Chinese AI Labs Are Winning the Price War, and Western Labs Are Starting to Answer With Price Cuts of Their Own

For much of the past year, the AI pricing conversation has run one direction: Chinese labs, led by DeepSeek and increasingly joined by others, release models priced well below comparable Western offerings, forcing Western labs to either match on price or compete purely on capability instead. This month's GPT-5.6 Luna release, which paired capability improvements with an aggressive price cut of its own, is a visible sign that the second strategy — compete on capability, hold the price line — is giving ground to direct price competition even from the labs that have historically been able to charge a premium. The knock-on effect showing up alongside the pricing pressure is a memory chip shortage that's started reaching consumer markets — a reminder that inference at falling prices and rising volume still runs on physical hardware with its own supply constraints, and that a software-layer price war doesn't make the underlying compute and memory costs disappear, it just shifts who absorbs them. For businesses buying AI capacity, the practical read is that price is becoming a less reliable signal of capability gap than it was a year ago — a cheaper model this month may simply mean its lab is fighting harder for market share, not that it's meaningfully behind on quality.