The standard story about Big Tech compensation is that companies pay top dollar because the best engineers are genuinely rare and genuinely valuable. That’s true, but it’s incomplete. A significant portion of elite engineering hiring at the largest tech companies serves a second purpose that rarely gets named directly: keeping that talent away from everyone else.
This is talent hoarding, and it functions as a form of competitive moat that regulators, economists, and startup founders consistently underestimate.
The Economics of Defensive Hiring
When Google or Meta hires a machine learning researcher at $400,000 in total compensation, the calculus isn’t purely about what that person will build internally. It’s also about what they won’t build for anyone else. A senior engineer who joins a large platform and spends three years on incremental feature work is, from a competitive standpoint, a neutralized threat. The company has effectively purchased an option: the right to deploy that person’s skills, or simply the right to ensure a competitor doesn’t.
This logic becomes clearest when you look at hiring patterns around emerging technology categories. In the years immediately following the public emergence of large language models, major tech companies accelerated AI hiring far beyond their demonstrated ability to deploy those researchers productively. Insiders at several companies described teams growing faster than projects existed to absorb them. The hiring wasn’t irrational. It was defensive.
The same pattern played out during the early smartphone era, during the cloud infrastructure buildout, and during the initial wave of autonomous vehicle development. Waymo, Cruise, and Argo AI collectively employed thousands of engineers in a market where, for years, no autonomous vehicle was generating meaningful revenue. The companies weren’t building businesses at that headcount. They were occupying territory.
Compensation as a Barrier to Entry
When a small startup competes for an engineer who has a Google offer on the table, it isn’t just competing on salary. It’s competing against the full compensation architecture that only a handful of companies can build: vesting schedules designed to retain over four years, equity grants that refresh before the previous grant fully vests, and benefits infrastructure that costs tens of millions annually to maintain at scale.
The result is a structural barrier. As we’ve argued before, much of this compensation resembles a lottery ticket more than a wage, but the lottery ticket still pulls talent toward large incumbents and away from the startups most likely to challenge them. A founding engineer at a Series A company typically takes on real financial risk. An L5 at Amazon takes on almost none. For someone with a mortgage and a family, that asymmetry is decisive.
Startups can occasionally overcome this with the genuine upside of early equity. But that works only for a narrow slice of candidates at a narrow slice of companies. The top decile of engineers, the people who could actually build something that threatens a major platform, often have enough financial stability to be risk-averse. Incumbents know this.
The No-Poaching History Is Instructive
The most explicit version of talent hoarding was exposed in a 2010 Department of Justice investigation into agreements among Apple, Google, Adobe, Intel, Intuit, and Pixar not to cold-call each other’s employees. The agreements were designed to suppress wage competition and reduce the mobility of engineers across companies. The companies eventually settled a related civil lawsuit for a reported $415 million.
What made those agreements notable wasn’t that they were unusual. It was that they were written down. The underlying incentive, limiting the movement of skilled labor to reduce competitive threat, hasn’t gone away. It’s just expressed through legal mechanisms: non-compete clauses (still enforceable in many states), garden leave provisions, and equity structures that make leaving financially painful regardless of what a new employer offers.
The FTC’s 2023 proposed rule to ban non-compete agreements recognized this explicitly. The agency estimated that non-competes suppress wages and reduce labor mobility across roughly 30 million American workers. The tech industry’s share of that figure, concentrated among its highest earners, represents a particularly consequential drag on the formation of new companies.
The Counterargument
The reasonable defense of aggressive Big Tech hiring is that these companies are simply paying market rates for scarce skills, and the market is working as intended. If a researcher can command $400,000 from Google, that reflects genuine value creation, not anticompetitive hoarding. The companies aren’t forcing anyone to accept their offers.
There’s something to this. Talented engineers who join large companies do, often, produce real work. Google’s infrastructure advances and Meta’s open-source contributions represent genuine outputs, not just warehoused talent. And the alternative, a world where large companies are somehow restricted in who they can hire, raises obvious problems.
But this argument conflates two things. The question isn’t whether large companies have the right to hire freely. They do. The question is whether a specific subset of hiring decisions is better explained by competitive neutralization than by productive deployment. The evidence, particularly during technology transitions when large companies consistently hire faster than they can onboard, suggests the answer is frequently yes.
The market-rate defense also ignores the structural nature of the problem. When a handful of companies can collectively set the compensation floor for an entire talent category, “market rate” loses its neutral meaning. The market rate for AI researchers isn’t determined by supply and demand across a competitive hiring landscape. It’s substantially determined by what the five or six companies with the most to lose from AI disruption are willing to pay to control the supply.
What This Costs the Industry
The practical consequence is a slower-moving startup ecosystem than the raw volume of venture capital would suggest. Startups that can’t compete for senior engineering talent have three options: hire less experienced people and accept a capability gap, wait until engineers leave large companies voluntarily (typically on a four-year vesting cycle), or pay acquisition prices for tiny teams through acqui-hires. None of these are efficient.
This matters most in precisely the domains where new entrants would create the most competitive pressure on incumbents. AI infrastructure, semiconductor design, distributed systems at scale: these are areas where the talent pool is thin, the barrier to entry created by compensation is highest, and the incumbents have the most to protect.
Talent hoarding isn’t a conspiracy. It doesn’t require coordination or explicit intent. It emerges naturally from the incentive structure facing any large company that can afford to be defensive. But the fact that it’s rational doesn’t make it neutral. It’s a market distortion, paid for through compensation packages, enforced through vesting schedules, and measured in the startups that never got built.
