The Plumber Speech

Larry Fink did not discover the skilled trades shortage. He discovered that talking about it, publicly, at scale, from the top of the largest asset manager on earth, sounds like leadership.

In his 2026 annual shareholder letter, the BlackRock CEO declared that America needs more plumbers and fewer lawyers. That the post-war consensus (go to college, get a degree, enter a profession) had been overdone. That the economy had mispriced intellectual labour upward and manual competence downward, and that artificial intelligence was about to make that mispricing impossible to ignore.

He is right about most of that. The question is what he does with being right.

What Fink Actually Said

The shareholder letter is worth reading directly. Fink does not hedge. "We probably overdid it," he writes of the post-war push toward universal college attendance. "We said to all the young people, go to college, go to college, go to college." The result, in his framing, was a generation of bankers, lawyers, and media professionals who probably should have been electricians, and an economy that now faces a structural shortage of the workers it most urgently needs.

The moment that landed hardest was a comment he says he made directly to members of the Trump administration: "We're going to run out of electricians to build AI data centers."

That sentence is doing a lot of work. It frames the skilled trades shortage not as a social problem, a question of dignity, mobility, and opportunity, but as an infrastructure bottleneck. The reason to produce more electricians is not that electricians deserve better than they got from the college-or-nothing consensus. It is that the data centers cannot be built without them. The urgency is operational, not moral.

The $100M Problem

BlackRock announced a $100 million investment to train 50,000 workers in skilled trades over five years. The announcement received coverage proportionate to the number. A hundred million dollars is a large sum in most contexts.

In this context, the math is uncomfortable. The United States needs an estimated 300,000 additional electricians over the next decade. More than 200,000 currently working electricians are approaching retirement. The net deficit, before accounting for rising demand from AI infrastructure expansion alone, is in the hundreds of thousands. BlackRock's programme trains 50,000 people over five years.

Google has pledged $15 million toward skilled trades training. Microsoft's Brad Smith has called the electrician shortage "the single biggest challenge" for data center expansion. These are companies collectively spending hundreds of billions of dollars on AI infrastructure, and their combined contribution to solving the labour shortage that threatens it is a fraction of a rounding error.

The $100 million is $2,000 per trainee. A single large-scale data center project can cost $2 billion or more, with electrical work accounting for 45 to 70 percent of construction costs. The arithmetic is not subtle.

The Model That Built This Shortage

The college-for-everyone model did not emerge from nowhere. It was actively promoted by the financial sector, by universities with incentives to expand enrolment, and by employers who began requiring degrees for roles that had never needed them. Credential inflation was a rational response to a system that rewarded credentials, and that system was shaped, in part, by the same capital that is now expressing disappointment in its results.

BlackRock is not uniquely responsible for this. But the firm that manages $11 trillion in assets did not spend the last three decades funding vocational training programmes or lobbying against the degree requirements that locked tradespeople out of advancement paths. The concern about over-credentialism is real and legitimate. The timing of its expression, at a moment when the shortage has become operationally inconvenient for AI expansion, is worth noting.

Fink's admission that the system "overdid it" carries more weight when it comes from someone who helped build the system. It also raises a question he does not answer: what is owed to the people who followed the advice that is now being revised?

The Numbers Behind the Shortage

The labour market data for recent college graduates is genuinely alarming, and it is not confined to trades versus degrees. Unemployment among workers aged 22 to 27 with college degrees has reached 5.6 percent, the highest since 2013, excluding the pandemic. Job postings for degree-holding entry-level workers fell 16 percent between August 2024 and August 2025. Applications per available role rose 26 percent over the same period.

The supply of credentialled workers has not declined. The demand for them has. The hiring door for young workers is already narrowing, and it is narrowing fastest for precisely the roles that the college-for-everyone model was designed to produce: analysts, researchers, junior lawyers, entry-level programmers. These are not manufacturing workers displaced by robotics. These are knowledge workers being displaced by systems that can do knowledge work.

Journeyman electricians currently earn between $120,000 and $200,000 annually, including overtime and supervisory work. The premium on practical, physical competence is not a future projection. It is already the market price. Fink is not predicting this shift. He is describing what has already happened and calling it a policy lesson.

Why Fink Is Right About Lawyers

There is one part of the plumber speech that deserves more credit than it typically receives: the lawyer warning is not rhetorical. It is structural.

Legal work, at its core, is the application of rules to facts. You gather information, you match it against a body of law, you identify risk, you draft a response. These are precisely the operations that large language models perform at scale, with increasing accuracy, and at a fraction of the cost of a billed hour. McKinsey estimates that 44% of all legal tasks are technically automatable with current AI. Goldman Sachs puts 17% of legal jobs at direct displacement risk, a figure that understates the picture because it counts roles, not the tasks within them.

The tools are already deployed. Harvey AI and CoCounsel (the latter a partnership between Thomson Reuters and OpenAI) are live inside major law firms. LexisNexis has integrated generative AI into its core research platform, the one that trained most of the lawyers now using it. Clio's Vincent AI handles client intake, deadline extraction, invoice generation, and document drafting. A 2025 survey found that 79% of legal professionals already use AI in some form in their practice, and 73% plan to expand that use.

Contract review time has been reduced by 75% in documented case studies. Document drafting that once required a junior associate billing four hours now takes minutes. Harvard Law School research found 100x productivity gains on specific tasks at AmLaw 100 firms. Notably, none of those firms currently anticipate reducing headcount as a result. But that is not the relevant metric. The relevant metric is what happens to the next generation of junior lawyers who were supposed to learn the profession by doing exactly the work that AI now does faster and cheaper.

The deeper problem is not replacement. It is deskilling. The junior lawyer's traditional role of document review, legal research, and first-draft motions served a dual purpose: it produced work product and it built expertise. If that work is absorbed by AI pipelines, the senior lawyers of 2035 will have been trained on what, exactly? The profession is bifurcating: top-tier strategic advocacy and complex courtroom work remain human, while the commoditised layer that used to feed talent into that tier is being compressed from below.

Fink says we need fewer lawyers. The market is already agreeing with him, not by eliminating lawyers, but by eliminating the conditions that produce the next generation of them.

Who This Transition Is Actually For

The advice to learn a trade is sound for someone choosing what to do with their life now. It is considerably less useful for the 28-year-old programmer who took on student debt to enter a field that is being compressed by the same AI tools that are creating demand for electricians. Retraining narratives tend to assume a fluidity of human capital that does not match how lives are actually built.

Amazon's explicit use of AI to justify 16,000 layoffs did not come with a skills retraining programme proportionate to the displacement. Meta's elimination of 15,000 corporate roles while doubling AI capital expenditure was not accompanied by an electrician apprenticeship scheme. The companies at the centre of the AI and jobs debate are not meaningfully investing in the transition they are accelerating.

The people most affected by this transition are not abstract. They are people in existing careers that are contracting, or people entering a labour market that is rewarding different skills than it rewarded two years ago. The advice to "learn plumbing" is structurally correct for the incoming generation. For the generation in the middle, it is a suggestion without infrastructure.

The Infrastructure Paradox

There is an irony embedded in this situation that deserves to be named. The AI expansion driving white-collar displacement is physically constrained by a shortage of the workers who cannot be automated. You cannot train a language model to wire a data center. You cannot deploy a chatbot to install the cooling systems that keep the servers running. The electricians and HVAC technicians are the rate limiter on the technology that is eliminating the roles of people who went to college to avoid becoming electricians and HVAC technicians.

This is not a small irony. The $10 trillion in infrastructure investment that the United States needs by 2033 (Fink's own figure) is physically impossible to deliver without the workers that the system spent decades undervaluing. The hidden costs of AI infrastructure are already showing up in electricity bills and grid capacity constraints. The labour cost is showing up next.

The companies building AI at scale have, in effect, created the conditions for the shortage that now threatens their expansion. That is not a failure of foresight unique to any single firm. It is a structural consequence of an industry that optimised for software and neglected hardware, in the broadest sense of that word.

"We're going to run out of electricians to build AI data centers."

Larry Fink, CEO of BlackRock

What Would Actually Help

The things that would actually close the gap are not secrets. Federal investment in vocational training at scale: not $100 million across five years, but programmes measured in the tens of billions, sustained over a decade. Removal of the regulatory and credential barriers that prevent mid-career transitions into licensed trades. Union density in skilled trades to ensure that the wage premium persists as labour supply increases, rather than being competed away. Honest accounting of who is bearing the cost of the AI transition, and policy frameworks that distribute that cost differently than the market currently does.

None of these are things a shareholder letter can deliver. Some of them are things that firms managing $11 trillion in assets are positioned to influence, if they choose to use that position for something other than well-timed admissions.

Fink is correct that the education system made promises it could not keep. He is correct that the trades were undersold. He is correct that AI is accelerating the moment of reckoning. The plumber speech is not wrong. It is just insufficient, and delivered by someone whose industry helped write the script it is now revising.