The Endpoint Refresh Cycle Just Got a Lot More Expensive, and Japan Is Feeling It Twice

A global memory shortage and a weak yen are colliding in Japan. For IT leaders here, they don't just add up. They multiply.

If you manage endpoints for a Japan-based organization, your next hardware refresh cycle is about to test every assumption in your budget model. Two forces are colliding right now, and for IT leaders here, they don't just add up. They multiply.

The first is the global memory shortage. AI data center buildouts have absorbed the lion's share of global DRAM and NAND production, and the numbers are stark. Contract prices for standard memory have climbed by well over 50% in a single quarter this year, with some components more than doubling versus 2025 pricing. Analysts don't expect meaningful relief until 2027 or 2028. Every device that ships with memory or storage inherits that cost, whether the vendor absorbs it quietly or passes it straight through in the price list.

The second is the yen. USD/JPY spent most of the summer sliding toward ¥163 to ¥164, a 40-year low, before Japan and the US carried out a rare joint intervention in early August, the first coordinated yen-buying operation between the two countries since 1998. That pulled the rate back to around ¥157 to ¥158 as of this week, though the yen has already given back close to half the intervention's initial gains, and traders are watching for another round if it drifts back toward ¥160. The pressure behind the slide, wide interest rate differentials and Japan's fiscal position, hasn't gone away, just the acute spike. For organizations here, even the improved rate is still a second markup layered directly on top of the first. Hardware priced globally in dollars gets more expensive to convert into yen at exactly the same moment the dollar price itself is rising.

For standard frontline endpoints, the laptops and desktops that make up the bulk of most device fleets, this shows up as straightforward sticker shock at renewal or refresh time. Vendors who used to give you room to negotiate on list price now have less room to give, because their own input costs have moved. If your refresh cadence and budget were built on 2024 or 2025 pricing assumptions, they need a hard look before the next cycle, not after the quotes come back.

For specialized users, the pressure compounds further. Developers and data scientists depend on exactly the components in shortest supply, and so does anyone running AI or ML workloads locally. High RAM and large, fast storage are the obvious ones. GPUs face their own demand crunch on top of that.

Apple's own lineup is a useful case study. Over the past several months, Apple has quietly pulled the highest-RAM configurations of both the Mac Studio and Mac mini from sale. There's no 256GB or 512GB Mac Studio anymore, and no 32GB or 64GB Mac mini in some configurations, with Apple citing the same memory constraints. Lead times on what remains have stretched to two or three months in some cases. If your organization standardizes on high-spec hardware for specialized roles, the device you speced eighteen months ago may simply not be orderable today, regardless of budget.

Consumer behavior is already showing what this looks like once it reaches the end user. When Apple raised iPhone prices in Japan by 8% to 11% in mid-July, citing the weak yen, one of the country's largest used-device marketplaces saw order volume more than double within hours of the announcement, a spike its own executives called rare, comparable only to a similar price hike four years earlier. It's worth watching whether the same substitution shows up in the corporate channel: more organizations extending laptop life or buying certified refurbished instead of new. If it does, it will look a lot like this.

A few practical shifts are worth considering for anyone managing endpoint budgets in Japan right now. Model FX volatility explicitly instead of tracking only USD list price, and build cost scenarios across a range of exchange rates instead of a single point estimate. This is also the moment to find out what rate finance is actually planning to use for next fiscal year's budget. Most multinationals set a single hedged rate at the start of the year and hold procurement to it, which means the cost pressure doesn't arrive smoothly. It shows up as a step change whenever that rate gets reset, and if next year's rate gets locked in anywhere near the recent lows, or drifts back down again once this month's intervention effect fades, the gap between the old hedge and the new one lands all at once instead of gradually. Segment your refresh cadence more deliberately: frontline devices may tolerate an extended cycle of four to five years instead of three, given how much of the current cost increase comes from memory and storage, not CPU-driven obsolescence. Reconsider where specialized compute actually needs to live. For GPU- and memory-hungry workloads, cloud workstations or centralized compute may now work out better than provisioning high-spec local hardware that's back-ordered or unavailable at any price. Engage vendors and distributors earlier than usual, given how much lead times have lengthened, and treat multi-year framework agreements that lock in pricing and allocation as worth the commitment even if it means signing further out than you're used to. It's also worth reassessing "gold standard" specs. If your standard build was set when memory was cheap, ask whether every role actually needs the spec it was assigned.

This isn't unique to Japan in its origin. The memory shortage is a global, AI-driven supply story. But the yen, even after this month's intervention, means IT organizations here are absorbing a cost that peers in dollar-denominated markets aren't facing at the same scale. Budgets that were already tight going into this cycle are getting tighter, and vendors are already pricing that reality into their quotes.

Curious how other IT leaders here are adjusting refresh cycles and procurement strategy in response. I'd welcome the conversation.