Gamers and designers, brace yourselves: Nvidia is cutting GPU shipments

The clearest signal is coming out of Europe. Tom’s Hardware reports that a retailer shared a letter from a supplier: the store is allowed to order no more than five GeForce RTX 5070 units per listing, while the RTX 5070 Ti and anything above it is marked as completely unavailable. The explanation in the letter is vague and boils down to “the current market situation.”

Korean news agency Newsis says AMD could start raising GPU prices as early as January 2026, with Nvidia potentially following in February. Against that backdrop, there’s also a gloomy forecast for the flagship: the RTX 5090 could supposedly approach $5,000 by the end of the year despite a $1,999 MSRP. The main driver, according to the same reporting, is memory. Manufacturers are shifting capacity toward HBM for AI accelerators, leaving less supply for consumer GDDR7 and pushing those prices higher. Germany’s Golem.de also ties the risk of price increases in Europe to rising memory costs, especially GDDR and DRAM, and expects the impact to show up as early as Q1 2026.

TrendForce, citing South Korean outlet Kbench, adds another data point: Nvidia is already considering cutting production of some mainstream models by 30–40%, with the RTX 5070 and RTX 5060 Ti among those mentioned. At the top end, TrendForce also points to reports that custom RTX 5090 cards from board partners have been quoted above $3,000–3,500.

The AI boom is forcing gamers and creators to compete with deep-pocketed data centers, and the outcome is usually predictable. For manufacturers, it makes sense to prioritize the segment with higher margins and demand backed by long contracts. Accelerators like Nvidia’s H200 can sell for tens of thousands of dollars per unit, and rising HBM3E costs are pushing overall system pricing higher. On top of that, some consumer cards are being snapped up by firms that rework gaming GPUs for AI cluster workloads and add nonstandard cooling. Even if that remains a niche practice, those batches don’t end up on regular retail shelves, which makes shortages more visible.

In stores, the pattern tends to look the same. First, models with more VRAM disappear faster and increasingly end up as “special order” items. Second, purchase limits start to appear. A similar situation has already shown up in Japan: in Tokyo’s Akihabara electronics district, including at Tsukumo eX., some listings reportedly came with “one card per customer” rules.

Nvidia has not publicly confirmed either production cuts or new supply rules. Still, the signals together suggest the market is sliding into another phase where pricing and availability depend not only on gamer demand, but also on how quickly the AI industry absorbs memory and hardware for itself.