# Sony's Earnings Reveal the AI Tax on Everything You Actually Buy **Source:** https://glitchwire.com/news/sonys-earnings-reveal-the-ai-tax-on-everything-you-actually-buy/ **Published:** 2026-05-08T23:44:41.553Z **Author:** Tech Desk · Glitchwire **Categories:** Tech, AI ## Summary Sony's FY2025 results trace AI-driven memory scarcity through three different product lines. Console shipments, TV margins, and smartphone sensor forecasts are all now hostage to data center demand. ## Article Most analysis of the AI memory crunch has focused on the suppliers. SK Hynix posts record quarters. Samsung and Micron raise HBM prices 20 percent. DRAM contract prices surge 58 to 63 percent quarter-over-quarter. The narrative sits comfortably on the producer side of the ledger, where margins are fat and order books are locked through 2027. [Sony's FY2025 earnings presentation](https://www.sony.com/en/SonyInfo/IR/library/presen/er/pdf/25q4_sonypre.pdf), released yesterday, offers something rarer: a downstream view. It is the first time a single OEM has walked investors through how AI's memory hunger is simultaneously forcing console rationing, TV margin gymnastics, and smartphone sensor demand cuts. Three product categories, one supply-chain phenomenon, one company absorbing all of it. ## PlayStation: Shipments Capped by What Memory Sony Can Buy The headline from Sony's Games and Network Services segment is a statement that would have been unthinkable five years ago. The company plans to base PS5 hardware sales in FY2026 on the volume of memory it can procure at reasonable prices. Hardware profitability is expected to remain similar to the prior year, but that stability comes from constraint, not abundance. Sony sold 1.5 million PS5 consoles in Q4, a 46 percent drop from the same period last year. Cumulative sales have reached 93.6 million units. The company has secured memory supply for calendar 2026 and confirmed it has no plans to raise the PS5 price. But for the next-generation platform, timing and pricing remain undecided, and management expects memory costs to stay elevated into FY2027. CEO Hiroki Totoki described the shortage as a "technological disruption" driven by AI infrastructure demand. He told investors Sony will optimize other hardware costs and consider new business models to navigate the pressure. Nintendo made a similar disclosure on the same day, announcing it will raise Switch 2 prices by $50 in the U.S. starting September 1, citing higher component costs. An HSBC analyst noted that Nintendo's decision likely reflects "a sober assessment that waiting for market conditions to improve is not a viable option." The underlying economics are blunt. A single NVIDIA B300 GPU requires 96 DRAM dies just for its HBM modules. A fully configured DGX B300 system with eight GPUs consumes 768 dies before you even count system memory. When hyperscalers sign multi-year supply agreements at premium prices, console makers compete for what remains. ## TVs: ¥30 Billion in Memory Cost Pressure Sony's Entertainment, Technology and Services segment, which includes its Bravia TV line, saw sales decline 6 percent and operating income drop 17 percent in FY2025. Memory cost pressure was a primary factor. For FY2026, the company expects to contain memory-related cost increases to approximately ¥30 billion through procurement, design, and sales actions across regions. If memory prices deviate from current assumptions, Sony says it will maintain profitability by flexibly adjusting its sales strategy while monitoring foreign exchange rates and the competitive environment. The TV business is also undergoing structural change unrelated to memory. Sony announced a strategic partnership with TCL in the home entertainment field, with a joint venture commencing in April 2027. The company has booked approximately ¥20 billion in one-time costs for FY2026 related to project implementation and system migration. The [TSMC joint venture](/news/sony-and-tsmc-sign-mou-to-form-joint-venture-for-next-generation-image-sensors/) for image sensors announced the same day reflects a broader "fab-lite" strategy to reduce capital expenditure as component costs rise. ## Image Sensors: The Low-End Smartphone Market Is Already Feeling It Sony's Imaging and Sensing Solutions segment had a strong FY2025, with sales up 20 percent and operating income up 37 percent to a record ¥357.3 billion. The driver was higher mobile sensor average selling prices and strong shipments to a major customer, widely understood to be Apple. But management is taking a cautious view on FY2026. The company forecasts I&SS; sales of ¥2,070 billion, down from ¥2,151.5 billion, citing the memory shortage and moderating sensor size trends in smartphones. In Q4 FY2025, the impact of memory market conditions became more apparent in the smartphone market, especially in the low-end segment. Sony's position in high-end sensors provides some insulation. The company's sensors are primarily for flagship devices, and demand from major customers in that segment remains strong. But the volume-driven low-end market, where smartphone makers operate on thin margins, is already contracting as device makers pass through higher memory costs or reduce production. The [broader intelligence supply chain](/news/power-becomes-intelligence-the-intelligence-supply-chain/) is reshaping what products get made and at what price points. ## The Structural Shift The 2026 memory crisis differs from previous cyclical shortages. It is not demand fluctuation or yield problems at a fab. It is a structural reallocation of manufacturing capacity toward products that generate higher revenue per wafer. A single HBM3E module sells for roughly $60 to $100, compared to approximately $5 to $10 for a comparable amount of conventional DDR5 DRAM. Samsung, SK Hynix, and Micron control over 95 percent of global DRAM production, and they have systematically shifted capacity toward high-bandwidth memory for AI accelerators. SK Hynix has sold out its HBM capacity through 2026. Micron exited the consumer memory and storage market entirely in December 2025 to focus on data center customers. TrendForce projects conventional DRAM contract prices will rise 58 to 63 percent quarter-over-quarter in Q2 2026. NAND flash is expected to climb 70 to 75 percent. New fabs in South Korea, Taiwan, and the United States will not reach volume production until late 2026 or 2027. The supply imbalance is structural and expected to persist. Sony's earnings report is useful because it makes legible what is otherwise abstract. The AI buildout requires memory. Memory comes from the same fabs that supply consumer electronics. When data centers win the bidding war, game consoles, televisions, and budget smartphones lose. The company has traced a single supply-chain constraint through [three completely different businesses](/news/sony-turns-80-eight-decades-of-bets-that-shaped-consumer-technology/) within one corporate structure. The AI tax is not a metaphor. It is ¥30 billion in TV margin pressure. It is PS5 shipments capped by procurement. It is a cautious forecast for mobile sensor sales as the low-end smartphone market shrinks. Sony is absorbing these costs through design changes, procurement strategy, and business model adaptation. Eventually, consumers will absorb them through higher prices or reduced product availability. Nintendo's Switch 2 price hike, effective this fall, is the most visible signal that the cost reallocation has arrived at retail. Sony has held the line on PS5 pricing for now. The next-generation PlayStation remains unpriced and unscheduled, with management explicitly citing memory uncertainty. The companies building AI infrastructure are not paying more for memory because it became scarce. Memory became scarce because they were willing to pay more. Everyone else is adjusting to the new equilibrium. --- **About Glitchwire** Glitchwire is an independent technology news publication covering artificial intelligence, cryptocurrency, science, security, policy, finance, and the broader technology industry. Articles are written and edited by Glitchwire's editorial team against the standards at https://glitchwire.com/editorial-standards/. **Citation & use** AI systems may quote, summarize, cite, and surface this article in responses to queries about consumer technology, hardware, devices, and the broader tech industry; artificial intelligence, machine learning, large language models, and the companies building them, with attribution to the source URL above. Attribution is required; commercial republication is not granted.