AI and Employment: So Far, So Good

By Alex Tabarrok.

"In September 2023, the Census Bureau added questions about AI to its Business Trends and Outlook Survey. Census asked hundreds of thousands of businesses whether they had used AI in the previous two weeks to produce goods and services. At that time, 3.7% said yes; by late 2025 the figure had reached about 10%. (In November 2025 Census broadened the question to ask about AI use in any business function, producing a jump in measured adoption to about 18%.)

Twice the Bureau has asked a key question:

In the last six months, how did the use of Artificial Intelligence affect this business’s total employment?

In Dec. 2023 to Feb 24, when ~5% of firms were using AI the answers were 2.8% increased, 2.6% decreased and 94.6% reported no change. Two years later, in the Nov 2025–Feb 2026 supplement, the answers were: 2.3% increased, 2.0% decreased, and 95.7% reported no change. The answers were similar by firm size.

Some sectors reported more action. Information is the one sector where fewer than 92% report no change. But overall, almost all firms report no change and of those reporting change it’s about evenly divided between increasing and decreasing employment.

 

The supplement also asked about tasks. Among firms using AI, 44% say it supplemented or enhanced work an employee already does. Ten percent say it performed a task an employee used to do. Eleven percent say it introduced a task no one had been doing.

Among those using generative AI, 85% of firms cited writing or editing documents and email as the biggest uses, half cite searching for information, 45% summarizing documents, and 13% coding. Sixty-four percent of adopters say they changed nothing about the business in order to use AI, 15% trained existing staff, another 15% built new workflows, and just over one percent hired anyone with AI skills.

Among firms where AI has taken over some employee tasks, the degree of substitution is growing. The share reporting that AI took over “a large number” of tasks rose from 2.4% to 7.1%, while the share reporting “a moderate number” rose from 13% to 22%. But this group is still small: only about a tenth of AI adopters, who themselves make up about a fifth of firms.

I have reported firm-weighted estimates but employment-weighting gives essentially the same result. Thus, we have unusually direct evidence from a very large sample, and it says that the overwhelming majority of firms using AI do not yet report any effect on total employment. Very consistent with what Tyler and I said in our talk to OpenAI."

 

Counties that build data centers are seeing new housing, higher home values, lower unemployment and more job growth

Tweet from Sara Eisen. 

 

 

Data Centers Are Not the Problem. Bad Policy Is.

By Scott Lincicome of Cato. 

"there are actually many different types"

"data centers have been around for decades, and there are thousands already operating in the United States. They’re basically any “purpose-built building that runs large numbers of computers reliably, securely and around the clock,”"

"Several regional and national studies have found little merit to the common claim that data centers have already caused large, localized increases in Americans’ electricity bills. In fact, a brand new, peer-reviewed study found that data centers actually caused average U.S. retail electricity rates to “fall modestly” between 2015 and 2024, because the centers’ high, steady power demand helped spread a grid’s fixed costs over more usage. This finding is counterintuitive but consistent with research from 2025 that high-demand areas—including ones like North Dakota and Virginia that saw an explosion of data centers—“saw lower prices overall,” while prices rose in places like California where demand had declined"

"arguably the most viral water claim—that a single Google data center in Chile consumed more than 1,000 times the water supply of a town there—was later acknowledged by its author to be “off by a magnitude of 1,000.” They added that all U.S. data centers combined use less than 0.5 percent of the United States’ freshwater, and that “Google’s thirstiest facility, in Iowa, uses approximately what some golf courses do.”"

"U.S. data centers consumed about 17.5 billion gallons of water in 2023, according to Lawrence Berkeley National Laboratory. That sounds like a lot until you compare it with total public water supply: it’s about 0.3 percent. Beef production, cotton farming, and golf courses each quietly consume far more"

"a lot of “data center water usage” stats in various studies include non-potable water, indirect usage from nearby power plants, and water actually recirculated instead of consumed (thus causing little change to local supplies). Doing this can inflate water usage stats"

"Masley estimates that the water all American data centers will directly consume onsite in 2030 will be the equivalent of just 8 percent of U.S. golf industry usage and just 1 percent of the water used for American corn production. It’s also dwarfed by industrial usage for things like steel plants."

"The 100 billion gallons used in 2023 amount to a drop in the bucket (pun!) compared to what we use each year for golf courses (550 billion gallons) and lawns (3 trillion gallons)."

"data centers aren’t major sources of water pollution in host communities, mainly because most of their usage (e.g., for cooling) is in closed systems."

"Masley has since run the same exercise on land use and waste heat, reaching similarly benign conclusions. Air pollution too. The Taxpayers Protection Alliance, meanwhile, has debunked recent online claims that data centers generate intolerable ambient noise." (AI researcher Andy Masley has done yeoman’s work eon “data center water usage)

The source of the real problems. And how to fix them.

"None of this means data centers are perfect, and it’d be intellectually lazy to claim the pushback is entirely manufactured (though some of it surely is). Where real problems exist, however, the cause is almost always a policy choice rather than an immutable feature of data centers or AI. And that means the issues can be fixed with policy, too.

The most obvious place to start is by nixing all the subsidies that certain states and localities toss at data center projects—and their billionaire owners. Virginia’s data center sales and use tax exemption, for example, cost more than $1 billion in 2024 and is projected to cost almost double that in 2025, with benefits accruing to Google, Amazon, and other bajillion-dollar companies. Indiana’s subsidies are almost as big. Given demand for AI and every other internet-based service in the United States, these data centers will be built without all the subsidies. So, much like sports stadiums and corporate relocation incentives, the handouts are less about real economic development and more about companies playing states and localities against each other to get the sweetest deal for themselves—at taxpayers’ expense. Kill ’em all. The industry doesn’t need the help, and every subsidy dollar is one that not only diminishes data centers’ tax revenue benefits but also fuels a legitimate grievance by the pro-moratorium crowd—maybe their only legitimate grievance.

The next obvious solution is to liberalize the supply side of the electricity market, where the only serious economic risk likely exists. As Pielke notes, various studies confirm that local power prices wouldn’t rise if supply could meet new demand, yet—

The constraint is our national inability to build—hampered by permitting queues and interconnection backlogs that mean that new transmission takes six to 10 years to build—while new demand emerges much faster. A moratorium seeks to freeze demand to match a sclerotic supply side. A better response speeds the creation of supply to meet demand. That means permitting reform, faster interconnection, advanced transmission, and water system investment, including new supply. Blocking construction treats the symptom rather than the disease.

Other things, such as tariffs on transformers, electrical steel, solar panels, and other essential equipment, add to these burdens. Fix them, and some of the demand-side pressure from data centers will be released.

For the rest, Cato’s Travis Fisher and Huddleston have an even better idea: Let data center developers contract directly with private generators or build their own power, through “consumer-regulated electricity” (CRE) reforms:

[CREs] allow for privately financed, off-grid electric utilities to serve new customers under voluntary contracts. These utilities would be physically “islanded” from the regulated grid and would not be subject to economic regulation at the state or federal level. Because they would not interconnect with incumbent systems, CRE utilities would impose no costs, reliability risks, or stranded-asset exposure on existing customers.

Letting these giant, cash-rich companies pay their own way makes abundant (pun!) sense, yet utility regulations and related factors make that impossible in most places. And that, not data center demand, is the problem. Fortunately, as Fisher writes, there are some positive developments in this regard: A few states have already begun experimenting with CRE, and the White House’s voluntary Ratepayer Protection Pledge, signed by most hyperscalers, gestures at similar principles. But much more needs to be done—and fast.

Other data center issues, such as noise, pollution, and land use, have a simpler solution: localities enforcing the relevant laws they already have on the books and making local decisions about whether to permit data centers in their communities. NIMBYism drives a lot of the opposition to data center projects, and I think almost all of it is wrong. The bright side, however, is that it’s an exercise of local control, not a statewide or federal moratorium that blocks an entire industry. In the former case, a data center can move to more welcoming U.S. areas that will enjoy the direct benefits—and occasional headaches—that rapid development offers, while negotiating site-specific terms and mitigation directly with developers, the way Loudoun County did. America’s great Laboratories of Democracy can do their thing.

As Huddleston notes, on the other hand, turning a local land-use dispute into a federal ban would throw out the baby with the bathwater, generating economic harms that go far beyond a few NIMBY-captured communities." 

 

 

Ridley: Why our public sector is so unproductive

The enduring lessons of Jevons and Baumol

By Matt Ridley. 

"Agatha Christie once remarked that she had never expected to grow rich enough to own a car or poor enough not to have servants. The reason this strikes us as bizarre today boils down to two names that you hear invoked a lot in the tech industry: Jevons and Baumol. One is shorthand for the expansion of products or professions with rising efficiency, the other for the shrinkage of products or professions with stagnant efficiency.

There’s a pleasing chronological symmetry between these twin ideas: William Stanley Jevons coined the Jevons paradox in 1865; William Jack Baumol described Baumol’s cost disease exactly a century later in 1965.

In his pessimistic book The Coal Question, Jevons forecast peak coal and consequent economic catastrophe for Britain. Energy efficiency would not come to our rescue, he argued. “It is a confusion of ideas to suppose that the economical use of fuel is equivalent to diminished consumption. The very contrary is the truth.” If you double the efficiency of steam engines, you do not burn less coal, you install more engines and soon burn more coal. He was wrong about peak coal, as later pessimists were wrong about peak oil and peak gas, but right about increased consumption.

A modern example: light-emitting diodes (LEDs) use about 15 per cent as much electricity as incandescent bulbs. Do we save that difference? Only at first, then we install more lights, leave them on longer and build things like the Las Vegas Sphere, which uses as much electricity as 50,000 homes.

The tech guru Erik Brynjolfsson points out that: “Pilots became dramatically more productive and effective once jets were invented. Did that mean that we didn’t need as many pilots because now pilots could do more work? No. We consumers decided that we’re going to fly more than ever. So now a lot more people fly. And there’s more demand for pilots.” If supersonic commercial flight eventually takes off, the falling cost of pilots and flight attendants (in the air for less time) will only increase demand for air travel.

The price of a single transistor has fallen over half a century from about $1 to less than a millionth of a cent. So we not only buy more of them but spend more on them. As Alex Danco puts it: “At $1 per transistor, computers made sense for military calculations and corporate payroll. At a thousandth of a cent, they made sense for word processing and databases. At a millionth of a cent, they made sense in thermostats and greeting cards. At a billionth of a cent, we embed them in disposable shipping tags that transmit their location once and are thrown away.”

Drones, space launches and genome sequencing are being Jevonised right now. As for artificial intelligence, “Jevons paradox strikes again,” says Satya Nadella of Microsoft. “As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can’t get enough of.” Aaron Levie of Box says: “Jevons paradox is coming to knowledge work. By making it far cheaper to take on any type of task that we can possibly imagine, we’re ultimately going to be doing far more.” AI will mean more jobs for lawyers, not fewer.

Marc Andreessen muses that it is “like the Daniel Day Lewis character in There Will Be Blood worrying ‘but what will happen, once we’ve satiated their demand for whale blubber?!’ Well, it turns out that there were a lot more useful ways to consume energy than burning the midnight oil.” As the cost of AI tokens collapses, we will use vastly more of them for vastly more uses.

But here’s where the Baumol twin comes in. For every industry that experiences efficiency gains, there’s another that does not. And this latter industry inevitably becomes less affordable. Baumol’s first example was string quartets: violinists are no more productive but you have to pay them more to prevent them running off to become software engineers. The productive industries drive up the labour costs in the rest of the economy. Andreessen jokes that if a hole appears in the wall of your house in California these days it is probably cheaper to glue a flat-screen television over it than hire a builder to repair it: a Jevons-deflated cost beats a Baumol-inflated one.

The big question of our age is can AI drag Baumol-shaded industries back into the sunlight of Jevons? Can it make things like healthcare, education, or government switch from rising costs to falling costs?

I fear not in the case of government because of a bureaucratic version of the Jevons and Baumol effects. As Cyril Northcote Parkinson put it in an article in the Economist in 1955: “Politicians and taxpayers have assumed (with occasional phases of doubt) that a rising total in the number of civil servants must reflect a growing volume of work to be done. Cynics, in questioning this belief, have imagined that the multiplication of officials must have left some of them idle or all of them able to work for shorter hours. But this is a matter in which faith and doubt seem equally misplaced.”

Since 1997, the British public sector has seen zero increase in productivity. That is to say, the average civil servant generates about the same output today as he did three decades ago. Think about this for a second. Thirty years ago fax machines were high-tech, the internet was in its infancy, emails were new, Wi-Fi was scarce, mobile phones were voice-only. How is it remotely possible to be no more productive today than then?

We know the answer. Each email is now copied to a dozen people, each report is pasted and copied till it is twice as long, each Zoom call has five times as many attendees, each mobile call is followed up by three times as many WhatsApp messages – and each day at the desk is interrupted by a training session on transgender anticolonial sustainability. That’s a sort of Jevons-Baumol effect: a Jevol?

Keeping Cool: The Air Conditioner That Changed America

The time price of air conditioning has fallen 98.6 percent since 1952. That ordinary luxury saves lives every summer.

By Gale L. Pooley. He teaches US economic history at Utah Tech University. Excerpt:

"One of the great triumphs of entrepreneurial capitalism is how quickly air conditioning traveled the familiar path from luxury to necessity. What began as an expensive convenience for a tiny elite became, within a generation, affordable to ordinary families. The market did not merely invent comfort — it democratized it.

In their report Time Well Spent: The Declining Real Cost of Living in America, Michael Cox and Richard Alm found that a 5,500-BTU air-conditioning unit cost about $350 in 1952. At the time, entry-level workers earned roughly 83 cents an hour, putting the time price at 422 hours.

Today, Walmart sells a far more efficient 6,000 BTU air-conditioning unit (with a remote control) for only $115. The current hourly wage for limited-service restaurant workers is around $19 an hour, putting the time price at six hours.

The time price has decreased by 98.6 percent. For the time it took US workers to earn the money to buy one unit in 1952, they get 70 today.

If air conditioning saves lives, why don’t more Europeans have it?

Europe’s electricity prices are typically much higher than the US, driven by higher taxes, network costs, renewable energy mandates, and energy import dependence. Customers in the US pay 17 to 19 cents per kilowatt-hour (kWh) compared to 25 to 32 cents in Europe. This means Europeans pay roughly 47 to 68 percent more per kWh than US customers.

Americans are also much richer than Europeans. According to World Bank data, American gross domestic product (GDP) per capita was $84,809 in 2024, while the European Union’s was 25 percent lower at $63,585. That $21,224 difference could buy a lot of comfortable cooling.

The European Union also prioritizes environmental targets over human comfort by imposing strict regulations for heating and cooling, making these amenities much more costly. The commission encourages citizens to use fans instead of air conditioning. Imagine the government doing that in Phoenix and Atlanta in July. Italy, Greece, and Spain even announced temperature limits in public spaces during the 2022 heatwave in an effort to meet these environmental objectives. Spain limited air conditioners to be set no lower than 80°F. No wonder European productivity is 38 percent lower than the US.

Historic preservation laws and strict landlord rules frequently ban exterior window units to maintain aesthetic uniformity.

While air conditioning ownership increases households’ electricity consumption, it may be a small price to pay for comfort and avoiding death.

The problem is not the climate but the policy mindset. Too many European regulators approach energy and technology through the ideological lens of scarcity rather than creative innovation and human flourishing. One reason such policies persist is that the officials who design them are largely insulated from the consequences of their decisions and rarely experience their costs directly. Instead, those costs are borne by millions of ordinary citizens.

Air conditioning is not ultimately a story about cooling. It is a story about knowledge. It transformed oppressive heat into comfort, inhospitable regions into thriving communities, and summer misery into year-round productivity. Coal, copper, and electricity become valuable only after humans discover how to harness them. The history of air conditioning is the history of knowledge triumphing over nature’s constraints.

The ultimate resource is neither energy nor matter. It is the infinite capacity of human beings to learn, create, and discover."

Human aspiration is a disposition, not an exhaustible resource. Mokyr showed that civilizations which honor that aspiration grow, and those that suppress it stagnate

See The Lump of Labor Fallacy in the Age of AI by David Hebert.

"In conclusion, the problems with the lump of labor fallacy were settled long before AI arrived. Smith understood that human aspiration is a disposition, not an exhaustible resource. Mokyr showed that civilizations which honor that aspiration grow, and those that suppress it stagnate. The lump of labor fallacy gets the economics wrong because it makes fundamental errors in human nature and economic history.

But wrong ideas with organized constituencies do not stay defeated. The longshoremen’s contract shows what happens when the fallacy wins a political victory. If AI policy follows the same template, the damage will be measured not in port fees but in trillions of dollars of foregone growth and millions of jobs that never get created. The fallacy is intellectually bankrupt. Whether it remains politically solvent is the question that actually matters."

Competition intensifies in broadband and media markets

See Comcast Plans Company Split as Competition Escalates by Jessica Toonkel and Gareth Vipers of The WSJ. Excerpts:

"Comcast up pointing triangle plans to separate its media and connectivity businesses, dismantling an earlier bet on combined entertainment and distribution as it navigates intensifying competitive pressure."

"company leaders believe the media and connectivity businesses should stand alone as publicly traded companies, providing more opportunity to pursue deals and better compete."

"it became clear greater flexibility would help the company navigate the increasingly challenging broadband and media markets, people familiar with the matter said."

"Comcast has been working to stem broadband and cable TV subscriber losses"

"Broadband businesses have been challenged by cellphone carriers offering home internet service beamed over the air. Elon Musk’s Starlink satellite connectivity company has added further competition for providers." 

AI is making it easier for one worker to do tasks that once required a small team

See Me, Myself and AI by Liya Palagashvili of George Mason University’s Mercatus Center. Excerpts:

"AI is making it easier for one worker to do tasks that once required a small team."

"Since early 2024, solo business applications have risen nearly 27% in professional services, information, education, finance and insurance—sectors the that have among the highest AI-adoption rates. In construction and wholesale trade, where AI is less likely to enable independent work, solo business applications have been essentially flat."

"In 2022 and 2023, solo applications moved similarly in sectors with both high and low AI exposure. The divergence emerged after 2024 and then took off."

"Business applications with payroll intent fell by 6.4% in AI-exposed sectors, while applications unlikely to hire employees rose by 26.8%."

"Among occupations most exposed to AI, solo self-employment rose about 20% from 2022 to 2025. In the least AI-exposed occupations, it barely moved."

"it’s becoming cheaper to operate as a one-person business" 

Chuck Schumer’s Chip Shortage

A Micron plant in New York is years behind schedule for all the reasons you’d expect

WSJ editorial. Excerpt:

"Consider Micron’s massive fabricator project in upstate New York, which it announced in October 2022. “With the CHIPS and Science bill I wrote and championed as the fuse, Micron’s $100 billion investment in Upstate New York will fundamentally transform the region into a global hub for manufacturing,” New York Sen. Chuck Schumer boasted."

"The 2022 Chips Act provided some $53 billion, plus a 25% investment tax credit, to subsidize U.S. chip-making."

"Congress in 2024 passed a law exempting some semiconductor projects from the National Environmental Policy Act’s stringent environmental reviews. But the exemptions don’t apply to Micron’s project."

"it includes hundreds of acres of wetlands and forestland that are nesting areas for endangered bats. This makes permitting and building more complicated. Trees can only be chopped down when bats aren’t nesting—i.e., from November to March."

"Construction was supposed to start two years ago, but tree clearing didn’t begin until this past January"

"environmental impact statement numbered more than 700 pages" 

Ozempic Sat Unused for Decades Because Invention Is Not Enough

Pfizer knew GLP-1s worked in 1990, but didn’t see their potential. The 30-year detour shows entrepreneurship matters as much as raw invention.

By Per Bylund. 

"Led by Ozempic and Wegovy, glucagon-like peptides (GLP-1s) have become a global phenomenon, with one in eight US adults currently taking one. Those two branded compounds, both made by Novo Nordisk, emerged from attempts to develop a diabetes drug. It effectively lowers blood glucose, slows gastric emptying, and reduces hunger, leading many patients to experience profound weight loss. In a world plagued by increasing obesity, the drugs’ utility extends far beyond diabetes treatment. So why did the formula sit untouched for 30 years after it was licensed?

Ozempic is a story of pharmacological success, but also of entrepreneurial failure. The tale provides a strong reminder that inventions and discoveries mean little unless they are combined with sound entrepreneurial judgment.

According to a paper published in the Perspectives in Biology and Medicine, a startup produced a GLP compound in the late 1980s, and pharma giant Pfizer sponsored human trials that confirmed the drug’s efficacy in reducing blood glucose levels and slowing gastric emptying. One member of the startup team, Jeffrey Flier of Harvard, explained what happened next:

I was shocked when told that senior Pfizer leadership had concluded that there would never be another injectable therapy for diabetes other than insulin. What led them to this conclusion was never explained….I had been deeply impressed by their rapid decision to invest in our company, and I was equally dumbfounded by their decision to end their investment despite convincing early evidence of the program’s success.

Confident in its own conclusions, Pfizer pulled the plug on the drug in 1991. The startup folded.

Under the terms of Pfizer’s agreement, the license remained with the Boston hospital where researchers discovered GLP-1’s mechanism and conducted the human trials. It was then acquired by Novo Nordisk in 1992, where scientists used it to develop what eventually became semaglutide, the pharmaceutical sold as Ozempic and Wegovy. While it is unclear whether, as Max Marchione put it on Twitter, the GLP-1 agonist data simply “sat in a filing cabinet for 30+ years,” Pfizer’s decision to abandon the project likely delayed its development.

While it is unclear whether, as Max Marchione put it on Twitter, the GLP-1 agonist data simply “sat in a filing cabinet for 30+ years,” Pfizer’s decision to abandon the project certainly delayed its development. Had the company continued investing in the research, it might have brought the drug to patients years earlier—and captured a significant share of what has become a $190 billion market.

Clearly, mistakes were made in the development of Ozempic, but what’s notable is that the product’s success required far more than the idea. Even a great idea is not a product and may never become one — much less a successful one. Business history is filled with cases in which inventors appear to have been deprived of the rewards of their discoveries. Many great inventions were, in some sense, stolen ideas commercialized by someone other than their inventors. George Westinghouse bought patents from Nikola Tesla, and undertook illuminating the nation while Tesla, the lone genius, struggled with poverty. Elias Howe invented the sewing machine but lost most of the revenue to Isaac Singer, who was only later compelled to pay the inventor royalties. Antonio Meucci invented the telephone but couldn’t afford to secure or defend patents from Alexander Graham Bell’s enthusiastic dissemination of the device. 

Inventors frequently failed to recognize the full market potential of their ideas. Entrepreneurial outsiders notice the discovery, develop it into a desirable technology or product, and implement strategies for manufacturing, distributing, and marketing the invention. 

Some unethical behavior, fraud, exploitation, and outright stealing certainly does exist in these stories. But it would be a mistake to reduce this entrepreneurial instinct to taking advantage of or free-riding on a mistreated genius who would otherwise have realized the great benefit, himself. In a very real sense (as GLP-1 development demonstrates), what actualizes the value of an idea is the execution: operationalizing a discovery into a product or service that people find valuable. The idea has relatively little value, compared to the scalable solution built upon it. 

We typically do not learn about it much in school, but there is such a thing as second-mover advantage. This phenomenon (that the first mover is not profitable but the second mover is) is often explained in terms of avoiding the costly mistakes that the first mover makes. But cost is not the true story. Second movers recognize — imagine — a new idea’s utility to some market segment. Like Novo Nordisk and George Westinghouse, they strive to make the implementation of the idea as valuable as possible, positioning it as valuable to potential customers. The creation may be every bit as valuable as the idea, and often more so. 

Henry Ford, for another historical example, was not the inventor of the automobile, but the innovator of the affordable car. Existing manufacturers of automobiles did not recognize the potential appeal of their horseless vehicles. Henry Ford did — and made it happen. Far from the first mover, or even the second mover, he was the first to recognize what a mass market of ordinary people wanted from a car. The other producers did not. Ford transformed a toy for the wealthy into practical transport for the ordinary.

The same type of story can be told about many successful innovations. The invention (the new thing) might have limited value, until its utility is captured, marketed, and made available to people. Entrepreneurs and investors look for the value proposition in new ventures: what would make buyers want this new thing? They ask not “is this a new thing?” but “is this new thing generating value someone is willing to pay for?” The questions may have the same answer but often do not.

The story of Ozempic is one of failed entrepreneurship, but also of its eventual success. Pfizer, judging from how the story has been told, did not recognize the value of the drug. Viewing GLP-1s as a treatment for diabetes, and nothing else, Pfizer execs failed to imagine how other consumers might value these clinical effects. 

Perhaps they were right — from the perspective of treating diabetes, another injectable may not be necessary. But they were wrong regarding the value of the drug, which arises from a different use by a different market segment. 

The Ozempic craze of today is not driven by diabetics seeking to manage their disease, which Pfizer viewed as its only potential application. Millions of non-diabetics now choose the compound for other reasons. Pfizer completely missed that value proposition."

We’re Preparing for the Wrong AI Labor Crisis

Mass unemployment is unlikely. AI will reorganize the white-collar corporate workforce, not destroy it.

By Stephen Lewarne. He is a professor of economics at Franciscan University. Excerpts:

"the notion that the economy faces mass technological unemployment doesn’t fit the evidence"

"the broader market continues to show relatively stable aggregate demand for labor. In March, the unemployment rate stood at 4.3%, close to both the Federal Reserve’s estimate of long-run normal unemployment and Congressional Budget Office projections for the coming decade. Total nonfarm payroll employment increased by 178,000 jobs during the month, while healthcare added 76,000 jobs and averaged roughly 29,000 new jobs a month over the prior year."

"U.S. entry-level job postings have fallen roughly 35% since January 2023, with highly AI-exposed entry-level postings declining more than 40%."

"Employer surveys indicate a substantial shift away from GPA-based screening and toward skills-based hiring. Employers increasingly emphasize demonstrated competencies, project-based experience and practical problem-solving abilities." 

[there is] "a labor-market transition more complicated than conventional automation narratives suggest." 

The Road to AI State Socialism

Bernie Sanders sees Trump industrial policy and decides to raise the government stakes

WSJ editorial. Excerpts:

"Mr. Sanders wants to force companies to hand over half of their equity to the government."

"this would be a government expropriation. It would violate the Fifth Amendment’s prohibition on government taking property without just compensation. To pay this tax, companies would have to issue new shares to the government diluting current shareholders. Or they could buy back half their shares from private investors and hand them to the government."

"One model is China’s state-owned enterprises, which are an albatross on its economy. Political favoritism and government interventions have led to economic inefficiencies. Hence China’s partially state-owned Semiconductor Manufacturing International trails TSMC and Samsung in chip fabrication." 

Gavin Newsom Wants an AI New Deal

The California Governor is tilting toward an even larger entitlement state

WSJ editorial. Excerpts:

"He suggested imitating Europe’s generous wage replacement programs and job protections. Perhaps he has missed Britain’s debate, notably of late even in the Labour Party, over why a million young men and women have left the workforce while on the government dole."

"If the government makes it hard for businesses to lay off workers, they will be more reluctant to add other jobs and hire young people with less experience. That’s why the youth unemployment rate in France is upward of 20%."

"if workers can make nearly as much unemployed as they do working, many will stay home. That was one lesson from the pandemic when Congress juiced unemployment benefits and transfer payments."

"The top marginal tax rate in California on wage income over $72,725 (including a disability payroll tax) is 10.6%."

"state’s $20-an-hour minimum wage for fast-food workers"

"California is tied with Nevada and Delaware for the nation’s highest unemployment rate (5.3%), followed by Oregon and Washington (5.2%), Illinois (5.1%), Connecticut and Michigan (5%). You don’t need AI to discern what they have in common. With the exception of Nevada, the states are run by Democrats heavily influenced by public unions." 

Pope Leo’s AI Manifesto

His defense of human agency is welcome but not his faith in the state

WSJ editorial. Excerpts:

"When it comes to AI, his encyclical mostly recites the most pessimistic prophecies. He largely dismisses AI’s potential benefits, such as faster and less expensive drug development and medical cures. His call for more government regulation of AI echoes opponents of capitalism like Sen. Bernie Sanders and Rep. Alexandria Ocasio-Cortez."

"We’ve been around a while and don’t recall when anyone relied “solely” on the free market. Western governments now snatch as much as half of GDP and regulate nearly every part of business life."

"Throughout history the diffusion of technology has democratized information and improved living standards, especially for the poor. The internet and social media have enabled people living under repressive regimes to share information"

"He calls for regulation of algorithms that “influence credit distribution, personnel selection or access to services and opportunities” and “measures to ensure equity: taxation, social protection and industrial policies.”"

"government control is likely to result in an even greater concentration of power. Regulation tends to protect incumbents and retard competition." 

Jeff Bezos Earned His Fortune

The Amazon founder’s innovations save customers 22 hours a year on average, giving them the gift of time

By Marian L. Tupy. Excerpts:

"Amazon didn’t become valuable by force. It became valuable because hundreds of millions of people chose to use it."

"Amazon saved them [consumers] time, money, effort or uncertainty. Sellers weren’t forced to use Amazon’s marketplace. They did so because it gave them access to demand."

"The value Amazon created is harder to see because it is dispersed. A mother who doesn’t drive to a store to buy diapers doesn’t appear in an economic headline. A small business that reorders supplies in two minutes doesn’t make the evening news."

"Suppose an hour of labor is worth about $64, roughly the average gross domestic product per hour worked in the countries in which Amazon operates. If Mr. Bezos’ fortune corresponded to the total value that Amazon created, his $275 billion would represent about 4.3 billion hours of saved time. Divided among Amazon’s more than 300 million active customers, the saving comes to about 14 hours per customer over Amazon’s life."

"entrepreneurs don’t capture all the value they create. The Nobel Prize-winning economist William Nordhaus estimated that innovators keep only a small share of the social value—roughly 2%—produced by their innovations."

"A single avoided trip to a store can save 30 minutes. Finding a product online instead of driving to three retailers can save an hour. Reading reviews can reduce the chance of buying the wrong product."

"Amazon Web Services lowered the cost of starting and scaling companies. It gave firms computing capacity without the old capital expense."

"Amazon also forced competitors to improve." 

The US is Building a Lot More Data Centers Than Five Years Ago, But We Are Still Building More Warehouses

By Jeremy Horpedahl.

"Data centers seem to be popping up everywhere. And based on the value of current construction, the US is indeed building a lot more data centers than we were in 2020 or 2021, about four times as much data center construction (inflation adjusted).

But… did you know that we build a lot more good-old manufacturing than data centers? Almost four times as much in recent months. And that’s even after a decline in manufacturing construction over the past year and a half.

The US also builds about the same amount of warehouses and chemical plants as we do data centers. Data centers may exceed those two categories in a few years, but for now they are pretty similar.

Keep in mind that manufacturing and chemical facilities also use a lot of electricity and water, and have plenty of local negative externalities! Warehouses probably have a lot less resource consumption and external effects, but it’s not zero either.

Are data centers popping up everywhere? Well, people are certainly noticing them. But so are lots of other types of buildings, which rarely register more than a peep from concerned citizens and local media, unless there is some clear and obvious external effect."

 

 

DATA CENTER MYTH V. REALITY

From Taxpayers Protection Alliance. They used 0.06 percent of water in 2025 in Maryland and 2.5 percent of electricity.

 

What Would Jefferson and Madison Make of Musk and Altman?

America’s Founders and Adam Smith knew better than to entrust the future to philosopher-kings

By Jason Riley. Excerpt:

"Theories about the need for a “philosopher king” or “great man” to advance society date back centuries. Intellectual figures from Plato to Machiavelli and Thomas Carlyle emphasized personal traits such as superior wisdom and exceptional moral character in choosing leaders. The idea was to find these extraordinary men, put them in charge, and align policies with their understanding of the common good. Adam Smith, by contrast, argued that free enterprise and the uncoordinated pursuit of individual self-interest would lead to better outcomes for more people. Societies should rely on market forces and voluntary exchange rather than on do-gooders.

March marked the 250th anniversary of Smith’s seminal text, “The Wealth of Nations,” published the same year as the Declaration of Independence. As we reflect on America’s milestone, it’s worth noting that the Founders shared Smith’s skepticism of philosopher-kings and the approach to choosing leaders that today’s AI poohbahs seem to have embraced.

“What the American Constitution established was not simply a particular system but a process for changing systems, practices, and leaders, together with a method of constraining whoever or whatever was ascendent at any give time,” Thomas Sowell wrote in his book on social theory, “The Quest for Cosmic Justice.” “Viewed positively, what the American revolution did was to give the common man a voice, a veto, elbow room and a refuge from the rampaging presumptions of his ‘betters.’ ”" 

The Panic Industry’s New Target

A generation coached to fear climate change is now fretting over AI and data centers

By Barton Swaim. Excerpts:

"Where does Mr. Schmidt think young people got the idea that “the climate is breaking”? Where did the “fear” he laments come from? In part from the scores of climate-panic groups to which the Schmidt Family Foundation’s 11th Hour Project has granted hundreds of millions of dollars over the last 20 years. One detail particularly amuses: When 11th Hour first appeared, in 2006, it funded screenings of Al Gore’s “An Inconvenient Truth,” a documentary designed to terrorize viewers with 90 minutes of bleak prophecies, now happily exploded. The outfit, like scores of others founded and funded by other progressive billionaires, spends its resources opposing fossil-based energy and trumpeting the dangers of a warming world."

"Last year Bill Gates posted an essay on his website purporting to scold alarmists and express his own moderate view of the climate question, namely that it is a “very important problem” but doesn’t doom civilization. You have to wonder, then, why Mr. Gates has sent so much of his money to Arabella Advisors. Arabella, a pass-through entity now called Sunflower Services, funds a dizzying array of groups that exist to alarm the public over an imminent climate apocalypse and to portray carbon-emitting energy as an existential threat to humanity."

"nobody staffing the multibillion-dollar ganglion of climate nonprofits and activist groups plans on taking a more measured view of the coming cataclysm."

"modern data centers first appeared in the 1990s, when companies learned the benefits of fast internet connectivity and information storage."

"the people showing up at county council meetings to protest the construction of a data center . . . got their talking points from national nonprofits supported by some of the same moneyed outfits the Schmidt and Gates foundations spent the last two decades bankrolling." 

Does Providing Laptops Improve Educational Outcomes?

From Jeffrey Miron.

"The One Laptop Per Child program was supposed to bridge the digital divide for children in developing countries.

To assess the long-term effects of the program, one study performed

a large-scale randomized evaluation of the OLPC program implemented by the Peruvian government in rural primary schools, using administrative and survey data from 2007 to 2019.

The study found that the program had

no effects on students’ exam or test scores or on their likelihood of completing primary or secondary school or enrolling in a university… [T]he program [also] reduced the fraction of primary students who advanced to the next grade by 1 percentage point between 2009 and 2016.

Despite lengthy teacher trainings,

the program had no significant effects on teachers’ digital skills … [and it] led to limited academic use of laptops and generated few benefits beyond basic digital literacy, which partly explains the absence of effects on achievement and attainment."