TSMC Price Increase 2026: Apple and NVIDIA Reportedly Face Higher Prices Across Every Advanced Node — 74% of Wafer Revenue

The first report came from Culpium, the independent outlet run by former Bloomberg journalist Tim Culpan. Culpium published the exclusive on June 23, 2026, and major tech outlets including Tom's Hardware followed. TSMC has not issued a denial — but it has not officially confirmed the story either. As it stands, this is reporting based on anonymous sources.

What was reported

Here is what Culpium reported, in summary (Culpium, June 23, 2026; Tom's Hardware).

The key point is the breadth of the increase. Price hikes on advanced nodes (the newest process generations in the miniaturization race) have been reported repeatedly over the years, but this time the scope reportedly extends beyond the latest 3nm and 5nm generations to include 7nm, a process that has been in service for years. Added together, nodes from 7nm onward account for roughly 74-75% of TSMC's wafer revenue. If the report is accurate, three-quarters of the company's revenue base is being repriced.

Note that no across-the-board percentage for the increase has been reported. Hikes on legacy nodes (mature processes — older production lines) remain at the level of scattered customer accounts and are unconfirmed.

Why can TSMC raise prices?

The answer is simple: customers have almost no alternative. Only a handful of foundries (contract chip manufacturers) in the world can mass-produce leading-edge logic chips, and only TSMC has consistently met customer requirements on performance, yield and volume all at once. How TSMC pulled this far ahead is covered in Why TSMC matters: the one company the world's phones and AI cannot exist without.

Supply and demand are also on TSMC's side. In the latest 2nm generation, TSMC stands alone, and its 2026 capacity is reported as effectively sold out to Apple, NVIDIA, Qualcomm and AMD. A 2nm wafer is said to cost about $30,000 — and customers are still fighting over allocation. The full picture of this supply crunch is laid out in 2nm was sold out before it was ever made: where TSMC, Intel, Samsung and Rapidus stand (in Japanese). A shop whose shelves are sold out rewriting its price tags — as pricing dynamics go, it is no more complicated than that.

There is a cost story too. The investment required for a leading-edge fab keeps ballooning, and overseas expansion — starting with the Arizona fabs — is considered more expensive than production in Taiwan. Why fab construction costs have exploded is covered in why a leading-edge chip fab now costs over $15 billion (in Japanese). Whether a company can pass ever-rising costs through to customers is the most direct test of its bargaining power. This report suggests TSMC answered that test with an aggressive, all-nodes-at-once move.

One more factor is easy to miss: switching costs. A chip design is built around a specific manufacturing process, so moving to another company's fab mid-stream takes something close to a full redesign in effort and time. Even a customer unhappy with the increase cannot realistically walk away from TSMC in the short term.

The champion raises prices — what is the challenger doing?

A contrasting move is playing out in Japan. Rapidus CEO Atsuyoshi Koike said on July 8, 2026 that the company will price its 2nm generation at a level at least equal to, or slightly below, TSMC's (Nikkei, July 8, 2026; the reference price is 3 to 3.5 million yen per wafer). The details of that pricing strategy are covered in Rapidus puts a price on 2nm: ¥3-3.5 million per wafer. The champion is reported to be notifying customers of higher prices while the challenger hangs out a lower price tag, angling to catch any customer discontent — a contrast that lays the structure of the foundry market bare.

That said, Rapidus's 2nm has not yet reached volume production. Even at equal prices, whether customers would move away from TSMC — with its proven yield and supply volume — is a separate question, and in the short term Rapidus does not threaten TSMC's pricing power.

How should investors read this?

If the report is accurate, it is a direct positive for TSMC's gross margin. Raise prices on roughly 74-75% of wafer revenue, and revenue and profit stack up without shipping a single additional wafer. And because the increase comes amid a supply crunch, the risk of orders walking away is limited.

For fabless customers (design specialists with no factories), it is a cost headwind. Apple, NVIDIA, AMD and Qualcomm each face a choice: pass higher manufacturing costs into their own product prices, or absorb them in their margins. Products with strong demand, like AI chips, can pass costs through easily; consumer products like smartphones cannot — watch how the impact varies customer by customer.

The first checkpoint — Q2 results announced on July 16, 2026 — delivered a 67.7% gross margin, above the top of guidance (covered in detail in TSMC's record Q2: net profit up 77%, 66% of revenue from AI servers, in Japanese). But that reflects performance before any price increase. If the hikes proceed as reported, their effect will show up mainly in future contracts, so gross-margin guidance for the second half and management commentary remain the checkpoints to watch.

What are the caveats?

The point this article most wants to stress: every report traces back to one outlet. The original story is Culpium's exclusive, and the follow-on coverage — including Tom's Hardware — cites Culpium. There is, at this point, no independent corroboration from multiple sources. Single-outlet stories based on anonymous sources are often directionally right and wrong in the details.

The numbers also carry open questions. No across-the-board percentage for the increase has been reported, and this article deliberately cites no specific hike rate. Spillover to legacy nodes remains at the level of scattered customer accounts. TSMC's silence is circumstantial evidence — but silence is not confirmation.

The precise state of knowledge, then, is this: TSMC is reported to have notified customers of price increases across all advanced nodes, and TSMC has not denied it. The reliable way to settle whether it is true is to track gross margins through upcoming earnings reports (Culpium, June 23, 2026; Tom's Hardware).