If you tried to buy a laptop or build a PC in 2026, you felt it before anyone explained it. Prices were higher than expected. Memory kits cost far more than they did a year ago. A mid-range laptop that sold for $800 last year now costs close to $1,050.
It feels like the chip shortage never ended. In a way, it never did. But the memory chip shortage of 2026 is not the shortage of 2022. The old one is over. A new one took its place, with a different cause, different victims, and a different end date.
This guide explains the global chip shortage 2026. It covers why a new one started, who controls the supply now, and what it means for anyone buying a computer or a phone this year.
What happened to the chip shortage of 2021 and 2022?
The first shortage began with the pandemic. People bought laptops, consoles, and home office gear at the same time. Car makers cancelled chip orders, then tried to buy them back when demand returned. Factories could not keep up, and lead times stretched past a year for some parts.
Then demand cooled. By 2023, the panic was over. Inventories piled up. Some parts were hardest to find in 2022. Car microcontrollers, 4G chips, and power management chips moved from shortage to surplus. Consumer PC processors went the same way. The market was left with plenty of ordinary chips and falling prices.
That is the key point. The shortage everyone remembers is finished. What came next is a different problem, and it started in a different place.
Why is there a shortage again in 2026?
The new shortage comes from AI chip demand. AI data centres need enormous numbers of specialised chips. Each AI accelerator needs a processor, a stack of high-bandwidth memory, and advanced packaging to join them together. All three are in short supply, and all three pass through a small number of factories.
The first bottleneck is the foundry. TSMC makes almost all of the most advanced chips. Its 2nm orders are already booked out to 2028. Demand for its leading-edge wafers is higher than supply, and its 2nm capacity will only double by 2027.
The second bottleneck is TSMC CoWoS packaging. Every modern AI chip must pass through TSMC’s CoWoS packaging lines. Those lines are fully booked, with lead times of 52 to 78 weeks in 2026. Nvidia holds roughly 60% of that capacity, against about 11% for AMD. The gap between packaging supply and demand was about 20% in early 2026. TrendForce expects it to narrow to about 10% by the end of the year. That is progress, but it is still a shortage.
The third bottleneck is memory, and the HBM memory shortage is the tightest link of all. SK Hynix’s 2026 output of high-bandwidth memory is already almost entirely allocated. Its share of HBM revenue was 50% in the second quarter of 2026, with Samsung at 33% and Micron at 18%, according to Counterpoint Research. Nvidia’s own supply commitments jumped from $119 billion to $279 billion in a single quarter, mostly for memory and factory capacity.
Demand is so strong that even finished chips are spoken for before they exist. Chinese firms have reportedly ordered more than 2 million NVIDIA H200 chips for 2026 delivery, against about 700,000 in stock. Put together, the picture is clear. The shortage is no longer one problem. It is a stack of three, and each layer depends on a handful of suppliers.
What is HBM, and why does it matter so much?
HBM stands for high-bandwidth memory. It is memory stacked in layers and placed right next to the processor, so data moves fast enough to feed an AI chip. A normal memory chip cannot keep up. Without HBM, the fastest AI accelerator in the world would spend most of its time waiting.
Only three companies make HBM at scale: SK Hynix, Samsung, and Micron. SK Hynix HBM leads the market, with Samsung and Micron behind it. Together, Samsung and SK Hynix control about two-thirds of all DRAM production and close to 90% of HBM output. When AI demand surged, these companies moved factory capacity toward HBM because it earns far more per wafer than ordinary memory. That left less capacity for the memory in phones and PCs.
AI data center chips sit at the centre of that demand. Data centres are expected to consume about 70% of all memory chips made in 2026. Consumer devices get what is left. That is the simple maths behind the price rises.
Why did RAM and SSD prices jump so much?
The RAM price increase of 2026 moved faster than at any point since the pandemic. TrendForce data shows conventional DRAM contract prices rose 90% to 95% in the first quarter of 2026. They rose another 58% to 63% in the second quarter. NAND flash, the chip inside every SSD, climbed 55% to 70% over the same period. The industry started calling it “RAMageddon”.
Prices kept climbing after that. TrendForce forecast further increases of 13% to 18% for DRAM and 10% to 15% for NAND in the third quarter. Gartner’s DRAM prices 2026 forecast is a rise of 125% across the year as a whole. NAND flash prices are expected to rise 234%. It expects worldwide semiconductor revenue to pass $1.3 trillion for the year.
The effect on buyers is direct. Memory now makes up close to 40% of a PC’s component cost, against about 15% before, according to Omdia. DDR5 prices show it best: memory kits have tripled or quadrupled, with entry-level kits starting at about $350. Laptop makers have quietly shrunk SSD capacity in some models to hold prices down. Intel’s new Panther Lake platform arrived with a reported 20% cost premium built in.
Phones felt it too. Counterpoint Research found that phones already on sale had risen about 15% on average by September 2026. New models were launching at prices roughly 25% above their predecessors. Memory rose to about 34% of an iPhone 18 Pro’s bill of materials, from about 10% a year earlier, according to TrendForce.
Who actually controls the chip supply?
The semiconductor supply chain is short, and every link is concentrated. A chip starts as a design, gets printed by a foundry, joined to memory by a packager, and tested before it ships. At each step, one or two companies dominate.
| Stage | Main suppliers | Status in 2026 |
|---|---|---|
| AI chip design | Nvidia, AMD, Broadcom | Demand far above supply |
| Leading-edge foundry | TSMC, with Samsung and Intel behind | 2nm capacity booked to 2028 |
| HBM memory | SK Hynix, Samsung, Micron | Tightest link, output allocated |
| Advanced packaging | TSMC CoWoS | Fully booked, 52 to 78 week waits |
| Lithography machines | ASML | Sold out into 2027 and 2028 |
ASML deserves a mention because it sits one level deeper. This Dutch company makes the extreme ultraviolet machines that print the most advanced chips. There is no real alternative, and its machines are sold out into 2027 and 2028. No new fab can open without them.
Geography matters as well. Advanced logic is anchored in Taiwan, memory in South Korea, and the most advanced machines in the Netherlands. The US and EU chip laws of 2022 and 2023 funded new fabs, and some of those plants are coming online in 2026. They help, but they cannot replace the existing hubs overnight.
How the shortage changed: 2022 vs 2026
| 2022 | 2026 | |
|---|---|---|
| Cause | Pandemic demand and car orders | AI data centre boom |
| Scarce parts | Car MCUs, 4G chips, power chips | HBM, CoWoS packaging, 2nm wafers |
| Consumer chips | Short | Mostly in surplus |
| Prices | High, then falling | Memory soaring |
| End in sight | Yes, by 2023 | New capacity from 2027 |
PC prices 2026: what does this mean for computer buyers?
The most visible effect is on shipments. IDC’s preliminary figures put global PC shipments at 62.7 million units in the third quarter of 2026, down 20.1% on a year earlier. Omdia counted 58.1 million desktops, notebooks, and workstations, down 21.2%. IDC expects the PC market to decline 8.9% across 2026 as a whole.
Prices are doing the opposite. IDC estimates average PC selling prices will rise 4% to 6% in a moderate case and 6% to 8% in a worse one. The cost is being passed to buyers, not absorbed by makers. A mid-range laptop that cost $800 a year ago now sells for closer to $1,050.
The same pressure reaches every device with memory in it. That is nearly all of them. Phones, tablets, consoles, and even smart TVs carry DRAM and NAND flash. When those two parts get dearer, everything gets dearer.
Graphics cards are a special case. They need both advanced chips and large amounts of memory. High-end models were already hard to buy. The Blackwell GPU shortage has pushed some buyers toward older or smaller models.
What about phones?
Phones face the same maths as PCs, with one extra twist. A phone chip now runs AI features on the device itself, from photo editing to voice typing. That work needs fast memory and a capable processor in a tiny space.
Google’s Tensor chips show where this is heading. They are designed around on-device AI, so a Pixel phone can run demanding features without sending everything to a server. That design only works if the memory and the processor keep pace with each other. When memory supply tightens, the whole balance shifts.
The result is visible in prices. Phones already on sale rose about 15% on average by September 2026, according to Counterpoint. New models launched about 25% above their predecessors. Budget phones felt it most, because memory is a bigger share of a cheap phone’s cost.
Should you buy now or wait?
This is the question every buyer asks, and the honest answer depends on need. If your laptop is failing, waiting will not help. Prices are not expected to fall soon, and the part you need may cost more in six months.
If you can wait, you can. The earliest realistic point for real price relief is late 2027, and full normality may not return before 2028, according to Micron’s own guidance. That is a long time to nurse a dying machine.
- Buy the memory you need now, not the memory you might need later. A 16GB laptop is fine for most people. Go for 32GB if you edit video or run virtual machines.
- Check the SSD size before you buy. Some makers have shrunk storage to hold the sticker price down. A 256GB drive fills up fast.
- Consider the previous generation. Last year’s laptop or graphics card is often nearly as fast and far cheaper, and it is usually in stock.
- Upgrade what you have first. A fresh SSD or more RAM can add years to a working machine, and it costs far less than a new one.
- Set a budget and stick to it. In a rising market, the most expensive mistake is buying a machine you did not need because a discount appeared.
One more thing. Do not panic-buy. Shortages create rumours, and rumours create queues. Buy what you need, when you need it, and ignore the noise.
When will the chip shortage end?
New chip manufacturing capacity is coming, but chip factories take years to build. The first big wave of new fabs is expected online from mid-2027. TrendForce expects the severe shortage to start easing in 2027, helped by orders spreading to other packagers and by TSMC’s expansion.
Memory is the slowest part to fix. SK Hynix expects memory shortages to last through 2030. Micron has warned that markets could tighten further in 2027 and 2028. CoWoS packaging should improve sooner, with the supply gap narrowing from about 20% to about 10% by the end of 2026.
So the honest answer is split. The worst of the packaging squeeze should ease in 2027. Memory prices will take longer. Anyone planning a purchase in 2026 should assume today’s prices are the new normal, not a blip.
Questions buyers keep asking
Is the global chip shortage 2026 over?
The old one is over. Its replacement is still going strong. Ordinary chips are plentiful. Advanced chips and memory are scarce, and the shortage has moved from cars and consoles to AI data centres and memory.
Why are RAM prices so high in 2026?
AI data centres buy most of the world’s advanced memory. Chipmakers moved capacity toward HBM because it pays more. Less capacity is left for ordinary DRAM, so prices rose sharply.
Will laptop prices come down in 2027?
Probably not much. New capacity arrives in 2027, but memory is expected to stay tight. Prices may stop rising, but a return to 2024 prices is not expected soon.
Is it a good time to build a gaming PC?
It is an expensive time. Memory and graphics cards cost far more than a year ago. If you must build now, buy a mid-range card and a previous-generation processor. Waiting until late 2027 may get you far more for your money.
The bottom line
The chip shortage did not end. It changed shape. The semiconductor shortage 2026 came from AI demand, not from a pandemic. Back in 2022, the problem was too much demand for ordinary chips. This time, the cause is too much demand for advanced memory and packaging.
For buyers, the message is simple. Ordinary computers are available. Memory is dear. Prices will stay high into 2027. Buy what you need, check the specs carefully, and do not wait for a return to old prices that may not come.















