In my 15 years in Silicon Valley, I found venture capitalists to be the shrewdest and smartest individuals there. The best of them have had previous experience as engineers or managers, sometimes both. They had experience at some of the best, most successful companies but also experience at some of the worst-managed companies (which far outnumber the well-managed) and from where you sometimes learn better lessons than you can at the best. Being a successful venture capitalist is a bit like playing three-dimensional chess. You invest today, but you have to envision how many different processes will play out, including not just your own startup’s engineering activity, but also what its competitors will do, what the customers will want years later when your product is ready, and what the broader trends in the Internet might be, all in order to end up selling the company you are building at a healthy profit at some point in the future.
So if I were going to invest $52 billion in the semiconductor industry, I would choose a few venture capitalists (VCs) to manage the investment. I would not choose working VCs—they have actually done a pretty poor job funding semiconductor startups in recent years. I would choose retired VCs with no financial interest in any of the companies they are investing government funds in. I am thinking of people like Vinod Khosla, John Doerr, Eric Schmidt, and John Chambers. The first two are experienced VCs with outstanding track records and the second two are retired executives of successful tech companies who have done venture investing. They have all demonstrated an ability to understand the global tech industry and take the long view. Their only reward for this work would be national recognition. They are all already wealthy so money would not be an issue.
Readers will know that Congress has just enacted the Chips and Science Act, which provides for $52 billion of federal financial support to semiconductor companies that build fabrication facilities (fabs) in the U.S. So if you are a taxpayer, you are investing $52 billion in the semi industry. The details of how this money will be doled out have yet to be fully worked out, and will likely include loans, grants, and tax credits, but financial analysts are suggesting that the support will be worth several billion dollars each to the bottom lines of Intel, Micron, and others over the years 2023 to 2030. This is the right policy, because the U.S. is too dependent on chips fabricated in Asia and the Chips Act will lead to more fabs being built here. I support the Chips Act as does the Coalition for a Prosperous America. I also have friends who work at some of the companies in question and I am glad to see them getting federal support.
But there is a better way to do this. And it comes in three steps: problem, solution, and implementation.
The VCs would be managing these strategic investments for the U.S.A. and the American people. The U.S. needs to be less dependent on chips manufactured in Asia, for several reasons. One objective is to strengthen our national security, so if the flow of semiconductors from Asia was disrupted for any reason, our production of civilian and military gear could continue. Another key objective is to ensure that the U.S. continues to be at the cutting edge of technology, since semiconductor manufacturing processes are changing and progressing all the time. The first thing the VCs would ask Congress would be: what is the specific problem we are trying to solve.
The U.S. problem is that we consume billions of dollars of electronic products containing semiconductors but we produce only a small (and shrinking) share of the global output of semiconductors. The global output of semiconductors last year was $556 billion, and many industry experts expect this to double to $1 trillion or more within ten years. Semiconductor market growth will follow the growth in the number of devices, from kitchen appliances to EVs to fighter jets, that use a growing number of semiconductors.
I will quickly summarize the problem, which I see as coming in three buckets: industrial, national security, and knowledge. The industrial problem is that many other industries are dependent on semiconductors and subject to interruption if shipments from Asia are disrupted, as they have been for over a year now due to the COVID-related supply chain snafus. This article from Car and Driver gives some insight into how many auto functions are dependent on semis these days. The functions that some auto companies have been forced to omit from recent production include: touchscreen, cruise control, HD radio, heated seats, LED headlights, GPS navigation systems, and adjustable seats. Aside from autos, many other products, from laptop computers to home toasters to military equipment, are all dependent on semiconductors and could be at risk of losing production if semiconductor supplies are in doubt. Moreover, with the growing trend in the semi industry of concentrating production in fewer, larger fabs, there is a real risk that the current supply chain snafu will not be the last.
The national security problem stems from the China challenge. China’s Made in China 2025 plan calls for China to produce at least 70% of the semiconductors it consumes. This goal lies behind huge investment in semiconductor design and manufacturing. According to BCG, the China share in semi manufacturing is today around 15% (compared to the U.S. share of 12%) and by 2030, China’s share will be 24%, largest in the world, and ahead of both Taiwan and South Korea. China already has two successful semiconductor companies, SMIC and YMTC, and more are likely to follow. Tech analyst Dylan Patel of SemiAnalysis has pointed out that YMTC’s memory chips are now technically equal, and in some cases better, than those from western producers while SMIC is now shipping specialized processors based on 7nm technology, something Intel has yet to achieve. If China achieves the sort of dominance in semiconductors that it has in other core industries like steel, the entire world risks being dependent on China for another crucial technology and product sector.
Finally, the knowledge problem is that if China achieves leadership and domination in semiconductors, the knowledge sectors where America leads today could very well also migrate to China. Today the U.S. is by far the dominant nation in chip design. But manufacturing and design work closely together and China could achieve leadership in design too. The rise of China-based software companies like TikTok and Zoom shows that China has the capability to design cutting-edge software.
In a financial analysis I published last year, I argued the U.S. chip industry is losing out because it invests less in capital spending than foreign chipmakers. In June this year, analyst Patel looked at the investment patterns in both semiconductor startups and large companies and his conclusion was headlined: Why America Will Lose Semiconductors. Patel argued that the investment trend in the U.S. semi industry “is terrifying for prospects of American hardware dominance. Not only is most of the assembly done in China, but the most well-funded startups in the semiconductor field and the most IPOs in the semiconductor field are occurring there well. China by far is building the most fabs, which is driven by their favorable tax and regulatory policies as well as massive subsidies. The US is a tiny share of worldwide spending.”
The solution to the problem is for the U.S. to have a critical mass of manufacturing in each critical sector of semiconductor production. This is realistic and achievable, as the U.S. still has many competitive advantages in the semiconductor industry. Unfortunately, the figure of 12% of global semi manufacturing does not help us solve the problem because it is too general. The U.S. needs critical mass in each of the key sectors. Below is a table I have assembled showing some of the most critical semiconductor sectors. This is by no means a comprehensive list but illustrates a path towards a true solution.
Our biggest dependences are in memory chips, where our production is very small and LEDs, where our production is nonexistent. These two kinds of chips are found in automobiles and many products used in business, in the home, and by the military. We have listed microcontrollers or legacy logic as a separate category because of their importance to so many of the small systems found in automobiles.
One approach that could address this problem would be to set out targets for the U.S. share of global production in each of the most critical sectors of semiconductors. The U.S. generally accounts for about 30% of global consumption of technology products, which are the largest users of semiconductors. So if the U.S. produced 20% of global output, it would be able to fulfill the majority of domestic needs.
The 20% figure is chosen as an illustration. In reality, an in-depth investigation should be done of each sector and a realistic target chosen. The target should include not just chip fabrication but the upstream and downstream support industries such as production of silicon wafers and the packaging of chips. Packaging in particular is a complex, high-tech process and without a sophisticated packaging industry, chip fabrication is unlikely to succeed. One of the secrets of Taiwan’s success is a focus on support industries like packaging.
|Sector||Global Market Size, 2021 ($B)||U.S. Mfg Share (estimate)||Leading U.S. companies||Target|
|Microcontrollers (Legacy logic, MCU)||$20||15%||TI, NXP, Global-Foundries||20%|
|Other||$207||5%||Renesas, Global- Foundries||20%|
Source: SIA, Company statements, BCG, Imarc Group, CPA estimates
Equipped with these targets, the VCs managing the government’s semiconductor manufacturing plan would then find chip companies willing to sign up to reach the target levels of production. While the semiconductor industry is dominated by a few giant producers in each sector, with a pot of $52 billion of cash, the VCs could find some smaller companies to get into segments of the industry, to help the government avoid excessive dependence on industry giants.
Not a Subsidy. A Commercial Transaction
When VCs invest in a company, they are always careful to structure the transaction to ensure that incentives are aligned. The investment of the $52 billion would be structured so that each company receiving funds commits to increasing U.S. production to agreed targets within a fixed time period, and maintaining that level of capacity for a period of ten years. If the company doesn’t meet the target, it pays back the investment.
With the current Chips Act, incentives are not aligned. U.S. semiconductor companies are still fundamentally incentivized to move production to China, for three reasons. First, costs are lower in China (mostly because of subsidies, labor costs are insignificant in a fab). Second, the Chinese market is larger than the U.S. market, accounting for 35% of sales last year as opposed to just 22% for all of the Americas, according to Semiconductor Industry Association data. Suppliers want to be close to their customers. Finally, the Chinese government is actively putting pressure on semiconductor companies to shift operations to China.
The Chips Act passed last month does not change the fundamental aims of U.S. semi companies. It just throws a nice big tax credit into the mix, incentivizing them to produce in the U.S. But it will not stop their gradual, steady drift towards China.
By requiring funding recipients to give back the funds if they don’t hit targets and hold them for a period of years, this proposal would drive a fundamental change in the objectives of the recipients. Before taking any funding, they would be forced to consider how they would make a profit on chips manufactured in the U.S. They would likely contribute to improving the plan. For example, they might suggest that a credit be provided to buyers of devices like laptops that include chips made in the U.S.A. Or they might even propose an effort to reshore laptop computer production. Most of the smart people I know in the semi industry say that the U.S. must train more graduate students in material sciences and other STEM fields for the U.S. to regain leadership in semi manufacturing. In the VC-led scenario, chip companies would likely insist on this, and the federal government could find a way to incentivize our universities to incentivize our youth to take an interest in these vital fields.
The people who understand the chip industry best are first, those who work in it, and second, its customers. By making the government funding a commercial transaction, the VCs managing this plan would ensure that everybody’s incentives were aligned, and everybody involved had an interest in finding a way to make it profitable for chip companies to manufacture a certain portion of their output in the U.S.
Note that this also addresses the congressional critics. Sen. Rick Scott called the Chips Act a “reckless spending spree,” while Sen. Bernie Sanders asked why the federal government is giving away $52 billion to companies where employees are extremely well-paid, CEOs like Pat Gelsinger earn over $100 million a year, and stock buybacks are rampant.
In the VC scenario, this is not a spending spree or a giveaway. It is a commercial transaction. We, the American people, need the security of domestic production for a certain portion of our output, and we are prepared to invest $52 billion to achieve that. If the recipients of the funds do not meet their obligations, they would pay back the money, just as a semiconductor company does now with its customers when it fails to deliver product.
Industrial Policy for Other Industries
As the Biden administration seeks to address domestic shortages in other industries, this proposal is highly relevant. Due to a timely combination of factors, including the auto production snafus and growing concern over the China threat, the Chips Act won bipartisan support. But it will be harder with other industries. A more commercial approach, where the government invests in return for companies’ agreeing to meet certain targets would be more politically acceptable. It would also be more likely to succeed.
Such an approach could be applied in important sectors like pharmaceuticals or rare earth mining and refining. There are many sectors where an intelligent approach is needed to counter China’s efforts to dominate global production. But the most important factor in such an approach is that it would align the incentives of the government and the private companies that are most knowledgeable in each area, and whose support is critical.
Industrial policy needs to move on from subsidies to a focus on agreed goals and objectives.