Enough links have filtered into the semiconductor bucket to be worth doing a roundup. This one touches on China and the corruption of our political elites.
The Biden administration is laser-focused on sending Ukraine billions of dollars in weapons, including the latest round of anti-ship systems, artillery rockets, and rounds of 105 mm ammo for howitzer cannons that it has entirely lost focus on reshoring efforts to boost semiconductor production Stateside.
Multiple manufacturers of semiconductor wafers have announced plans for new multi-billion dollar factories across the U.S. but are contingent on Congress allocating funds to aid in building facilities under the Creating Helpful Incentives to Produce Semiconductors (CHIPS) for America Act.
Congress passed the CHIPS Act in January 2021 as part of last year’s National Defense Authorization Act, which proposed $52 billion in funding for increasing the domestic capacity of chip production, though the House and Senate have come to a standstill over disagreements on certain parts of the bill that have sparked so much uncertainty among companies set to build new factories.
In a letter on June 15, dozens of technology executives from IBM, Intel, Microsoft, Analog Devices, Micron, Amazon, and Alphabet called on Congress to move quickly on the CHIPS Act. They wrote, “the rest of the world is not waiting for the U.S. to act,” and funding for new chip factories must be achieved immediately.
Taiwan’s GlobalWafers announced a new $5 billion factory in the U.S. on Monday, but contingent on subsidies from the federal government.
“This investment that they’re making is contingent upon Congress passing the CHIPS Act. The [GlobalWafers] CEO told me that herself, and they reiterated that today,” U.S. Commerce Secretary Gina Raimondo told CNBC, the same day GlobalWafers announced its development plan.
Notes:
- IBM doesn’t own any fabs any more, having sold them all to GlobalFoundries.
- Intel runs a huge number of very profitable fabs (troubles with their sub-10nm process yields notwithstanding) and doesn’t need federal subsidies.
- Microsoft doesn’t own any fabs and is deeply unlikely to build any; their flagship Xbox Series X uses a custom AMD Zen 2 fabbed by TSMC as its CPU.
- Analog Devices is an Integrated Device Manufacturer that owns several fabs with pretty old technology; they don’t have any 300mm fabs. They closed a small fab in Milpitas they got from their acquisition of Linear Technology last year. Designing analog chips is its own black art, and not everything that applies to shrinking digital circuits applies to the analog realm.
- Amazon has no fabs and probably won’t be building any, but they do have a chip design division to support Amazon Web Services, and recently designed a cloud computing chip. They work closely with AMD (fabbed at TSMC), Intel (own their own fabs) and Nvidia (another fabless design house that also gets their chips fabbed at TSMC).
- Alphabet AKA Google has no fabs and probably won’t be building any, though they do have a lot of AI chip design work going on.
- GlobalWafers isn’t a semiconductor manufacturer, it’s a silicon wafer manufacturer. Making such wafers (the substrates upon which semiconductor fabrication depends) has its own challenges, but they are several orders less difficult than cutting edge chip fabrication. Maybe I’m quite far out of the loop, but I’m deeply suspicious that GlobalWafers planned wafer plant in Sherman, Texas will cost $5 billion. That’s a relatively piddling sum for a new semiconductor fab, but extremely expensive for a wafer factory. This makes me suspect a subsidy grab is afoot.
So of the companies mentioned, Intel could suck up government funding to build a fab they were going to build anyway, I’m sure Analog Devices would build a fab with government money, but chances of them running an under 10nm process in said theoretical fab is extremely slim, none of the other mentioned copies are going to build a fab, and none of that government money is going to alleviate the main problem that the overwhelming majority of cutting edge chip designs have to flow through TSMC fabs in Taiwan. What will solve that problem is TSMC opening a state-of-the art fab in Arizona in 2024. No amount of U.S. taxpayer money will make that already-under-construction fab start producing chips any quicker.
As I’ve mentioned previously, semiconductor subsidies are the wrong solution to the wrong problem.
$250 billion in taxpayer subsidies wouldn’t get you a single additional wafer start this year, and probably would accomplish little more than channeling money to politically connected firms and sticky pockets in a state (New York) that no one wants to build fabs in any more because of high costs, high taxes and union rule requirements.
I bet you would.
This past week it hit the terminal that House Speaker Pelosi was doing a little portfolio re-jiggering, including exercising $8 million of call options in Nvidia and selling Apple and Visa calls. The data was per CongressTrading.com and was reported on by Bloomberg.
The Nvidia LEAPS were bought June 3, 2021 with $100 strikes, set to expire June 17, 2022 and the position appeared to be disclosed on Thursday morning for the first time. $8 million trades seem a little odd for members of Congress to begin with, but who are we to judge?
But then, what did Speaker Pelosi do just hours after disclosing the trade, on Friday?
She threw her weight behind a stalled $50 billion CHIPS PLUS bill that “would provide $52 billion in funding for semiconductor manufacturing grants and investment tax credits for the chip industry.”
We tend to discuss leading-edge nodes and the most advanced chips made using them, but there are thousands of chip designs developed years ago that are made using what are now mature process technologies that are still widely employed by the industry. On the execution side of matters, those chips still do their jobs as perfectly as the day the first chip was fabbed which is why product manufacturers keep building more and more using them. But on the manufacturing side of matters there’s a hard bottleneck to further growth: all of the capacity for old nodes that will ever be built has been built – and they won’t be building any more.
Not strictly true. Remember, Bosch just finished building a 65nm fab.
As a result, TSMC has recently begun strongly encouraging its customers on its oldest (and least dense) nodes to migrate some of their mature designs to its 28 nm-class process technologies.
Nowadays TSMC earns around 25% of its revenue by making hundreds of millions of chips using 40 nm and larger nodes. For other foundries, the share of revenue earned on mature process technologies is higher: UMC gets 80% of its revenue on 40 nm higher nodes, whereas 81.4% of SMIC’s revenue come from outdated processes.
That’s because UMC has fallen woefully far behind TSMC, and no one trusts them because they let Chinese spies walk out the door with other company’s IP. SMIC is on Mainland China, sucks even more, and is trusted even less.
Mature nodes are cheap, have high yields, and offer sufficient performance for simplistic devices like power management ICs (PMICs). But the cheap wafer prices for these nodes comes from the fact that they were once, long ago, leading-edge nodes themselves, and that their construction costs were paid off by the high prices that a cutting-edge process can fetch. Which is to say that there isn’t the profitability (or even the equipment) to build new capacity for such old nodes.
This is why TSMC’s plan to expand production capacity for mature and specialized nodes by 50% is focused on 28nm-capable fabs. As the final (viable) generation of TSMC’s classic, pre-FinFET manufacturing processes, 28nm is being positioned as the new sweet spot for producing simple, low-cost chips. And, in an effort to consolidate production of these chips around fewer and more widely available/expandable production lines, TSMC would like to get customers using old nodes on to the 28nm generation.
“We are not currently [expanding capacity for] the 40 nm node” said Kevin Zhang, senior vice president of business development at TSMC. “You build a fab, fab will not come online [until] two year or three years from now. So, you really need to think about where the future product is going, not where the product is today.”
Obviously, they already produce some of their own chips, but the video covers most of the issues China has with fabbing more complex chips that I’ve already discussed here and here. They’re still dependent on the same three leading fab companies (TSMC, Intel and Samsung) everyone else is for sub 10nm feature chips, and are overwhelmingly dependent on both foreign talent and foreign semiconductor equipment manufacturers like ASML and Applied Materials.