Last week the Senate voted overwhelmingly (96-4) to include an amendment to the National Defense Authorization Act (NDAA) that provided support to the American semiconductor industry. The amendment combined the earlier American Foundries Act and the Creating Helpful Incentives to Produce Semiconductors (CHIPS) for America Act. A nearly identical proposal was included in the House NDAA, so that bodes well as the bill goes into the Conference Committee. But the vote does not authorize or appropriate funding for these policies, it only creates the policy framework.
The pandemic and worsening trade tensions have shined a light on worrisome issues with the supply chains that American industry and consumers depend on. Semiconductors are an area where the U.S. is heavily dependent on Asia. While the U.S. still has most of the leading design companies, we have comparatively few of the leading manufacturing facilities. In that critical sector we trail Taiwan, an island claimed by China that is one quarter the size of the state of Florida, and then South Korea, Japan, and China. The U.S. has gone from being first in manufacturing at the dawn of the industry to fifth, and companies in those countries who are ahead of us are investing aggressively. American makers of the production tools used in semiconductor factories have been moving their manufacturing to Asia as well, because that’s where their customers are.
The recent Intel
INTC
announcement adds great urgency
If Congress needs any additional impetus, it only needs to look to Intel’s recent announcement that it will delay the launch of its next generation chips, and that it was considering contracting out more of its manufacturing to make up for the delays. This should ring alarm bells for a number of reasons:
- Intel has lost ground to AMD in the x86 microprocessor chip market because AMD has well designed products that are manufactured by Taiwan Semiconductor Manufacturing Company (TSMC). Don’t forget – a decade ago when AMD fell behind in manufacturing, it spun off its fabs (which became the foundation for GlobalFoundries) and turned to TSMC to produce its chips. TSMC is able to make AMD’s chips with its most advanced processes which are now even further ahead of Intel. The only place Intel can go to catch up with AMD is TSMC, so the pressure on them to outsource to there (Intel already is an important customer) will be intense.
- Apple
AAPL recently announced that it will shift away from using Intel microprocessors in its Mac computers to chips of its own design that will be fabricated at TSMC. One could argue that Apple was only a small share of Intel’s volume, but don’t forget – the last time Apple shifted its Mac chip supplier – from PowerPC chips made by IBMIBM and Motorola/Freescale to Intel – that marked a significant turning point in IBM’s decline and eventual exit from semiconductor manufacturing. - TSMC’s proposed fab in Arizona is only a token investment with insufficient capacity to support even a small fraction of the needs of Apple, Qualcomm
QCOM , NvidiaNVDA , Intel, AmazonAMZN , GoogleGOOGL , and other American design shops. The proposed fab has a capacity of 20,000 wafer starts a month, a drop in the bucket compared to ~250,000 wafer starts per month at a single one of TSMC’s Gigafabs in Taiwan. And TSMC has three Gigafabs going on four spread over three science parks on the island.
It’s going to take more than money
The NDAA amendments reflect a bipartisan willingness to devote substantial funding to this area. In addition to a final authorization and providing sufficient long-term funding, Congress should set out clear goals for technology development and economic competitiveness. Clear goals enable a market-driven race by companies to develop their own roadmaps to get there. For example, the NASA Commercial Orbital Transport Services program had the clear goal that the U.S. must have a domestic means of human space flight launch (i.e., stop relying on Russia) with all necessary technology to support it. The recent success of SpaceX with its Crew Dragon is a fulfillment of that goal. For semiconductors, the goals might be that we develop both the capacity and the economic justification to manufacture half of all next generation 5G/6G and AI chips domestically. Giving a single agency a mission focus around such clear goals will also avoid the inevitable dilution that comes with spreading the effort too broadly. NASA has demonstrated the success of such a focused approach in many other programs, as has NIH in the Human Genome Program and others.
Focus on the demand side
The bill ammendments do a good job of providing supply side incentives – incentives to build more foundries (the factories that manufacture chips), to invest more in R&D, to build up our capabilities in neglected areas like chip packaging, and to support the education and training of workers. A more sustainable approach would be to focus on the demand side, growing domestic demand in early markets for new technologies as a way of incenting the growth of local supply.
If we look historically at industries in which the U.S. has led—automobiles in the 1920s, computers, telecommunications, integrated circuits (ICs), the Internet, products using the global positioning system (GPS)—large early markets drove consumption and gave American firms incentives to innovate. NASA and the Air Force gave long-term contracts with milestone based funding to SpaceX (and to its competitors) to deliver payloads to orbit. Those contracts provided the company with cash flow to stay alive and develop innovations like reusable vehicles that have changed the game in space launch.
Having a large home market in which to “practice” is a significant advantage. As long as consumers will buy interim products as the manufacturer improves its production processes, demand can generate the cash a firm needs to grow, learn, and improve. As we look to restart our economy after the pandemic, using stimulus spending to drive demand for specific technology investments would be a double win. An initiative around electrical grid modernization would stimulate demand for power semiconductors, Group III-V and silicon carbide semiconductors, IoT chips, and energy storage devices, as well as create many jobs and contribute to reducing our carbon footprint. Demand is important as well for people – when there is demand, students will go there for careers.
Export markets are important
We should tone down any commercial cold war with China or other parts of the world. Chip makers who still manufacture in the U.S., companies like Texas Instruments
Opportunities to restore American manufacturing competitiveness in a sector don’t come around very often. But the bipartisan willingness to support and fund such an initiative, coupled with directing stimulus funding to create demand pull might be a once in a generation opportunity.