Human Capital Is Being Repriced
Block cut 40%. The stock surged 22%. A transmission analysis for allocators.
Executive Summary:
Block cut 40% of its workforce while growing gross profit 17% YoY. The stock surged 22%+. Wall Street is rewarding companies that replace headcount with AI, and CEOs are paying attention.
AI breaks the linear link between revenue growth and headcount growth in software companies, where personnel costs typically exceed 50% of revenue. This is an operating leverage regime change.
Macro impact is narrow: tech employment is under 5% of the U.S. workforce. The real transmission is K-shaped wage bifurcation and talent reallocation friction into traditional industries.
When the marginal cost of producing software approaches zero, the moat shifts from technical capability to distribution and proprietary data. Software valuation premiums need recalibrating.
Your portfolio carries an implicit assumption that human capital pricing is stable. Every path forward has costs, including doing nothing.
The Signal
Block cut 40% of its workforce on February 26, 2026. Headcount dropped from over 10,000 to under 6,000. CEO Jack Dorsey attributed the restructuring directly to AI, stating that “intelligent tools” have fundamentally changed what it means to build and run a company.
The stock jumped 22-25% in after-hours trading.
I have no particular interest in Block itself. Companies lay people off every month. But line up Block, Klarna, Salesforce, and Amazon’s moves over the past two years, and a pattern emerges. The pattern is what matters.
The Rule
AI is repricing human capital. For allocators, this is not a tech HR story. It is a regime change in corporate profit margins.
That sentence has three layers of transmission.
Transmission 1: Operating Leverage Regime Change
Software companies spend more than half their revenue on people. The exact ratio varies by business model and maturity, but 50% is a reasonable baseline across the sector. This is well understood. What is less appreciated is how tightly revenue growth has been bound to headcount growth: you wanted 30% more output, you hired 20-30% more engineers, PMs, and support staff, and that near-linear relationship put a structural ceiling on operating leverage that persisted for decades regardless of how sophisticated the tooling became.
AI removed that ceiling.
Block’s numbers tell the story. Full-year 2025 gross profit came in at $10.36B, up 17% year-over-year (source: Block 10-K). The business was growing, not shrinking. This was not a distress layoff. It was a restructuring driven by the realization that one senior engineer paired with AI tools can produce what two or three used to.
Cut 40% of headcount. Revenue holds steady. The personnel savings drop straight to the bottom line.
That is the regime change in operating leverage. Legacy valuation models assume software companies must keep hiring to keep growing, with personnel costs scaling proportionally. That assumption is breaking.
The incentive structure makes this self-reinforcing. Wall Street saw Block cut 40% and rewarded the stock with a 22%+ surge. The signal to every other CEO is unambiguous: the market pays you to do more with fewer people. Block will not be the last.
Context matters. Block had roughly 3,835 employees at the end of 2019. By 2025, that number had ballooned to 10,205 (source: Block 10-K, CNN reporting). Zero-interest-rate-era overexpansion, corrected in one move. But the new equilibrium will not snap back to 10,000. It will settle at 6,000 or lower, while revenue continues to climb.
That is the assumption allocators need to recalibrate.
Transmission 2: Silicon Valley Inflation, Not U.S. Inflation
The intuitive macro read is straightforward: mass layoffs of high-earning tech workers should depress wages, reduce spending, and eventually pull down CPI.
The scale does not support that narrative.
Broadly defined, tech employment accounts for somewhere between 3% and 5% of total U.S. employment, depending on how you draw the boundary. Tens of thousands of displaced engineers will hit San Francisco rents and high-end restaurant revenue. They will barely register on national CPI. This depresses Silicon Valley inflation. It does not move the U.S. number.
The more interesting transmission is talent reallocation. These engineers do not disappear. They flow into healthcare, financial services, manufacturing, and other industries that have been starved for technical talent. But there is a friction cost: traditional industries cannot match the $250-300K total compensation packages that Big Tech offered. The transition involves a pay cut and an adjustment period, not a seamless handoff.
For allocators, the signal is K-shaped wage bifurcation. Most software engineers will see compensation stagnate or decline as supply surges and the large employers freeze hiring. A small cohort with AI infrastructure skills, the people who can fine-tune models and manage large-scale GPU clusters, will see their compensation bid up aggressively in a talent arms race.
That bifurcation is itself an allocation signal.
Transmission 3: The Micro-Elite Firm
The third transmission path is the slowest and the least reversible.
When the marginal cost of producing software approaches zero, a three-person team with AI can ship what used to require thirty. The barrier to launching a product drops to a weekend and a few dozen dollars in monthly AI subscription fees.
The flip side is a competition explosion.
If anyone can build a SaaS product in three days, “the ability to write code” is no longer a moat. The real moats migrate to two things: distribution and proprietary data. You have a massive audience or a precisely targeted customer list, and what competitors cannot replicate is not your product but your reach. You have decades of non-public vertical industry data to fine-tune models, and competitors running the same foundation model cannot match your output.
For allocators, the implication is direct. Software company valuation premiums have been partly built on the assumption of durable technical moats. If AI is flattening technical barriers, the survivors win on distribution and data moats instead. The valuation anchor needs to shift accordingly.
The Overfitting Lens
Your portfolio carries an implicit assumption: that human capital pricing is stable. That the cost structures embedded in the companies you own will persist. Block just demonstrated that assumption is being challenged, and the market is rewarding the challengers.
There is a deeper layer. Elite education and credentialing systems optimize for a specific economic environment, the way a machine learning model optimizes for its training set. When the underlying environment shifts, as AI is shifting it now, the question is whether those finely tuned credentials generalize to the new regime or overfit to the old one. For allocators evaluating management teams and workforce quality, that is not an abstract question. It is a valuation input.
Where This Rule Breaks Down
The boundary conditions matter.
First, this rule applies primarily to industries where personnel costs dominate revenue: software, knowledge services, consulting, media. AI replaces a software engineer’s output far more readily than it replaces a semiconductor fab operator or a logistics worker. For hardware manufacturing, physical supply chains, and capital-intensive industries, the human capital repricing will be slower by an order of magnitude. The semiconductor supply chain is a clear example: the smarter AI gets, the more insatiable its appetite for compute hardware, making IC design and fabrication talent more scarce, not less.
Second, distinguish genuine AI-driven restructuring from “AI washing.” Some companies are using AI as cover for layoffs that are really corrections to zero-rate-era overexpansion. The diagnostic is simple: track revenue trajectory and R&D spending post-layoff. If revenue holds or grows and R&D investment increases, the restructuring is likely structural. If revenue declines in tandem, it is a business problem dressed up in AI language.
Third, transmission speed. This repricing unfolds over quarters and years, not days. It is a slow, largely irreversible regime shift, not a tradable event.
Three Paths, Each with a Cost
The first path is riding the margin expansion directly. Allocate to software companies benefiting from operating leverage gains, particularly those with high personnel cost ratios and demonstrated AI adoption. The cost: if AI productivity gains disappoint, the margin expansion narrative reverses and you absorb a valuation correction. This suits allocators with high conviction in AI productivity and tolerance for volatility.
The second is the picks-and-shovels play. Skip the question of which software company wins and allocate to the AI compute infrastructure supply chain instead. Valuations here are already stretched, and these are capital-intensive businesses. If the AI capex cycle decelerates, the correction could be severe. This suits allocators who prefer the “sell ammunition” logic over picking winners.
The third path is inaction. Maintain current allocations without adjusting for AI-driven human capital repricing. This is a legitimate choice, but understand what it implies: your portfolio carries an embedded assumption that headcount cost structures are stable. If the regime change is real, the market will gradually price it in, and your positioning drifts behind without you ever making a decision. Not choosing is itself a choice with a price.
Audit the human capital assumptions embedded in your portfolio. Then decide for yourself whether to adjust.
Disclaimer
This article reflects my personal investment philosophy. It is not investment advice. Make your own informed decisions.
Miyama Capital manages proprietary capital only and does not solicit external investors.
Kuan & CIO, Miyama Capital