This site may earn affiliate commissions from the links on this folio. Terms of apply.

With GlobalFoundries having appear that it volition abolish its 7nm deployments and terminate evolution of leading-edge FinFETs, the number of companies with leading-edge fabs has fallen to three. But of the three, Samsung is a bit of a special example — while it has foundry customers and has competed for Apple and Qualcomm business organization in the by, it also builds its own silicon to support its mobile phone efforts. It's a fleck of a hybrid — half Intel, half TSMC, but with a fraction the sales of either.

According to EETimes, GlobalFoundries is leaving the leading edge because the economics simply don't brand sense. GlobalFoundries was facing a $2B – $4B additional expenditure to ramp its wafer production to the xl,000 – fifty,000 wafer starts per month that are required to earn back a return on investment on the node. Co-ordinate to Caulfield, most of GlobalFoundries' customers have no plans for 7nm chips. Need for 14/16nm across the industry was half of what overall demand for 28nm, and 7nm could be one-half again equally low as 14nm.

"When we look out to 2022, ii-thirds of the foundry market place will be in nodes at 12nm and above, and then it's non similar nosotros are conceding a big part of this market place," Caulfield told EETimes.

None of these statements are surprising. If you're a computing enthusiast you lot live, in a very real sense, in a sort-of semiconductor bubble. The chips that we collectively care near — high-performance processors built into smartphones and PCs, including GPUs — are just a fraction of the total semiconductor industry. They constitute the highest-performing and highest-value products in the market on a per-unit basis, but in terms of unit shipments, they're absolutely dwarfed past the combined volume of microcontrollers and small SoCs for various embedded and low-power devices. You can probably name most of the companies that purchase and build these type of chips off the top of your head: Apple, AMD, Intel, Qualcomm, Samsung, MediaTek (to an extent), HiSilicon, Nvidia, Xilinx. Some of these companies, similar Samsung and Intel, have their ain fabs and will utilize their ain production for ramping next-generation nodes.

As the price of ramping next-generation nodes increases, information technology ways those costs must exist built-in by either the foundry, its customers, or both. And this may exist where the blossom began coming off the GlobalFoundries rose and the GF – AMD partnership. Back in 2022, AMD signed a v-yr WSA agreement with GF that was supposed to comprehend both the 14nm and 7nm nodes, with specific allowance for AMD to seek manufacturing flexibility if required, but with a articulate center towards guaranteeing that AMD would take a ramp on 7nm. Equally role of this negotion, AMD disclosed that it expected to buy $650M of wafers from GlobalFoundries in 2022. It later amended that amount in its form 10-K to note that it actually spent $1.1B on wafers from GlobalFoundries in 2022 and $700M in 2022.

AMD-WSA-Agreement

But — and here'southward the actually big "but" — co-ordinate to EETimes, AMD was also GlobalFoundries sole 7nm customer. That'south critically important when you consider that in 2022, much of AMD's total business was running out of GF. GlobalFoundries builds Polaris, it builds Vega, it builds AMD'southward entire Bulldozer family (which the company is still shipping), and it builds Ryzen. Apart from the consoles, GF had the entirety of AMD's concern, minus whatever 28nm GPUs the company might have shipped from TSMC in 2022. But the days when a single customer can bulldoze a leading-edge node are over, and some of you may remember a few years dorsum when AMD announced it would no longer pay for custom nodes. When you've got one customer, you're effectively building a custom node no matter what.

This shift marks a change in the fundamentals of foundry dynamics. In the former days, you could deploy a new node for cut-edge customers, knowing that y'all'd brand money off information technology for decades to come. TSMC, for example, earned 27% of its revenue in 2022 from nodes at 65nm or above, and 65nm is more than a decade erstwhile. More than than half of TSMC's acquirement all the same comes from the 28nm node and above. But this model assumes — in fact, requires — that customers regularly proceed to newer nodes from older ones. But only sure types of silicon benefit from FinFETs in the first place, which means the benefits of dice shrinks are themselves shrinking every generation beyond the entire industry.

The Opposite of What AMD Wanted

The original goal of spinning off GlobalFoundries was to challenge TSMC at the top of the foundry heap and compete for leading-edge customers. The specific point of this strategy was to amortize the price of moving to leading-edge nodes amid many customers rather than attempting to carry it entirely inside AMD. Paying the cost of a 7nm dice compress isn't as bad as paying the cost of an entire foundry, but with the cost of each new node rising, the $2-$4B in boosted investment required for GF to brand 7nm work dwarfs what AMD would pay GF for wafers in the near term.

It'southward articulate that, for whatever reason, GlobalFoundries wasn't able to convince potential customers to accept a chance on its ain 7nm every bit opposed to TSMC's. After years of struggling to compete at the leading edge and having to cancel its own efforts to do so at 14nm, the company's owners want to meet it turn a profit by focusing on edifice parts for customers that are profitable now equally opposed to chasing afterward Samsung, Intel, and TSMC in a futile attempt to earn customers and market share. It'southward an understandable conclusion, but an ominous one. As the number of customers for new nodes continues to decline, single customer wins become overwhelmingly of import. Samsung tin can keep its fabs fed with its own production sales, but the trend line here is impossible to miss:

Leading edge foundries over time

Nosotros used to have 19 firms competing at the leading edge. STMicro appear it would lean on foundries for production after 14nm and never put that node into production. Currently, there are v firms offering 14nm, 4 of which offered information technology as a leading-edge node: Samsung, TSMC, Intel, GlobalFoundries (leading-border, with GF using Samsung's IP) and and so UMC, which began offering 14nm as a new capability this summer. We may still see secondary foundries deploy on nodes like 14nm one time they are no longer leading edge, only even this is uncertain. The cost structures are squeezing companies out of the market. Whether solutions similar FDX can provide alternatives remains to be seen.

The goal of the GlobalFoundries spinoff was to create new customer opportunities and give AMD a partner in a more reliable position to deliver regularly foundry engineering science improvements. That dream largely failed to materialize. In all honesty, a difference from GF is probably best for both companies — even an incredibly aggressive Epyc and Ryzen sales ramp would not have supported AMD's 7nm needs within the necessary time frames. Meanwhile, GF wants to focus on the market segments where information technology might really take a chance of making a mark, and it's hard to fault that. But the continuing pass up in full client base of operations, alongside steep toll increases is an ominious preview of a futurity in which we may have just one company building on the leading edge — or mayhap no companies at all.

Now Read: GlobalFoundries Radically Restructures, Kills 7nm, AMD Moves All 7nm Production to TSMC, and Troubling News: AMD Moves 7nm GPUs Back to TSMC