---
title: "Why Apple Just Got RM2,600 More Expensive"
description: "Apple raised Malaysian prices by up to RM2,600 in June and the ringgit had nothing to do with it. Blame the memory chips AI data centres are buying up."
url: "https://www.mrmoneytv.com/articles/apple-price-increase-memory-chip-shortage/"
category: "Investing & Market"
author: "Finlit"
published: 2026-08-11
source: "Mr Money TV"
---
# Why Apple Just Got RM2,600 More Expensive
Apple raised Malaysian prices by up to RM2,600 in June and the ringgit had nothing to do with it. Blame the memory chips AI data centres are buying up.
## Key takeaways
- Apple raised Malaysian prices on 26 June 2026, by RM450 on the entry iPad and RM2,600 on a 14-inch MacBook Pro M5 Max. The worst case was the 16-inch M5 Max with 128GB and 8TB, which went from RM29,899 to RM43,499.
- Prices went up by spec rather than evenly across the range, which points at a single component. It's memory that repriced, not the processor.
- Only three companies make the high bandwidth memory that AI accelerators need: Samsung, SK Hynix and Micron. SK Hynix has said its memory capacity is essentially sold out for 2026, and DRAM contract prices rose as much as 98% in the first quarter of 2026.
- Gross margins on AI-grade memory have run as high as 85%, against roughly 35% to 50% for ordinary consumer memory, so makers point their limited factory capacity at the AI buyers. Micron shut its Crucial consumer brand in February 2026 after nearly 30 years.
- Memory runs in cycles. Micron's revenue fell from USD30.76 billion in FY2022 to USD15.54 billion in FY2023, then recovered to a record USD37.38 billion in FY2025.
- Moomoo's AI industry chain screen lists 329 listed companies worth roughly USD47 trillion combined, split across the application, algorithm and infrastructure layers. That's a much wider field to look at than NVIDIA on its own.
import VideoEmbed from '../../components/VideoEmbed.astro';
import Disclaimer from '../../components/Disclaimer.astro';
import BarChart from '../../components/BarChart.astro';
import PriceChart from '../../components/PriceChart.astro';
On 26 June, Apple redid its Malaysian price list. A 13-inch MacBook Air that cost RM4,699 the day before became RM5,499. Same machine, RM800 more, and the ringgit had nothing to do with it.
If this were ordinary inflation, everything would have gone up by roughly the same percentage. It didn't. The entry iPad went up RM450. A top-spec 14-inch MacBook Pro went up RM2,600. The pricier the machine, the harder it got hit, and that pattern points at one component inside the box. Work out which one and you also work out where a lot of the AI money is currently landing.
## 1. The price list gives the game away
These are the increases off Apple Malaysia's own pricing, cheapest machine first.
The 14-inch MacBook Pro with the M5 Max went from RM14,999 to RM17,599. The Mac mini, which most people think of as the cheap one, went from RM2,499 to RM3,299. The configuration that took the worst of it barely gets mentioned anywhere: the 16-inch MacBook Pro with 128GB of memory and 8TB of storage went from RM29,899 to RM43,499, a jump of RM13,600 on one laptop.
The machines that got hit hardest are the ones carrying the most memory and storage.
*Apple The Exchange TRX, Kuala Lumpur. Photo: [LegendaryLim](https://commons.wikimedia.org/wiki/File:Apple_The_Exchange_TRX,_Exterior_Night_View.jpg), [CC0](https://creativecommons.org/publicdomain/zero/1.0/).*
## 2. The expensive part is the memory, not the brain
Three chips do most of the work inside your laptop and you never see any of them. The CPU handles ordinary calculation. The GPU does the same kind of work in parallel and in much bigger chunks, which is why it's good at graphics and why NVIDIA became a household name. Then there's memory. Memory is the one that got expensive.
Think about your own memory for a second. You're not holding every fact about your life in your head right now. Ask about something from ten years ago and you have to go and dig for it. Try to juggle nine things at once and you start dropping them, and you end up with a headache. Your laptop works the same way. Open thirty tabs on a machine with 8GB of RAM and it crawls. Open the same thirty on a 32GB machine and nothing happens. You're paying for how much it can hold at once.
RAM is a chip. It gets manufactured, shipped and priced like any other component. For most of the last decade that price moved gently enough that nobody outside the industry paid attention.
*A DDR5 server memory module from SK Hynix, one of three companies making memory at AI scale. Photo: [4300streetcar](https://commons.wikimedia.org/wiki/File:SK_Hynix_DDR5_Tall_MRDIMM.jpg), [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/).*
## 3. AI is eating the world's memory
Training an AI model takes computing power, and it also takes somewhere to park enormous amounts of data while the work happens. Ordinary laptop memory is nowhere near fast enough. What the machines use instead is high bandwidth memory, HBM for short, a stacked chip built to shift data at speeds a consumer part cannot touch.
Samsung, SK Hynix and Micron make it at scale, and those three are the world's entire supply. Producing HBM also eats roughly three times the wafer capacity of standard DRAM for the same gigabyte, so an AI order takes a bigger bite out of a factory than the order size suggests.
The demand side is crowded. Anthropic, OpenAI, Amazon, Microsoft, Google and Meta are all building data centres at once, each trying to be the one whose model people actually use. The rough logic of that race is that whoever buys the most chips wins, so nobody is holding back. Data centres now take an estimated 70% of all memory produced worldwide, [SK Hynix has said its capacity is essentially sold out for 2026](https://tech-insider.org/memory-chip-shortage-2026-ai-consumer-electronics/), and factories are running flat out with order books already stretching into next year. DRAM contract prices [rose as much as 98% in the first quarter of 2026](https://www.thestar.com.my/tech/tech-news/2026/06/26/apple-raises-prices-for-macbooks-and-ipads-as-costs-soar-over-ai).
*Photo: [BalticServers.com](https://commons.wikimedia.org/wiki/File:BalticServers_data_center.jpg), [CC BY-SA 3.0](https://creativecommons.org/licenses/by-sa/3.0/).*
## 4. An 85% margin decides who gets the chips
Ordinary consumer memory is a decent business. Margins sit somewhere around 35% to 50% in normal years. AI-grade memory has been running at up to 85%, meaning that on a chip sold for USD100, about USD85 is profit and USD15 covers the cost of making it. The buyers signing those orders have billion-dollar infrastructure budgets and are not negotiating hard over a chip line item.
Put yourself in the seat of a memory maker. Your factory produces a fixed number of wafers a day and you cannot conjure more. One customer wants consumer parts at a 40% margin. The other wants AI parts at 85% and will take everything you can make. You reallocate, and the consumer side gets whatever is left over.
In February 2026, [Micron shut down Crucial](https://investors.micron.com/news-releases/news-release-details/micron-announces-exit-crucial-consumer-business), its consumer memory brand, after nearly 30 years, and redirected the output to data centre customers. Thirty years of retail business, closed, because the other customer pays better.
So when Apple goes shopping for memory, it's bidding against the data centre buyers for a shrinking pool. Every laptop, phone, console and TV needs the same chips, which is why other brands are likely to follow Apple's price list over the coming months.
## 5. Follow the money past NVIDIA
Say you run a nasi lemak shop. To open, you buy ikan bilis from a fishery supplier, who buys from fishermen in Setiawan. You buy timun from a vegetable distributor, who buys from farms. Your customer hands over RM20 and that RM20 gets broken up and passed down a chain of businesses the customer will never think about. That chain is the industry.
Most people see a busy nasi lemak shop and think about opening a nasi lemak shop. Meanwhile the fellow supplying kacang to every nasi lemak shop in the area doesn't care which one wins, because they all buy from him. He sits one layer back, and it's usually the steadier position.
AI is the same picture at a different scale. Everyone saw OpenAI, then everyone saw NVIDIA, and money piled into a single name until it got very expensive. NVIDIA has suppliers too, and those suppliers have suppliers.
Screening tools have made that chain easy to see. Moomoo's app has an industry chain view under its markets tab that groups listed companies by where they sit in a supply chain, and the AI chain alone holds 329 companies worth roughly USD47 trillion between them. Open the infrastructure layer and you get the names doing the unglamorous work: NVIDIA, Dell, Micron, AMD, Google, Cisco and a long tail behind them.
Those 329 companies split across three layers. The application layer is what you touch, ChatGPT and Claude and the rest. The algorithm layer is the maths underneath that produces the answer. The infrastructure layer is everything physically holding it up: memory, networking, data centres, racks, power, even the connectors and switches that move data fast enough to matter. A strong heart doesn't do you much good if the blood vessels are too narrow to carry the flow, and right now the vessels are the bottleneck.
If you want to look at that chain yourself, Moomoo is [where we opened an account](https://finlit.my/mmtv40), and code MMTV40 gets you their current deposit promotion. Some links here are affiliate links, and Finlit may earn a commission at no extra cost to you.
## 6. Nothing with an 85% margin stays at 85% forever
Go back to the nasi lemak shop. If word gets out that the seller is making 85% on every plate, somebody opens up down the road at 50% and still eats well. Then a third seller comes in at 30%. Economists call that margin normalisation, and it happens to most businesses earning far more than the going rate.
The memory makers are responding to their own boom by building plants, but a new plant takes two to three years to come online. Moore's Law is working in the background too: Gordon Moore, who co-founded Intel, observed that computing performance roughly doubles every 24 months while the cost of producing it falls. Stretch that over enough years and what's scarce and expensive today ends up cheap and ordinary.
Memory has been through this before, and recently.
Micron's revenue fell from USD30.76 billion in FY2022 to USD15.54 billion in FY2023, roughly cut in half, and the company posted a net loss of USD5.83 billion that year. Two years later it did a record USD37.38 billion. Anyone treating either of those years as the normal state of the business got the following year wrong.
The two schools of thought on where this goes are both reasonable. One says supply catches up around 2028 and prices normalise. The other says AI keeps advancing, each generation demands better chips than the last, and the shortage simply moves to whatever the next bottleneck is. Either way the number to watch sits in the quarterly earnings: revenue, profit, and the margin between them. A margin that starts sliding usually means new supply has arrived or somebody has undercut somebody.
## 7. What an analyst report is actually for
There was a time when getting professional opinion on a stock meant opening a brokerage account at every bank in town, because each one published its own research and you wanted to compare them. Now it takes one screen. Look up a stock on most broking apps and you'll find the analyst ratings consolidated: the highest price target anyone has published, the lowest, the average, and where the price sits today.
It's worth knowing which half of that report to trust. The facts are solid. Revenue, margin, the guidance management gave for the next two quarters, all of it is public information that every analyst receives identically. The opinion built on top is where they diverge. Micron has around 29 analysts covering it, which is 29 educated guesses about the same company, with no way of knowing in advance which one is right.
They're still worth reading, because these analysts sit down with management at least once a quarter and you don't. Warren Buffett did not build his Apple position without visiting the operation, and if you asked for the same access tomorrow you would not get it. So treat the range as a reference point and make the final call yourself. Where those analysts sit, and whose interests their report serves, is a separate question, and we get into it in the next episode.
## What to actually do with this
- If you were planning to buy a laptop, phone or console, the price is unlikely to drop soon and other brands are likely to follow Apple. Buy for what you need now rather than waiting out the shortage.
- Skip the spec you'll never use. Memory is the component doing the repricing, so the extra 16GB you tick out of habit is where a big chunk of the increase now lands.
- When you read about an industry, trace the chain backwards. Ask who supplies the company everybody is talking about, and who supplies them.
- Use an industry chain or sector screener to see the whole field before you decide. The obvious name is rarely the only option, and it's usually the priciest.
- Check the margin as well as the headline. An 85% gross margin describes this moment, and margins that high tend to attract competition.
- Read quarterly earnings for direction rather than the number. A margin that starts falling is telling you competition or supply has changed.
Your laptop got more expensive because of a decision made in a factory in Korea about who gets this month's output. That decision shows up in public filings, and the companies making it are all listed, so you can follow it as easily as anyone else can.
We broke the whole thing down on the channel, including a walkthrough of the industry chain screen:
## Frequently asked questions
### Why are Apple products getting more expensive in 2026?
Because memory and storage chips have repriced sharply, not because of currency movements. AI data centre construction has absorbed most of the world's memory production and pushed DRAM contract prices up as much as 98% in the first quarter of 2026. Apple raised Malaysian prices on 26 June 2026 by RM450 to RM2,600 depending on the model, and the increase tracks how much memory and storage a configuration carries.
### What is HBM and why is it causing a memory shortage?
High bandwidth memory, or HBM, is a stacked memory chip built for the enormous data throughput that AI accelerators need. Producing it consumes roughly three times the wafer capacity of standard DRAM per gigabyte, so an HBM order removes a much larger slice of a factory's output than a consumer memory order of the same size. AI buyers are taking that capacity, which leaves less for laptops, phones and consoles.
### How long will the memory chip shortage last?
Memory makers are building new plants, but a fabrication plant typically takes two to three years to come online, so real relief is unlikely before 2028. Analysts expect conditions to stay tight through 2027 and 2028, and SK Hynix has warned the imbalance could last longer than that. The counter-argument is that margins this high attract competition, which has historically pulled prices back down.
### Which companies make memory chips for AI?
Three companies dominate high bandwidth memory production: Samsung and SK Hynix of South Korea, and Micron Technology of the United States. Between them they supply effectively all the HBM used in AI accelerators. Their 2026 capacity is largely committed already, under multi-year supply agreements signed with chip designers and hyperscale cloud operators.
### How do you invest in the AI supply chain beyond NVIDIA?
By tracing where the money goes after it reaches the obvious names. An AI system has three layers: the applications people use, the algorithms behind them, and the infrastructure underneath, which covers memory, networking, data centres, connectors and power. Screening tools that group stocks by industry chain make those suppliers visible, and Moomoo's AI chain alone lists 329 companies. You still have to research a company before you buy it.