SAN FRANCISCO, Jan 16, 2026, 09:28 (PST)
- Nvidia says memory supply constraints are limiting output of GeForce RTX GPUs
- RTX 5070 listings flipped between in-stock and out-of-stock at major retailers on Friday
- GPU makers are weighing which models to build as scarce GDDR7 memory gets rationed
Nvidia said on Friday that “memory supply” constraints are limiting production of its GeForce RTX graphics cards, a squeeze that has kept the RTX 5070 hard to find in steady volume. (Overclock3D)
The RTX 5070 sits in the middle of Nvidia’s desktop lineup, aimed at buyers upgrading older gaming PCs. That band tends to show stress fast when supply slips, because shoppers compare prices across similar cards and walk away when markups jump.
The bottleneck is memory — chips used to store data for graphics processing — at a moment when artificial intelligence data centres are soaking up capacity. “We have to support memory consumption for AI infrastructure,” said Sungsoo Ryu, CEO of SK Hynix America, one of the suppliers to Nvidia, in an interview with Reuters. (Reuters)
Board partners have also been trying to calm nerves after mixed messages on supply. Asus, one of Nvidia’s biggest add-in-board partners — the companies that build custom versions of Nvidia’s chips — said it had no plans to stop selling its RTX 5070 Ti and RTX 5060 Ti 16 GB models, and blamed “incomplete information” after reports the cards were being wound down. (Asus)
Gigabyte CEO Eddie Lin said the shortage can force a blunt calculation: allocate scarce GDDR7 — the graphics memory used on RTX 50-series boards — to products that return more revenue per gigabyte. “They will calculate how much revenue … per gigabyte of memory,” Lin told an interview published by Tom’s Hardware. (Tomshardware)
On Nvidia’s own specifications, the RTX 5070 comes with 12 GB of GDDR7 memory on a 192-bit interface and a 250-watt total graphics power rating. Nvidia lists a 650-watt system power requirement and pitches the card alongside the higher-end RTX 5070 Ti. (Nvidia)
The RTX 5070 launched with a $549 manufacturer’s suggested retail price, and Nvidia delayed its Founders Edition version behind partner cards, pushing early buyers toward third-party models. Reviews at the time challenged Nvidia’s early comparisons to the far more expensive RTX 4090 flagship. (Theverge)
On Friday, a NowInStock tracker showed several RTX 5070 models briefly available at Amazon, B&H and Newegg, with some listings at $549.99 and others at $589.99 to $639.99. The same page logged repeated flips back to “out of stock” for multiple models. (Nowinstock)
Competition has tightened in the same price band. In the U.K., PC Gamer’s price watch listed the RTX 5070 at about £480 on Amazon, and said AMD’s Radeon RX 9070 was about £30 more while generally running faster — including with ray tracing, a lighting technique used in games — though it warned prices have started creeping up again. (Pcgamer)
But the mix could change quickly if the memory crunch deepens. If suppliers keep steering output toward data-centre parts, Nvidia and its partners may keep prioritising models that use less memory or deliver higher margins, leaving RTX 5070 stock uneven and street prices above MSRP.
A clearer signal will be whether retailers can keep cards available for days rather than minutes, and whether partner models settle closer to list price instead of living hundreds of dollars above it. For now, buyers are watching restocks, not new launches.
Nvidia has not said when the constraints will ease, and it has not published shipment targets for the RTX 5070 line. The next few weeks will show whether Friday’s flashes of supply become routine — or stay a brief window.
