As consumers are grappling with significant price hikes of smartphones due to the ongoing memory crisis, it has now come to light that TSMC is also hiking its wafer prices. According to Taiwan's Commercial Times, TSMC is preparing to raise its 3nm wafer prices by up to 15% in the second half of 2026. Due to relentless AI demand, TSMC's 3nm capacity is at full capacity, hence, the price increase.
AI is Eating TSMC's 3nm Capacity and Phone Buyers Will Pay the Price
In the past, TSMC's 3nm capacity was mostly used by smartphone and PC chip makers like Apple, Qualcomm and MediaTek. But due to the AI boom, Nvidia, AMD, Google, Amazon and multiple cloud providers are queuing up at TSMC's door to produce their AI accelerators and custom ASICs.
The foundry has increased monthly 3nm output from around 130,000 wafers to 175,000 wafers, but it's still not enough. According to industry reports, AI demand is outpacing supply by nearly three times.
Currently, a 3nm wafer costs around $18,000-$20,000 and TSMC's upcoming 2nm wafers are expected to exceed $30,000 each. A 15% bump on a $20,000 wafer is roughly $3,000 more, which will be ultimately passed down to end consumers.
For phone buyers, this is absolutely bad news. The Snapdragon 8 Elite Gen 5 and MediaTek Dimensity 9500 are both fabricated on TSMC's enhanced 3nm (N3P) node. The cheaper Snapdragon 8 Gen 5 and Dimensity 9500s are also built on the same 3nm node.
Note that even before this price hike, MediaTek was paying 24% more for the Dimensity 9500 wafer while Qualcomm was taking a 16% hit on the Snapdragon 8 Elite Gen 5. Now, with this fresh 15% hike, flagship smartphones are expected to cost even more.
That's the reason Qualcomm CEO Cristiano Amon flew to Seoul to meet Samsung Foundry president last month. According to reports, Qualcomm is considering Samsung Foundry to produce its upcoming Snapdragon 8 Elite Gen 6 chip on Samsung's 2nm (SF2) node. If it happens as discussed, Qualcomm would be returning to Samsung after nearly five years.
Apple is also doing the same thing and exploring alternative fabs for low-end iPhone chips. In May 2026, the Wall Street Journal reported that Apple and Intel have reached a preliminary chip-making agreement. Apple is reportedly going to use Intel's 18A node to produce low-end M-series and iPhone chips, possibly by 2027/2028.
Of course, Apple is still using TSMC's 2nm node for its Pro chips and has already booked 50% of 2nm capacity for 2026. Qualcomm may also keep the Snapdragon 8 Elite Gen 6 Pro on TSMC's 2nm node and shift the standard variant to Samsung.
But one thing is clear: only flagship chips would be produced on cutting-edge nodes at TSMC and rest of the chips might move to other fabs. As a result, flagship phones powered by flagship chips would be much more expensive in 2026 and beyond.
