LONDON – By the time Carl Pei signed off on the Nothing Phone (4a), the device he had approved no longer existed. The memory components inside it had doubled in cost between the moment his team committed to building the phone and the moment it reached a warehouse. The retail price went up accordingly. And if Pei is right, that is only the beginning.
The Nothing co-founder and chief executive laid out the situation on X on June 12, with a candor unusual for a smartphone company chief. Memory, he wrote, is now the single most expensive component inside a modern handset – more than the processor, more than the display, accounting for more than half of the total hardware bill in some models. The price has surged roughly 300 percent in certain categories. It will keep rising through next year, he added, offering no quarter to consumers hoping a future festive sale discount will bail them out.
What Pei described is the retail face of a structural deficit that has been building for two years. The global semiconductor market is not experiencing a cyclical dip in demand or a temporary logistics bottleneck. It is experiencing a fundamental reallocation of supply, one driven almost entirely by artificial intelligence infrastructure. The world’s largest cloud providers – the companies that run the data centres behind AI systems – are signing multi-year contracts with chipmakers, locking in production capacity that previously flowed to consumer devices. When Samsung and SK Hynix say their 2026 order books are already full, they mean full for the hyperscalers. The smartphone maker that calls in March to negotiate supply for a September launch is, increasingly, told to get in line.
Jensen Huang, Nvidia’s chief executive and the figure whose company sits at the center of the AI chip economy, confirmed the supply situation plainly: demand will exceed supply for years. What he did not say, but what the consumer electronics industry is now absorbing, is that someone will pay for that imbalance. The answer, as it almost always is, is the person at the end of the chain holding a phone.
The arithmetic of what that means is not abstract. Research firm Techarc found that average smartphone prices in India – one of the most price-sensitive major markets in the world – rose 7.9 percent in the first five months of 2026. For phones below the equivalent of roughly 120 dollars, the increase was 17.6 percent. These are not premium flagship upgrades. These are the devices that constitute the mass market, the handsets that first-time smartphone owners in emerging economies depend on, the segment where a cost increase of even twenty dollars reshapes purchasing decisions for millions of households.
The dynamics at the high end of the market work differently, and not necessarily in consumers’ favour either. Trade groups representing automakers, telecom vendors, and medical device manufacturers warned the Trump administration in early June that AI data centres are consuming memory chips at a scale already raising costs across unrelated industries. In smartphones specifically, the shift in chip economics is consolidating the market at the top. MediaTek and Qualcomm, the chipset suppliers that serve the mid-range and budget segments, are seeing their smartphone shipment volumes stagnate as brands grapple with margin pressure. Apple and Samsung, whose flagship lines carry enough pricing power to absorb higher component costs and pass them on, are gaining ground by default – not because their devices have become dramatically better, but because the economics of making a competitive device at any lower price point have become dramatically harder.
This is the quiet structural injury of a chip shortage that does not resolve cleanly: it rewards scale and punishes competition. The brands most exposed are the challengers – the Nothing Phones, the mid-tier Android lines, the regional manufacturers in South and Southeast Asia that have spent years building market share on the premise of diminishing component costs. That premise has collapsed.
Europe, characteristically, has responded with policy. The European Commission formally presented the Chips Act 2.0 on June 3, the most significant revision of the bloc’s semiconductor strategy since the original 2023 legislation. The package gives Brussels emergency powers to override commercial contracts between chipmakers and their customers during a supply crisis, compel manufacturers to prioritise EU-designated critical orders, and accelerate the permitting procedures that have made building semiconductor facilities in Europe materially slower and more expensive than comparable projects in Asia or the United States. The Commission acknowledged in its own documentation that permitting timelines in the EU run an average of 7.5 months longer than in competing jurisdictions – a gap that adds roughly three percent to the total cost of a new facility before a single chip has been produced.
The ambition is real. So is the distance between ambition and outcome. The original Chips Act targeted a 20 percent share of global semiconductor production for Europe by 2030. That target is already effectively abandoned: Intel cancelled its planned Magdeburg plant, Europe’s share of global chip production remains below 10 percent, and the €52 billion mobilised under the first act, while substantial, has not yet produced the advanced fabrication capacity the bloc needs. The Chips Act 2.0 enters the European legislative process now, with the Council deadline set for September. The phone in your pocket for the next two Christmas cycles will be priced under a supply regime the new law cannot yet touch.
Nvidia, meanwhile, has not waited for regulatory frameworks to shape its supply chain. The company struck an agreement with South Korea’s SK Hynix to develop next-generation memory architecture for AI data centres and consumer devices simultaneously – a partnership that positions Hynix as a critical supplier to both the AI infrastructure build-out and whatever high-end smartphone chips come next. The arrangement illustrates how the industry is already reordering itself around the AI demand signal: memory designed first for the data centre, adapted later for the handset, on a timeline and at a price that reflects the data centre’s priority.
Micron surpassed a $1 trillion market capitalisation in May, with analysts attributing the milestone to structural changes AI has driven to the global memory industry. That valuation tells its own story: the companies making memory have never been worth more, and the consumers buying the devices that memory goes into have never paid more for it. The gap between those two facts is the market Pei described – one where smartphone brands, whatever their size or ambition, have no meaningful leverage over the single component that now defines their cost structure.
Pei’s advice to consumers was blunt: if you plan to upgrade, do not wait. What he left unspoken is the more uncomfortable implication – that buying sooner may not be cheaper, only less expensive than waiting. The question of when the shortage eases does not have a reliable answer. The AI build-out that is consuming chip capacity is accelerating, not plateauing, and no policy measure currently in place in any major economy is designed to redirect that demand toward the consumer electronics industry. Europe’s Chips Act 2.0 is designed to build supply. It cannot, by design, suppress the demand that is eating it.
