Article • Dean Baker’s Beat the Press
The Sanders-Trump Plan for an AI Sovereign Wealth Fund: Don’t Buy It
Article • Dean Baker’s Beat the Press
Since I keep hearing people singing the praises of the possibility of left-right collaboration over the idea of an AI sovereign wealth fund, I’m going to take another opportunity to trash it. This is not just a bad idea; it is a seriously bad idea.
There are three main reasons:
I’ll take these in turn.
There seems to be an idea among some on the left that if the government runs a company, that we will have honest, competent people in charge who act for the public good. In a best-case scenario, this can be true. But that is very much the exception, and definitely not the case in Donald Trump’s America.
In the normal course of things, we can expect to get career civil servants who are generally honest but typically not big risk-takers. That is not necessarily a bad thing; we don’t want to take big risks with things like air traffic control, but it can be an argument as to why some things may be better left to the private sector.
But we are not dealing with the issue in the abstract; we have Donald Trump in the White House. And despite the intentions of people like Bernie Sanders, the Supreme Court just said that Trump can put anyone he wants in charge of every government agency, except the Federal Reserve Board. (The SCOTUS said the Fed is important.)
When we are talking about an AI sovereign wealth fund, we are talking about giving Donald Trump yet more power to steer the development of AI. This guarantees both corruption; more money will go to Trump donors and friends, and a lack of any real concern with public safety.
These problems were already bad enough with Trump using regulatory authority and the power to award or withhold purchases by the military and elsewhere in government. Handing the keys to Donald Trump through a sovereign wealth fund stake in AI would just make matters worse.
It is common to talk about industrial policy as though it is a decision to favor certain industries, as opposed to leaving things to the free market. Guess what folks, we always favor certain industries; we just tell lies to pundits and young children.
What are government-granted patent and copyright monopolies if not policies to favor certain industries? Seriously, this should be evident to everyone. These monopolies don’t fall from the sky; the government issues them to protect sectors like prescription drugs, computers, software, and entertainment.
This is arguably good policy, but you can’t say it is not policy, and just the free market. These monopolies can be longer or shorter, or stronger or weaker, and they don’t have to apply everywhere. It wasn’t always possible to patent life forms, software, or business methods.
The infrastructure we choose to support also directs the economy. Highways support autos and suburbs. Airports support airlines, airplanes, and tourism.
The argument over industrial policy is not whether to have it or not; it’s what the policy should be. Is it a good idea for the government to promote AI? Perhaps, but let’s hear the argument that resources are better spent on AI than housing, education, or medical care.
The last point is that we have a serious AI bubble, with stocks like Nvidia hitting market capitalizations as high as $6 trillion, which is close to 20 percent of US GDP. (Microsoft peaked at around 6 percent of GDP in the 1990s tech bubble.) SpaceX, which is primarily an AI company according to its registration statement, has a market capitalization of more than $2 trillion, even though it loses billions of dollars annually.
The overall capitalization of the stock market is close to 40 times annual earnings, which is more than twice its long-term average and near its peak in the 1990s tech bubble. This AI bubble has given us the world’s first trillionaire and hugely increased the fortunes of many other billionaires.
It is hard to see why anyone who is remotely progressive would want to see this bubble persist and grow even larger. Pushing the idea of an AI sovereign wealth fund promotes the bubble in two ways. First, it holds out the possibility of tens or even hundreds of billions of dollars of public money flowing into the stocks of the big AI companies. That inflow could give stock prices a substantial lift, and also give our AI billionaires a great opportunity to cash out.
The other way the talk of AI sovereign wealth fund could boost the bubble is by helping the industry’s sales effort. The big AI proponents have pushed the idea that AI will utterly transform the economy and displace a huge share of the labor force. There is zero evidence of this in the data to date, but the belief that this will happen is what fuels stock prices.
We should want policy that is based in reality, not sales pitches by AI CEOs. And we certainly should not want to help them with their sales pitch, as this AI sovereign wealth fund proposal does.
It is widely recognized that corporate tax avoidance/evasion radically reduces the effective tax rate below the 21 percent statutory rate. Many tax lawyers and accountants get rich by designing complex schemes to minimize corporate tax burdens. This is an especially big problem when we have an I.R.S. that is not very interested in collecting the corporate income tax.
A simple way around this problem is to require corporations to turn over a percentage of their stock, in the form of non-voting shares, that is equal to the targeted tax rate. That would currently be 21 percent, but if we targeted a higher tax rate, it could be 25 percent, 30 percent, or whatever was considered the appropriate level.
These shares would get the same dividends or share buyback terms as other shares, but they would have no voting power. This would mean that we don’t have to worry about Donald Trump running Anthropic or OpenAI, but we would get the intended tax revenue. As a sidebar, we would also put the tax shelter industry out of business, eliminating a considerable amount of economic waste.
This would be a great policy for progressives to push. There is no obvious downside to structuring the corporate income tax in a way that can actually be collected. That is a far better aspiration than buying into an AI bubble.