Why a 75% Margin Didn't Stop the Nvidia Price Increase

The Nvidia price increase reported by Bloomberg over the weekend comes with a detail that's easy to miss: this is a company already keeping roughly 75 cents of gross profit on every dollar of sales. Some of its largest customers have now been told that servers built around its AI chips will cost more than 15% more in many cases. The reason isn't a fatter margin. It's memory.

The higher prices hit systems shipping early next year, including the flagship Vera Rubin generation and the Grace Blackwell machines already humming in data centers. How much extra depends on the chip generation and how much memory a given configuration carries. The warning reached buyers indirectly — the contract manufacturers that assemble servers for Microsoft, Google and Oracle passed the news down the chain. Reuters relayed Bloomberg's report and said it could not independently confirm it. Nvidia did not comment.

Nvidia stock barely flinched, slipping about 1% to roughly $214.72. The real story sits underneath the ticker. The world is short of affordable memory chips, and the same squeeze that's about to make an AI rack more expensive has already started showing up in what Americans pay for laptops and phones.

Why the Nvidia Price Increase Landed Now

Nvidia doesn't just sell silicon anymore. It sells full systems, and those systems are stuffed with high-bandwidth memory and standard DRAM that Nvidia buys at market rates like everyone else. Its famous gross margin comes from the logic chips it designs. Memory is a bought-in component, and when the price of a bought-in component roughly doubles, the bill moves. That's a pass-through, not a markup.

It's also not isolated to data centers. Tom's Hardware has reported Nvidia raising prices across its GeForce gaming line, with increases reaching nearly 39% on some cards — the RTX 5070 up around 36%, the RTX 5060 up about 27%. When the same cost pressure shows up in a $300 graphics card and a multimillion-dollar rack, it's a supply problem, not a pricing strategy.

The Memory Squeeze Behind the Bill

Memory has quietly become the tightest link in the AI supply chain. It's the least glamorous part of the stack and, right now, the hardest to get. Two things explain almost all of it.

Three companies make nearly all the world's DRAM

Samsung, SK Hynix and Micron together account for more than 95% of global DRAM output, and all three have shifted wafer capacity toward high-bandwidth memory, the stacked variety that feeds AI accelerators and sells at far better prices. SK Hynix told investors last fall that its HBM, DRAM and NAND capacity was essentially sold out for 2026. Micron has walked away from the consumer memory business entirely to serve enterprise and AI buyers.

One AI server eats what eight ordinary ones do

A single AI server consumes eight to ten times the DRAM of a conventional one, so every new cluster drains the pool faster than a traditional data center ever did. Contract prices for DDR5 have more than doubled, and the retail evidence is brutal: a 32GB desktop memory kit that sold for under $90 in early 2025 was fetching several hundred dollars a year later.

What It Does to the Cloud Giants' Math

Microsoft, Google, Oracle, Amazon and Meta can absorb a 15% bump — but not quietly. Each of them has published capital spending plans built on assumed hardware costs, and those assumptions were set before memory went vertical. The likely outcome is not fewer AI projects but fewer racks per dollar, or capex guidance revised upward at the next round of earnings calls. Compute scarcity keeps demand from softening.

Why Your Next Laptop or Phone Costs More

This is where the Nvidia price increase stops being a Wall Street story. Research firm Gartner expects combined DRAM and SSD prices to rise around 130% by the end of 2026, pushing PC prices up roughly 17% and smartphone prices about 13% versus 2025. Some manufacturers are holding the sticker and cutting the specs instead — the same laptop as last year, with 8GB of RAM where 16GB used to be.

Cheap devices take the worst of it, because memory is a bigger share of a budget phone's parts cost than a flagship's. Gartner expects the sub-$500 laptop category to disappear entirely by 2028. If you were planning to replace a machine in 2027, the math argues for moving that up and buying more memory than you think you need, since most thin laptops solder it to the board.

What to Watch on Wednesday

Nvidia reports fiscal second-quarter results on August 26, and the number that matters most isn't revenue — it's the gross margin outlook. If margins hold near 75% while memory costs climb, the pass-through is working and customers are eating it. If guidance slips, Nvidia is absorbing part of the hit. Either way, listen for what the cloud providers say about capex in the weeks after, and check the price of a RAM upgrade before your next purchase.