Global memory chip costs have more than doubled in some segments over the past year, driven by voracious demand from AI data centres. Apple CEO Tim Cook warned this week that price rises for the company’s devices are now “unavoidable,” while the head of Microsoft’s Xbox unit disclosed that storage component costs had risen fivefold compared with two years earlier. The shortages extend beyond luxury electronics: a coalition of trade groups told the White House that an “urgent imbalance” in the memory market could cause sustained price increases for American households.

The mechanism is a sudden, massive build-out of AI infrastructure. Hyperscalers are hoovering up high-bandwidth memory and NAND flash for training and inference clusters, leaving consumer and automotive manufacturers competing for a constricted supply. Apple, which has previously absorbed component cost increases, now signals it can no longer shield customers. Its next iPhone, the iPhone 18 expected in September, could cost over $200 more, according to research firm Techinsights. Ford and Dell have issued similar warnings.

The economic consequences are matched by subtler societal ones. In hiring, the use of AI-generated thank-you notes is backfiring: actress and interviewer Anne Hathaway recounted receiving identical AI-written messages from multiple candidates, instantly revealing a lack of genuine effort. On social platforms, users are documenting an accelerating reliance on chatbots for deeply personal decisions—what to eat, how to dress, whether to leave a relationship. Wharton School fellow Cornelia C. Walther describes a drift toward “full-blown addiction.” Researchers caution that over-delegating choice to algorithms may weaken the cognitive muscles required for critical thinking and resilience to uncertainty.

Safety gaps are also surfacing. Ethical hackers at British security startup Mindgard discovered a jailbreak in ChatGPT’s image generator that produced photorealistic scenes of extreme violence, including a bound and bruised woman. The prompt contained no offensive words, slipping past content filters. OpenAI told the researchers it had resolved the flaw, but a simple variation of the original prompt was still able to generate equally graphic material a month later. The episode underscores the fragility of guardrails in generative systems.

Apple’s autumn product launch will be the first mass-market test of how far AI-driven component inflation translates into higher sticker prices. In Washington, the memory supply letter to the White House has yet to elicit a policy response, while OpenAI faces renewed pressure to demonstrate its safety mitigations are more than patchable feints.