Your Phone Just Got More Expensive Because Nvidia Needs Your RAM

Samsung has never lost money on phones. That streak is about to die.

Samsung has survived economic crashes, pandemic supply chain chaos, and the slow death of the smartphone upgrade cycle. Through all of it, the mobile division has always turned a profit. Until now.

According to Korean outlet Money Today, Samsung MX (mobile experience) head TM Roh has warned company leadership that 2026 could bring the first net loss on smartphones in Samsung's history. The culprit isn't weak sales — the Galaxy S26 is selling fine. It's not competition. It's not tariffs. It's DRAM and NAND prices, which have skyrocketed because every data center on Earth is buying every memory chip they can find to feed AI workloads.

Ryan Whitwam at Ars Technica has the receipts.

The math is absurd

Let me give you a number that should make you mad: Nvidia's upcoming Vera AI CPU — the successor to Grace — packs up to 1.5 terabytes of LPDDR5x memory. Per chip. A single rack-scale AI platform with 36 Vera CPUs and 72 Rubin GPUs will consume enough RAM to build 4,600 Galaxy S26 Ultras (12GB each).

One rack. Four thousand six hundred phones.

This isn't a metaphor. This is the actual bill of materials. The same LPDDR5x chips that go into your phone are getting vacuumed up by data centers at a rate the supply chain can't match. According to Counterpoint Research, RAM now accounts for more than a third of the total cost of building a budget smartphone — roughly double what it was before the AI infrastructure buildout kicked into high gear. Even on flagship devices, memory is north of 20% of the BOM.

The processor used to be the expensive part. The display was second. Now? RAM is eating everyone's lunch.

The prices are already here

This isn't a "coming soon" problem. Motorola just raised Moto G prices by up to 50%. Samsung added $50 to the Galaxy A37 and A57, $80 to the Z Flip 7 and Z Fold 7, and $100 to the Galaxy Tab S11. These aren't premium upsells — they're cost-passing dressed up as product updates.

The budget phone category, specifically, is getting annihilated. When RAM is a third of your BOM and RAM prices double, you can't build a $200 phone anymore. The math stops working. Counterpoint's data suggests the very concept of a budget smartphone becomes "suspect" in the coming years, which is a polite way of saying dead.

The beautiful irony no one's talking about

Here's where it gets genuinely funny: Samsung's semiconductor division just had its best quarter ever. An estimated $38 billion in profit for Q1 2026 — more than seven times what it earned in Q1 2025. Samsung Semiconductor is absolutely crushing it because they sell DRAM and NAND. The same price spike that's killing Samsung Mobile is making Samsung Semiconductor richer than it's ever been.

Same company. Two divisions. One's drowning, the other's swimming in cash.

It's the corporate equivalent of your left hand setting your right hand on fire and then selling it bandages.

It gets worse before it gets better

Samsung, Micron, and SK Hynix are all scrambling to expand production. Samsung is actually spinning down LPDDR4 lines to boost LPDDR5 output. But Nikkei Asia projects that even with best-case production improvements, DRAM output in 2027 could fall 40% short of demand.

Forty percent. That's not a supply chain hiccup. That's a structural mismatch between how much memory the world's AI ambitions require and how much memory the world can physically produce.

And here's the kicker: the only thing that changes this forecast is a substantial reduction in AI infrastructure demand. Which means every major tech company would need to simultaneously decide to stop building data centers. How likely is that?

What this actually means

There's a tendency to talk about AI infrastructure costs in terms of GPU prices and electricity bills. That's the visible part. But the second-order effects — the supply chain cannibalization, the component price cascades — are hitting consumer electronics in ways that aren't getting enough attention.

Your phone costs more. Budget phones might not exist in a few years. Laptop prices are creeping up for the same reason. The AI boom has a hidden tax, and it shows up in your device purchases whether you use AI or not.

Samsung might lose money on phones for the first time ever. If the world's largest smartphone manufacturer can't make the unit economics work, nobody can.

The AI industry spent two years telling us compute was the bottleneck. Turns out it's memory. And everyone's paying for it.