that build your laptop — fights over the leftovers.
Memory makers abandoned the consumer market
Three companies — Samsung, SK Hynix, and Micron — control the vast majority of global DRAM production. All three have shifted wafer capacity away from the ordinary DDR5 and LPDDR5 memory that goes into laptops and toward HBM and server-grade DRAM, because that’s where the margins are.
The economics are brutal for consumers. HBM isn’t just more profitable per chip — it consumes far more wafer area than standard DRAM. Every wafer converted to HBM removes several wafers’ worth of laptop memory from the market. The result: less consumer supply at exactly the moment demand is rising.
Micron went furthest. In late 2025 it shut down Crucial, its direct-to-consumer memory brand, exiting the consumer RAM and SSD market entirely to redirect capacity toward AI customers. The bet paid off spectacularly for Micron: its fiscal Q3 2026 revenue more than quadrupled to $41.5 billion at an 84.9% gross margin, and its stock surged. The bet did not pay off for anyone shopping for a laptop.
Component prices went vertical
The raw numbers from TrendForce’s contract price surveys tell the story better than any narrative:
DRAM (conventional, contract prices):
- Q1 2026: up 90–95% quarter over quarter — a record. PC DRAM specifically more than doubled.
- Q2 2026: up another 43–48% for PC DDR5 (early forecasts ran as high as 58–63% before moderating).
- LPDDR5X, the soldered memory in most thin-and-light laptops: up roughly 90% in Q1, the steepest increase in its history.
NAND flash (SSDs):
- Q1 2026: up 55–60%, with client SSD contract prices up over 40%.
- Q2 2026: up 70–75% — the first time in this cycle NAND outpaced DRAM.
DRAM and SSD price history: how fast it moved
| Period | DRAM (contract) | NAND / SSD (contract) | Retail reality check |
|---|---|---|---|
| Early 2025 | Stable, pre-shortage baseline | Stable | 32GB DDR5 kit: $80–120 |
| Q4 2025 | +30% QoQ; DDR5 up ~70% YoY, some parts +170% | Accelerating; NAND wafer prices up as much as 60% MoM in November | Spot prices begin vertical climb; Crucial exits consumer market |
| Q1 2026 | +90–95% QoQ (all-time record); PC DRAM more than doubles; LPDDR5X +~90% | +55–60% QoQ; client SSDs +40%+ | 16GB DDR5 module: $142 (March) |
| Q2 2026 | PC DDR5 +43–48% QoQ (moderated from an initial +58–63% forecast); DDR4 +35–40% | +70–75% QoQ — NAND overtakes DRAM for the first time this cycle | 16GB DDR5 module: $201 (April); cheapest 32GB DDR5 kit: ~$439 (June 15) |
| H2 2026 (forecast) | Still rising through Q3–Q4, but decelerating — buyers have hit their affordability ceiling | Enterprise SSD demand keeps squeezing client supply | Gartner: combined DRAM + SSD prices up more than 130% by year-end |
| 2027–2028 | First meaningful new fab capacity arrives late 2027 at the earliest | Same timeline | Prices stabilize but don’t return to 2025 levels |
Cumulative effect: standard laptop memory costs roughly three to four times what it did 18 months ago, and spot trackers recorded spikes above 2,000% on individual parts during the worst of the run-up.
What that looks like at retail: a 32GB DDR5 kit that cost $80–120 in early 2025 was selling for around $439 by mid-June 2026 — a three-to-four-fold increase. A 16GB DDR5 module averaged $201 in April 2026, up from $142 a month earlier. Even obsolete DDR4 roughly doubled or tripled. Spot price trackers recorded spikes well over 2,000% on some parts across the year. GDDR6 and GDDR7 graphics memory more than tripled in six months, which is why GPU prices are climbing too.
Memory chipmakers booked $97 billion in Q1 2026 revenue, up 81% year over year. That money came from somewhere. Increasingly, it comes from you.
Laptop makers passed it on — brand by brand
Dell moved first and hardest: 15–20% increases starting mid-December 2025. COO Jeff Clarke told investors he had “never seen memory-chip costs rise this fast” and called the situation unprecedented.
Lenovo notified customers that all quotes expired January 1, 2026, and urged them to lock in orders early. Its hedge: stockpiling component inventory at roughly 50% above normal levels, which bought it pricing flexibility competitors lack.
HP disclosed that memory now accounts for 15–18% of the cost of a typical PC — double the previous year — and that its stockpile would carry it only through the first half of its fiscal year, with margin pressure hitting from May 2026 onward. CEO Enrique Lores flagged H2 2026 as the danger zone.
Apple absorbed costs longer than anyone thanks to its long-term supply agreements, then capitulated on June 25, 2026 with increases of up to $300 on Macs and iPads (and $1,300 on the top Mac Studio). Counterpoint Research estimates component costs could add $150–200 per device to the next iPhone lineup as well.
Framework, the modular laptop maker, delisted standalone memory to fend off scalpers, criticized Dell’s hikes publicly, then had to raise its own prices anyway — twice — noting that SSD pricing in particular kept deteriorating.
Beyond laptops: Nintendo (+$50 on Switch 2), Microsoft (+$100–150 on Xbox), Sony, Valve (Steam Machine launched $300 above its intended price), boutique builders like CyberPowerPC and Maingear, and even Raspberry Pi all raised prices citing memory costs. Industry-wide, Consumer Reports and analysts tracked laptop increases of 15–30% across major brands in 2026, with IDC projecting 10–20% increases across PCs, tablets, and phones.
Laptop and device price hike timeline
| Date | Company | What happened | Size of increase |
|---|---|---|---|
| Late Nov 2025 | Micron | Shuts down Crucial, exits consumer memory entirely | Aftermarket RAM/SSD supply shrinks |
| Late Nov 2025 | Dell | COO Jeff Clarke calls memory costs “unprecedented” on earnings call | Warning issued |
| Dec 7, 2025 | CyberPowerPC | First boutique builder to reprice; Maingear urges customers to buy now | Unspecified |
| Mid-Dec 2025 | Dell | First major OEM to raise PC and server prices | +15–20% |
| Dec 2025 | Framework | Delists standalone memory to block scalpers, then raises laptop memory prices | Multiple rounds |
| Jan 1, 2026 | Lenovo | All existing customer quotes expire; new pricing takes effect | +15–20% signaled |
| Jan 2026 | Raspberry Pi | Temporary price increase on memory-heavy boards | Varies by model |
| Q1 2026 | Acer, ASUS | Notify clients of tougher contract terms and increases | +15–20% signaled |
| May 2026 | HP | Pre-shortage memory stockpile runs out; margin pressure begins | H2 2026 hikes flagged |
| 2026 | Nintendo | Raises US Switch 2 price, explicitly blaming memory costs | +$50 |
| Jun 25, 2026 | Apple | Raises Mac, iPad, and accessory prices; stock falls 6%+ | MacBook Air $1,099?$1,299; MacBook Pro $1,699?$1,999; MacBook Neo $599?$699; iPad $349?$449; Mac Studio (M3 Ultra) +$1,300 |
| Aug 2026 | Microsoft | Xbox console prices rise, citing the same memory costs | +$100–150 |
| Late 2026 | Valve | Steam Machine launches above its originally intended price | +$300 vs. plan |
Note the sequence: Apple, historically the company most able to absorb component swings through long-term supply agreements, held out six months longer than Dell and Lenovo — then raised prices anyway. Counterpoint Research pointed out that Microsoft, Samsung, Sony, Dell, HP, and Lenovo had all moved before Apple did. When the last holdout folds, that’s the clearest signal the cost pressure is structural, not negotiable.
The irony: AI PCs made it worse
Here’s the twist. At the same time AI demand was strangling memory supply, the industry was pushing consumers toward “AI PCs” that require more memory.
Microsoft’s Copilot+ PC certification mandates a minimum of 16GB of RAM, a 256GB SSD, and a neural processing unit (NPU) capable of 40+ TOPS. Local AI features — Recall, live captions, Studio Effects, on-device language models — all need memory to load into. The 8GB laptop, long the budget default, doesn’t qualify.
So the baseline spec of a “modern” Windows laptop doubled its RAM requirement precisely when RAM became the most expensive component in the machine. TrendForce notes memory can account for about 18% of an AI PC’s total bill of materials. PC makers are now reportedly rethinking their 2026 AI PC roadmaps because the machines they spent two years marketing have become uneconomical to build at mainstream prices. Gartner expects the broader AI PC rollout to slow specifically because of pricing.
The hidden price hike: shrinkflation
Not every increase shows up on the price tag. TrendForce’s Avril Wu predicted manufacturers would quietly downgrade specs to hold price points — and that’s exactly what happened. A $600 laptop in 2026 may look identical to its 2025 predecessor while shipping with 8GB of RAM instead of 16GB, a smaller SSD, or a dimmer display. TrendForce documented entry- and mid-range models being downgraded specifically to cut bill-of-materials costs.
The endgame is starker: Gartner projects the sub-$500 laptop segment will effectively disappear by 2028, because thin-margin machines can’t absorb memory costs at all. Gartner also forecasts average PC prices rising 17% while global shipments fall 10.4%; TrendForce forecasts notebook shipments dropping 13.5% in 2026, a sharp reversal from its earlier growth forecast.
When does this end?
Not soon. The consensus across TrendForce, Gartner, IDC, and the memory makers themselves:
- No meaningful new supply before late 2027. New fabs take two to three years to build and reach volume. SK Hynix is investing in a $91 billion cluster and plans to gradually double wafer output over five years; Micron expects meaningful new capacity in 2027–2028.
- The shortage could stretch further. SK Hynix has signaled shortages could persist through 2030 in some scenarios. IDC describes this not as a cyclical shortage but as “a potentially permanent, strategic reallocation of the world’s silicon wafer capacity.”
- Price growth is slowing — for a bleak reason. TrendForce notes increases decelerated in the second half of 2026 not because supply improved, but because consumers hit the ceiling of what they’ll pay.
- Prices won’t return to 2025 levels even after capacity expands. The demand floor from AI infrastructure is permanently higher.
What you should actually do
If you need a laptop, buy sooner rather than later. Every analyst tracking this market — Gartner, TrendForce, IDC, Counterpoint — says waiting is unlikely to help before 2028 at the earliest.
Consider last year’s model or refurbished. Year-over-year gains are modest, and pre-shortage inventory carries pre-shortage pricing. A 2024–2025 machine bought today dodges most of the increase.
Read the spec sheet, not just the price. Shrinkflation means the same price can now buy less machine. Check RAM and SSD capacity against the previous model year before assuming a deal is a deal.
Buy more RAM than you think you need — up front. With Crucial gone and aftermarket kits selling at 3–4x historical prices, upgrading later is now the expensive path. If the machine has soldered LPDDR memory, later isn’t even an option.
Don’t pay for “AI PC” branding you won’t use. An NPU only matters if your software uses it. If your workload is browsing, Office, and video calls, a well-specced non-Copilot+ machine at a lower price is the smarter buy.
FAQs
Why are laptop prices going up in 2026?
AI data centers are consuming most of the world’s memory chip supply. DRAM and SSD prices doubled or tripled, and memory now makes up roughly 35% of a laptop’s manufacturing cost.
How much more expensive are laptops in 2026?
Major brands raised prices 15–30%. Dell added 15–20% in December 2025, Lenovo repriced in January, and Apple raised MacBook prices by $200–300 in June 2026.
Is AI really the cause of the RAM shortage?
Yes. Samsung, SK Hynix, and Micron shifted production toward high-bandwidth memory for AI servers, which pays higher margins and uses more wafer space, shrinking laptop memory supply.
Should I buy a laptop now or wait?
Buy sooner. Analysts at Gartner, TrendForce, and IDC expect no meaningful supply relief before late 2027, and prices are unlikely to return to 2025 levels afterward.
Will laptop prices ever go back down?
Probably not to 2025 levels. New memory factories arrive in 2027–2028, but AI demand has permanently raised the floor. Some analysts see shortages persisting into 2030.
Why did Apple raise MacBook prices?
Memory and storage costs became unsustainable to absorb. Tim Cook called the shortage a “hundred-year flood” before Apple raised Mac and iPad prices on June 25, 2026.
What is happening to cheap laptops under $500?
Gartner expects the sub-$500 segment to effectively disappear by 2028. Rising memory costs make thin-margin budget machines unviable, so makers are cutting specs or dropping models entirely.
How much RAM should a laptop have in 2026?
Treat 16GB as the floor — it’s Microsoft’s Copilot+ minimum and the practical baseline for Windows 11. Buy it up front; aftermarket RAM now costs three to four times more.
The bottom line: the same wafers can become memory for an AI server or memory for your laptop, and right now the AI server pays more. Until hundreds of billions of dollars in new fab capacity comes online — 2028 at the earliest — every gigabyte in your next laptop is competing with a data center for the privilege of existing. The data center is winning, and you’re paying the difference.
