Opinion: CHIPS for America is not enough

Opinion: CHIPS for America is not enough
Feature articles |
It takes more than $22.8 billion to change the way global semiconductor players behave in a global economy. The US runs the risk of spending both too much and not enough.
By Peter Clarke

Share:

For people outside the semiconductor industry $22.8 billion seems like a lot of money.

That is the amount of money the CHIPS for America bill proposes is provided over the next five years to stimulate the US semiconductor industry (see CHIPS for America Act promises $22.8 billion in aid). In particular, the bill wants to encourage the building of wafer fabs in the United States and significantly reduce dependence on off-shore production. But it is not enough to ensure that outcome.

In essence it is part of an attempt to fulfill President Trump’s promise to “bring manufacturing back” to the United States. But it is not even clear that the $37 billion that the SIA originally lobbied for is enough (see US semiconductor industry seeks $37 billion government boost).

Even for politicians who deal with trillion-dollar annual budgets, the $22.8 billion to be given to the semiconductor sector seems like a lot of money. But it is not enough. Indeed, many politicians simply do not understand the semiconductor industry. Remember the European Commission’s forlorn pleas for the European semiconductor industry to unite almost a decade ago (see European Commission repeats call for “Airbus of chips”).

It is true that $22.8 billion will have some impact. Much of the money will end up getting wasted as government support money usually does, but some will make a difference. About half of the money is being targeted at R&D initiatives. It would be needed to develop the technical capabilities and ecosystem in the US to support leading-edge manufacturing. But again remember all that support that was spent on researching 450mm-diameter wafers.

Next: Everyone’s doing it!


The old argument that gets rolled out when support mechanisms are proposed is that everyone else is doing it and so if the home nation or region doesn’t do it then it is suffering under a disadvantage. What this argument usually fails to recognize is that the home nation is usually already providing support in various forms. It also fails to recognize is that subsidy is not usually what makes the difference. What makes the difference is vision, entrepreneurship and good execution.

One exemplar of that is leading foundry Taiwan Semiconductor Manufacturing Co. Ltd. The company was founded in 1987 by Morris Chang with the help of the Taiwanese government and Philips Semiconductor. While it can be argued that the company had access to low-cost capital in the early days, under the leadership of Chang the company quickly moved into a rhythm of charging enough for the wafers it produced to fund process development and intensive capital investment for new processes and wafer fabs to run them in. The 30-year rise of TSMC, based on Chang’s vision and near faultless execution, has resulted in a 2020 capital expenditure budget for that one compnay alone of between $15 billion and $16 billion.

The exemplar that $22.8 billion is not enough to propel a company or a country to the leading-edge comes in the form of Globalfoundries Inc. That foundry was created by the buy-out of manufacturing belonging to Advanced Micro Devices and Singapore’s Chartered Semiconductor in 2009. Both were pretty close to the leading edge at the time but struggling to justify the capital cost required to stay in the race. The buyer was Abu Dhabi’s Advanced Technology Investment Co. (ATIC), a subsidiary of the Mubadala sovereign wealth fund.

If you tote up how much Mubadala/ATIC has spent on assembling the pieces of Globalfoundries, supporting the company through years of loss-making, and on building the flagship wafer fab 8 in Malta, New York, it can be seen how expensive trying to lead in making chips can be.

Next: Billions of dollars


ATIC spent several billion dollars over several years acquiring AMD, Chartered and IBM Microelectronics. And in Malta, New York it is estimated ATIC has made a $12 billion investment in its leading-edge Fab 8. Back in 2018, Patrick Moorhead of Moor Insights and Strategy, said that the state of New York had offered GlobalFoundries $1.2 billion and in return for a $3.2 billion investment and a direct headcount of 1,200 people at Fab 8. What New York got, Moorhead said, was a $12 billion investment with 3,300 people employed.

Mubadala’s big bet on semiconductor production has therefore probably cost oil-rich Abu Dhabi something around $20 billion. And yet that was not enough to get Globalfoundries to the position where it could compete head-to-head with market leader TSMC.

In recognition of this GlobalFoundries made the strategic decision in 2018 to drop its 7nm FinFET manufacturing process and pivot away from the leading-edge and towards More-than-Moore manufacturing. That does not mean that GlobalFoundries cannot have a successful and profitable future as maker of a broad range of semiconductor technologies behind the leading-edge in miniaturisation. But it is not the future Abu Dhabi had mapped out when it first started taking a stake in AMD and even talked about bringing chip manufacturing to the Gulf region.

Beyond that there is talk of 2022 being the target for GlobalFoundries to stage an initial public offering of shares. One interpretation is that Abu Dhabi and Mubadala want out and are giving the company time to make some profit, create a reasonable IPO prospectus and thus allow Mubadala to get something back on its investment. The problem is we’ve been here before. Back in November 2012 Mike Noonen, then GlobalFoundries’ executive vice president of worldwide marketing and sales, said that the foundry was looking to stage an IPO once the foundry achieved profitablity.

Meanwhile the cost of setting down leading-edge wafer fabs continues to increase exponentially. For TSMC the cost of a fab at the leading edge is now about $15 billion.

Next: Reversing a trend


So when the CHIPS for America bill talks about $10 billion to incentivise domestic semiconductor manufacturing, that is enough money to support about two or three leading-edge investments. The US is already talking to two leading players (see US talks to Intel, TSMC about building local foundry fabs).

The way the CHIPS for America Act may work, is not in terms of the money its lays out, but instead in terms of the creation of a different financial climate on a long-term basis. It has to reverse a 30-year trend that was also marked by the rise of the fabless semiconductor sector.

The fabless chip sector flourished for two reasons. One: venture capital saw the advantage of avoiding capital-intensive investment. Venture capital investment later came to prefer software and internet service plays for the same reason, but that is a different story. Two: the existence of TSMC and United Microelectronics Corp. in Taiwan, and a few other foundries, who were prepared to take on the capital-intensive part of the chipmaking business if they could get enough scale.

The division between fabless design and foundry manufacture was itself entirely predicated on the antiseptically clean nature of the GDSII database format, which became the de facto industry standard for data exchange of integrated circuit or IC layout artwork, and the ability to transport such designs around the world on magnetic tapes or disks. Matters are somewhat more complex today but the principle remains.

If the US financial sector comes to see an economic advantage in domestic, capital-intensive investment then there is the possibility of more than two wafer fabs getting built. If we enter an era of high tariffs on electronic products and quotas for local content that could yet come to pass, but the success of CHIPS for America won’t be based on whether the US tax payer spends $20 billion or $30 billion in domestic manufacturing subsidies, which on its own is not enough.

Related links and articles:

CHIPS for America Act promises $22.8 billion in aid

GloFo rethinks its future, drops 7nm FinFET

Analysis: TSMC ponders US wafer fab while awaiting trade rule changes

US talks to Intel, TSMC about building local foundry fabs

US semiconductor industry seeks $37 billion government boost

Linked Articles
eeNews Analog
10s