- “America has to bet like this.” , Sen. Mark Warner, Chips Act co-sponsor
The “Chips for America Act”, passed by the Senate in June, calls for $50 billion to be spent to boost the US semiconductor industry, which would make it perhaps the largest act of pure “industrial policy” in US history.
have chips 100 times bigger than the previous such initiative, The 1980s and 1990s of the SEMATECH consortium, which addressed 14 semiconductor companies with government funding, were then seen as a “Japan threat” to American technological leadership. Chips is now fueled by the prospect of an emerging “China threat” – and a sense of potential vulnerability is widespread, including both economic pre-eminence. And Ideas of national security.
There is also a factor of intellectual contagion. The Korean government is talking about a similar subsidy ($65 billion) for its chip industry. Europeans have big plans – Mario Draghi talks about the need this week To overhaul the entire financial structure of the European Union, to partially support “Huge investment in semiconductors.Also this week Japan announced a new round of subsidies. And of course China has put semiconductor technology as the centerpiece of its “Made in China 2025” plans to achieve (at least) technological parity with the West. How can America just stand up? Is it really a “bet we have to make”?
Well, $50 billion is No Enough to “fix” our perceived weaknesses. They may not even be fixable, which is really to say that they may not exist. but $50 billion Is More than enough money, perhaps seriously, to distort the capital allocation process in this industry – especially given that there are indications that this generosity may be aimed at the wrong goals. The only real precedent – SEMATECH – is now seen as being ineffective. Some studies have concluded that government involvement through SEMATECH is in fact a Shortage In the overall level of domestic investment in chip technology. Could the chips – a massively large program – also misfire? Or even make things worse?
incompatibility, and worse
The outlines of the chips are described in previous column. The bill itself is little more than an outline. It proposes some general measures (e.g., traditional non-focused “incentives such as investment tax credits”) and some very specific goals (eg, “support for 3 nanometer transistor processes”). Like many pieces of draft legislation, CHIPS is rather inconsistent, and lacks detail. (The House has yet to act, and it will be interesting to see whether the measure gains or harms in consistency and clarity as the process progresses.)
The inconsistency in the bill reflects a general sense of confusion on the part of policy-makers about the “semiconductor industry”. Many of them are reacting to headline numbers, which can be misleading and sometimes fake in this business. For example, there are frequently quoted statistics On the decline in US market share in chip manufacturing:
- “Some semiconductors are actually made in the US. Only 12 percent of chips sold worldwide in 2019 were made in the US, down from 37 percent in 1990.
worrying? Maybe it’s worse? Commerce Secretary (Gina Raimondo) recently told Congress He –
- “The US once led the world in making leading-edge semiconductor chips. Today we produce 0% of those chips in the US, 0%! That’s a national security risk and an economic security risk. The Defense Department has been warning us for years.” Giving…[We have to] protect yourself We are completely dependent on Taiwan and China for vital supplies.”
The fact is that these numbers are almost meaningless for a number of reasons. Raimondo’s “0%” is a finely constructed distortion (and one wonders if she knows it, or if her staff knows it).
This is not to say that there is no cause for concern. Yet more careful analysis than the legislative process may be required to veil the statistical fog, and to present a more accurate quantitative assessment of our true weaknesses. There is always a tendency to boil down complexity – and the semiconductor world is massively complex – to simple sound bites. And of course for a slice of the $50 billion pie the diversity of interests would generate many conflicting interpretations.
So there are many questions as to which policy makers and legislators should consider before using cash-laden helicopters for errant goals. Start with the two most basic.
- Where exactly is America vulnerable? Where is the threat concentrated in the vast semiconductor ecosystem?
- What is the nature of this vulnerability, and what is the best way to address it?
After clarifying the analytical framework, the answer to the first question is simple, almost indisputable.
The answer to the second question is highly debatable, and very reasoned. This is where the policy-making process goes astray.
1. Where are we weak?
For classical infrastructure spending by types of roads and bridges, see Funding that is easy to target. Look for potholes, traffic jams, crumbling bridges. Success is measured in miles of paved, alleyways, dug tunnels and spans replaced.
But for an almost unimaginably complex global semiconductor ecosystem, the answer is much more complicated. “Semiconductor industry” is an odd concept. There are at least six major segments, which organize about six different business models with different cost structures, different value-creation potential and different technology foundations. and varying degrees of economic power and risk.
in the letter:
- chip design: This segment includes high-value “fableless IC” companies such as Nvidia and Qualcomm, and more recently, “end users” such as Apple and Amazon, that design (but not manufacture) advanced integrated circuits.
- integrated circuit manufacturing: These are foundries that transfer designs to silicon under contract from fabless IC companies.
- Chip Packaging: These companies complete the manufacturing process of converting silicon wafers into chips – a rigid or “produced” form suitable for end users to assemble into digital devices of all kinds.
- Semiconductor Manufacturing Equipment (SME): Manufacturers of capital equipment used by all segments in the manufacturing process, but especially by foundries.
- EDAElectronic design automation companies provide software to designers to automate the process of developing complex chip designs.
- Core IP: These companies sell or license “pre-designed” functional blocks of intellectual property, which imaginable IC companies can incorporate into their chips.
(These sections are described in more detail in the previous column.)
An obvious answer: it’s the foundry section
US vulnerability is concentrated almost entirely in the fabrication/foundry segment.
Overall, the US position is strong. The US dominates (alone or with its partners) four of these six segments, accounting for the lion’s share of enterprise value added (shown here in the price/sales ratio – the amount of enterprise value generated for every dollar of revenue). generates. China’s share in those areas is negligible.
In fact, more than half is built in the end-user value-added design segment – where the US dominates and Chinese firms have a modest presence.
The fifth segment – IC packaging – is a very diverse industry, less technologically sophisticated, more commoditized, with fewer barriers to entry and many alternative sources of supply. It generates the smallest part of product value and enterprise value. The US has a 28% market share; Together with our close allies – Korea, Japan, Europe – we have about 52% stake. The risk is 29% in Taiwan. China has 14%. American risk in this segment seems manageable.
The US vulnerability is concentrated in the manufacturing/foundry segment. Several US semiconductor firms specialize in the design and are “fictitious”. They rely on foundries to make their products. The foundry sub-segment is where the potential vulnerability is most acute, especially if we take into account Taiwan’s geopolitical risks. That’s the simple answer.
2. What is the nature of US vulnerability?
Analysis of the semiconductor fab/foundry landscape typically divides the US vulnerability into two further questions:
- How much manufacturing capacity does the US control?
- Where does America stand in terms of leadership for the next generation of construction technology?
these are wrong questions for asking– but they often outline the discussion of the problem so that it is useful to first answer them on their own terms.
[Caveat: The question of vulnerability combines two kinds of risk: supply chain disruption risk, and geopolitical risks (which is a…