Daily-Dose

Contents

From New Yorker

From Vox

Wolfspeed’s silicon carbide 200mm semiconductor wafer.

All of this means the cost of building a fab can range from $1 billion to $20 billion, depending on the complexity of the chips that are being manufactured. This is the primary reason that the recent surge in demand for chips — fueled in part by the demand for more laptops and more cars — did not immediately result in more chip fabs. Because these plants take years to greenlight and construct, chip companies aren’t eager to spend billions on building more factories, since demand could always subside. This is partially why governments often intervene and provide incentives to build more chip factories.

Case in point: New York officials spent decades trying to attract a semiconductor company to Marcy, where New York state has funded a nanocenter associated with the SUNY Polytechnic Institute. Wolfspeed only agreed to take over the site after another company backed out and New York offered to subsidize the fab with a $500 million grant — about half of its total construction costs. Now, even more money is on the horizon, not just for another Wolfspeed factory, but for possibly even bigger fabs, including a new $100 billion megafab in Ohio built by Intel, which the administration is hoping will regain “the leading edge” and start building the same kind of advanced chips that TSMC makes. President Joe Biden, in his most recent State of the Union address, said that this facility, once it’s built, could provide as many as “10,000 new good-paying jobs.”

The big chip bet

Before any of that can happen, officials say the US needs to pass a $52 billion package called the Chips Act, which would subsidize the construction of several new fabs. Currently, the bill is packaged within a broader proposal called the United States Innovation and Competition Act, legislation focused on competitiveness with China. While the House and the Senate version of this plan aren’t exactly the same, the initiative has the support of Republicans, Democrats, the White House, and the major chip companies. The support from the industry isn’t surprising; each of these companies could theoretically receive up to $3 billion to build a new factory, and another $2 billion may be earmarked specifically to build a fab that would exclusively focus on more basic chips used in cars.

Proponents of the massive bill argue that it’s the bare minimum because other countries are still subsidizing chip manufacturing, too. Back in 2014, China launched a $150 billion effort to boost its own semiconductor industry over the next decade, and the country has imported fewer and fewer chips in recent years. South Korea plans to spend as much as $65 billion on its own national chip initiative. The European Union also has its own $49 billion Chips Act, and its member countries, including Spain and Germany, will soon launch their own incentive programs.

“The clock is ticking,” John Neuffer, the CEO of the Semiconductor Industry Association, a trade organization that represents American chip companies, told Recode. “Decisions are being made today as to where to site those fabs.”

Courtesy of Wolfspeed

A Lucid electric vehicle cuts the ribbon at the new Wolfspeed chip fab in Marcy, New York, on April 25.

Not everyone loves this approach; it’s effectively a corporate subsidy for companies that are already extremely profitable. Sen. Bernie Sanders has been highly critical of the Chips Act, and has said that chip companies should have to give up equity in exchange for massive grants. Others have argued that these companies would build new factories in the US regardless of federal incentives, since they also have reasons to steer clear of potential geopolitical conflict. And critics point out that chipmaking is not quite the jobs-creator that it’s sometimes advertised as, with most actual chip manufacturing being done via automation.

There’s no guarantee the funding will work. The US may not have enough of the specialized workers that chip manufacturing requires to support the number of fabs that officials want. Will Hunt, an analyst at Georgetown’s Center for Security and Emerging Technology (CSET), estimates that eight new fabs may require at least a few thousand foreign workers, since many of these facilities need to hire people with previous experience working in semiconductor manufacturing. Another concern is that the US’s lengthy regulatory and permitting process could slow down the construction of new factories, and the US already builds new fabs at a slower rate than countries in East Asia. Even after these facilities are constructed, they may not produce the number of chips or jobs that companies promise.

A senior economic official at the White House told Recode that while the $52 billion will boost American chip manufacturing, it won’t be enough to produce the number of chips the US consumes. Still, the government thinks that gaining this manufacturing expertise could be critical during a future emergency. After all, the pandemic has illustrated time and time again that when supplies are short, countries will try to secure the world’s most sought-after products — whether it’s chips, masks, or vaccines — and can even use them as a way to influence international relations. Governments would rather other governments be dependent on them than the other way around. In other words, they want bargaining chips.

So it’s not surprising that semiconductors have become that leverage. These tiny little chips are ubiquitous and have become a necessity in most people’s everyday lives. There’s no indication that’s changing anytime soon, especially since more powerful devices — which use even more powerful semiconductors — are always being rolled out. As long as the world depends on this technology, countries will want as much control over chips as they can get. That means that even with Wolfspeed’s factory now open for business, the US still has a long road ahead.

That’s an alarming-seeming but ultimately misleading number; most of it is accounted for by changes in international trade (imports, which reduce GDP numbers, rose as businesses knew war in Ukraine was coming) and the shrinking of inventories held by businesses as consumers bought them up. “Final sales to domestic purchasers,” which strips out those factors, was up 0.6 percent for the quarter, or 2.6 percent annually.

However you parse the release, though, this seemed as good a time as any to discuss the long-run future of growth in the US and the rest of the world. I’m getting pretty worried about it, to be honest! And I’m worried because of an economics paper by NYU’s Thomas Philippon with the unassuming name “Additive Growth.”

The implication of Philippon’s paper is as simple as it is disturbing: We should expect economic growth to slow down in the long run, and the big leaps forward of the last couple centuries may be an aberration.

This conclusion is far from certain, and it goes against decades of assumptions on how to model economic growth. But Philippon brings a lot of data to bear on his thesis, which makes some intuitive sense, and even the possibility of it being true should alarm us.

The big pile of stuff we know how to do

Philippon’s paper is not concerned with economic growth per se, but with a variable that is central to explaining long-run growth: total factor productivity, or TFP.

Defining or getting a handle on the idea of TFP can be tricky. Technically, it’s just a residual: the annual growth rate of TFP is what you get when you look at annual economic growth, and remove the growth attributable to an increase in labor (more hours worked) or capital accumulation (more factories built, labor-saving machines purchased, etc).

For now, think of TFP as measuring something like “how well humans are able to use labor and tools to do stuff.” If TFP grows, that means we can get more economic output out of the same people and stuff we already have.

That makes TFP the “secret sauce” behind economic growth more generally. Labor inputs can increase, of course — but people can only work so many hours, and don’t really want to work too many hours. Population growth helps there, but less so when it comes to per capita economic growth, which arguably is what matters most.

Capital accumulation — using more labor-saving equipment and tools — helps too, but there’s only so much money to invest. The key is using the resources you have more effectively — and TFP measures, roughly, how effectively we’re using our resources. We can get more effective at using our resources through advancements in science, business management, and other changes.

Economic growth is generally modeled exponentially: our economic output grows by a set percentage every year, and while that percentage varies, it also compounds on itself. TFP is usually modeled the same way. If you have $100 growing 2 percent each year, that’s exponential. If it instead just gains $2 every year, that’s linear.

What Philippon does is attempt to assess whether TFP actually does, in practice, grow exponentially. He first looks at two datasets covering TFP in the US and finds, instead, linear growth since World War II: TFP does not increase by a set percentage each year, but a set amount (0.0245 points, if you’re curious) each year. It doesn’t compound; it just gradually, steadily grows. You’re getting $2 a year, not 2 percent of an ever-increasing pile.

Extending the data back to 1890, he finds linear growth, but with a break: slower growth from 1890 to 1933, and faster after 1933, but steady and non-exponential in each period. He then extends the analysis to 23 relatively wealthy countries, from Japan to Germany to Spain. A linear model fits better here, too.

Linear growth implies, as Philippon writes, that “new ideas add to our stock of knowledge; they do not multiply it.” It also implies a slowing of economic growth over the long run. Or, as Jason Crawford, who writes about the history of science and technology at the Roots of Progress, put it, “GDP per capita can continue to grow without bound, but that growth will slow over time.”

How doomed are we, exactly?

The US and other rich countries have experienced a well- documented decline in productivity growth, especially TFP growth, since 2004 or so. Philippon’s findings could help explain why that is. The slowdown is only there if you assume TFP should be growing exponentially. If you assume mere linear growth, it’s not that things have gotten worse in recent decades. It’s just that they were never that good.

That’s an alarming conclusion, mostly because from the standpoint of human history, the past few centuries have been very good. Before the 17th to 18th century or so, human economies grew extremely slowly. Agriculture showed little productivity growth, meaning there was a fixed population that farming societies could support. Living standards varied mostly based on how many people were around; when the population suddenly shrank (as in the Black Death in Europe) people grew richer on a per capita basis, and when the population swelled the opposite occurred. This is known as the “Malthusian trap.”

“Until about 1800, the vast bulk of people on this planet were poor,” Joel Mokyr, an economic historian at Northwestern, once noted. “And when I say poor, I mean they were on the brink of physical starvation for most of their lives.”

That pattern started to break down in the 17th through 19th centuries, a process sometimes shorthanded as the “Industrial Revolution,” but including a wide variety of cultural, scientific, technological, and economic changes. Long story short: productivity sustainably grew for the first time in human history. And it grew, by historical standards, quite rapidly, such that a far lower share of people alive in 2022 are on the brink of starvation than were in 1800, even though the population needing food has never been greater.

Philippon identifies this break in human history in his paper, looking at TFP data about England. TFP growth is linear throughout English history, he emphasizes, but the rate of linear growth became suddenly higher around 1650 and 1830. These “break points,” he argues, correspond to the so-called first and second industrial revolutions (the first characterized by textile work and the discovery of steam power, the latter by mass production, industrial steel/plastic, electricity, etc.) beginning in England. I’m less persuaded of this than Philippon — the precise timing of the industrial revolution(s) is among the most hotly debated topics in all of history, and 1650 in particular feels a bit early to date the start of it.

But if he’s right, that means that humanity would need another, similarly important break to prevent economic growth from slowing long-term.

Philippon’s paper is bracing, but I’m not 100 percent sold. I share Tyler Cowen’s question about the usefulness of treating TFP as a real, observable attribute of the economy.

What is TFP, anyway? It’s sometimes shorthanded as a measure of “technological progress,” but it doesn’t really measure technological change; TFP can grow or shrink without changes in technology, and technological changes can occur without affecting TFP. The late economist Moses Abramovitz famously dubbed it a ”measure of our ignorance about the causes of economic growth” in a 1956 paper.

We should try to shrink that ignorance — but how meaningful are changes in the unexplainable share of economic growth? They probably mean something, but it’s hard to say what. TFP has been incredibly useful for comparing productivity between, for instance, firms, to identify which are more effective and efficient; here, for example, is a great paper on productivity of smallholder African farms.

But I’m not sure TFP holds the same explanatory weight in explaining the growth of countries over a long period of time.

All the same, Philippon’s paper should at the very least open up an important new direction for research. One observer I respect concluded, after reading the paper, that if true, it means “Human extinction looks much likelier.” I wouldn’t go that far. But it asks an important question, and answering that question correctly matters a lot.

From The Hindu: Sports

From The Hindu: National News

From BBC: Europe

From Ars Technica

From Jokes Subreddit