TSMC is turning into a victim of its own success as the world's preferred chip foundry, leaving its heretofore prized customers such as Apple in a bind of sorts as they suddenly find themselves crowded out by AI hyperscalers.
In its frustration, Apple is now reportedly exploring the possibility of dividing up its silicon load between Samsung, Intel, and TSMC rather than remaining largely TSMC-exclusive.
Apple is looking for contingencies by tentatively probing Intel and Samsung as additional vectors for manufacturing its custom chips
According to Bloomberg's Mark Gurman, Apple has already held "early-stage talks" with Intel for using its chipmaking services. Concurrently, a number of Apple executives have also visited Samsung's under-construction Texas fab for a preliminary exploration of its viability for Apple silicon.
Critically, Apple is exploring options for the "main processors of its devices," and not just low-volume silicon in some of its more niche products.
Do note that in recent months, both GF Securities and DigiTimes have disclosed that Apple might opt for Intel's 18A-P process for its lowest-end M-series chips that are expected to ship in 2027, as well as non-Pro iPhone chips in 2028. GF Securities also went a step further by noting that Apple's bespoke ASIC - expected to launch either in 2027 or 2028 - will leverage Intel's EMIB packaging.
In fact, Apple has reportedly signed an NDA with Intel and already procured PDK samples of its advanced 18A-P process for evaluation purposes. Do note that Intel's 18A-P process is its first node to support Foveros Direct 3D hybrid bonding, which allows for stacking multiple chiplets through TSVs.
Even so, Apple now appears to be going beyond one-off, product-based chip fabrication partnerships, and actively searching for viable ex-TSMC options.
This comes as Apple conceded during its latest earnings call last week that it'll take several months for the supply of Mac Studio and Mac mini devices to catch up with demand, pointing to agentic AI as a major source of demand for these products.
Apple had gone on to note that Mac-related supply constraints were not due to a shortage of memory resources but rather because of the limited availability of advanced nodes at TSMC.
The situation has become so extreme that Apple has now pulled the base variant of the Mac mini, which sported the M4 chip, 16GB RAM, 256GB of storage, and a retail price of $599, from its configurator. Similarly, the Mac Studio with 512GB in storage has also been pulled.
While TSMC is trying to expand its 3nm capacity by repurposing some of the 4nm lines that now lie vacant due to reduced demand from low- and mid-tier smartphone SoCs, the resulting 25,000 wafers per month in additional capacity is seemingly not enough to unclog lines choked with oncoming demand, especially from AI hyperscalers, who are willing to pay premia of between 50 and 100 percent for rush orders.
In fact, according to TrendForce, TSMC is only adding new 3nm capacity of around 3,000 wafers per month, while adding 55,000 wafers per month in new 2nm capacity. Clearly, its focus has already shifted towards the new 2nm node.
While Apple has reportedly reserved around half of TSMC's 2nm lines, its 2nm-based products won't launch until well into H2 2026. Meanwhile, Apple desperately needs additional 3nm-based chip making capacity to keep its product lines chugging along. It is for this reason that Apple is now exploring the use of Samsung and Intel fabs even as it remains unsure of the quality of the resulting chips. This is a classic 'damned if you do, damned if you do not' type of situation.
Follow Wccftech on Google to get more of our news coverage in your feeds.
