Future Technology[Hardware Weekly] AI just ate your RAM upgrade
Three chipmakers, three very different bets this week, and all of them land on your wallet. AMD stopped selling chips and started selling whole racks. Nvidia keeps colonising the laptop you are about to buy. And the RAM inside it costs six times what it did a year ago.
The Big 3
AI ate the RAM supply, and your next upgrade is paying for it
A gigabyte of memory has gone from about $2.80 in 2025 to around $12 in 2026, a sixfold jump in a single year. Not a shortage of raw materials, a redirection of priorities: the same manufacturers making everyday DDR5 also make the high bandwidth memory that feeds AI accelerators, and HBM pays far better, so production lines have shifted toward it.
Why it matters: this is the moment the AI boom stops being an abstract headline and shows up on your invoice, and it hits hardest for anyone running a NAS, home lab or memory-heavy build.
Nvidia wants to be inside every laptop, not just every GPU
Nvidia unveiled its RTX Spark Superchip for laptops and desktop PCs at Computex, pushing further into AI PC silicon rather than sticking to standalone graphics cards. It is chasing the same on-device AI processing race as Qualcomm and Intel, all fighting to own the chip that decides how fast your machine handles local AI tasks.
Why it matters: the GPU maker becoming a platform player changes who you are actually buying from when you pick a new laptop, and it will not be the last chip company to blur that line.
AMD stopped selling chips and started selling whole racks
AMD's Advancing AI event confirmed EPYC Venice, a 256 core, 512 thread server chip built on Zen 6 and TSMC's 2nm process, delivering up to 1.7x the performance of the previous generation. It ships alongside the Helios rack system, AMD's answer to Nvidia's full data centre bundles rather than a single component competing on a spec sheet.
Why it matters: when your rival sells racks, selling a faster chip is not enough anymore, this is AMD picking a fight on Nvidia's own turf.
Quick Hits
Google Pixel 11 launches this month with the new Tensor G6 chip, built for faster on-device AI rather than benchmark bragging rights.
Samsung's Galaxy Book6 laptops lean harder into on-device AI features, part of the same chip race reshaping every laptop launch this year.
Apple's MacBook Neo undercuts the whole lineup at $599 or £599, the cheapest MacBook Apple has ever sold.
Qualcomm's Snapdragon Wear Elite moves wearable chips to a 3nm process with a dedicated NPU, aiming to run billion-parameter AI models on a watch instead of a phone.
Pick of the Week
Corsair Vengeance RGB DDR5 32GB (2x16GB) 6000MHz
A mainstream 32GB kit that sits in the sensible middle for most builds, and with memory prices climbing the way they are, the case for buying now rather than waiting is stronger than usual.
Who it is for: anyone planning a PC build or upgrade this year. Skip it if you already have enough memory for your workload, prices are not coming down soon enough to justify buying ahead of need.
Price: check current price on Amazon, it moves often given the market right now.
Check current price on Amazon →
Worth Waiting For
Google Pixel 11 Pro Fold
Tipped for later in 2026, the Pixel 11 Pro Fold is expected to carry the same Tensor G6 chip as the standard Pixel 11 into a foldable body. Nothing is confirmed by Google yet, treat this as a credible rumour rather than a locked launch.
Why wait: if a foldable is what you actually want, buying a standard Pixel 11 now and regretting it in three months is the classic mistake. Worth holding out to see official confirmation first.
Forwarded this? Get your own hardware briefing at futuretechnologyhq.com/newsletter
Happy building. Nath, Future Technology
Some links in this newsletter may be affiliate links. We only recommend products we genuinely think are worth your time.