← Home

Three Headlines in One Week That Tell You the AI Build-Out Just Hit a Cycle Marker

Nvidia beat earnings handily. Meta laid off 8,000. Samsung went on strike. Three headlines, one story: the AI cycle is rewiring real labor and real supply chains, fast.

Three signals landed inside 36 hours this week. Nvidia reported Q1 results that beat consensus on revenue, EPS, and Q2 guidance, with Data Center growth of 92% year over year. Meta began the first round of an 8,000-person layoff, framed as an AI-efficiency drive. Samsung Electronics' largest union started an 18-day strike on May 21 involving more than 45,000 workers, with memory chip production for AI accelerators in the disruption zone.

The three stories landed in different sections of the financial press and got covered as discrete events. They are not discrete events.

I have been around for a few of these moments now. The pattern I am describing has happened before, and it tends to mark the inflection point in a capital cycle rather than the beginning of it.

What is happening: a small number of vendors at the top of the supply chain (Nvidia, TSMC, ASML, the HBM duopoly) are capturing margin at a rate that is rewriting their position in the global economy. A larger number of integrators in the middle (Microsoft, Meta, Google, Amazon) are absorbing those vendors' pricing while restructuring their own workforces to make the math work. The labor at the bottom of the supply chain (Samsung's union members, the contract manufacturing workforces across Vietnam, Malaysia, and Taiwan) is doing what labor has done in every previous cycle. It is using the moment of supplier leverage to renegotiate.

The Nvidia print is the cleanest version of this story. $81.6 billion in Q1 revenue with 85% YoY growth shows the math of a company that has structurally captured a critical economic input. The pricing power that produced $1.87 in EPS against an $80 billion-plus revenue base is what concentration looks like before regulators or competitors arrive.

The Meta layoff is the cleanest version of the integrator story. 8,000 cuts at $350K each is roughly $2.8 billion in annual run-rate savings. The 2026 capex line is $125-145 billion. The cuts fund a margin recovery that lets the company maintain operating-income growth while spending the capex. The next round, if Mark Zuckerberg's prior playbook holds, comes in 9-12 months.

The Samsung strike is the cleanest version of the labor story. HBM memory is the supply-chain pinch point for every Nvidia accelerator that ships. The union timed the action with full awareness of the leverage that pinch point provides. They are right about the leverage.

The reason this matters is that all three landed in the same week. The pattern of these three landing together is what every prior cycle has looked like at this stage. The dotcom build-out had the equivalent moment when Cisco's earnings rolled over while Yahoo was cutting headcount and the dock workers in Long Beach were threatening a strike over container traffic. Different industries, same week, same pattern.

What I would expect to follow this is something like 12 to 18 months of headcount and supply-chain volatility, with the top of the stack continuing to print well. After that, the question of whether the demand side justified the capacity build-out gets asked. The cycle is visibly compounding now. The question of when it stops compounding gets asked later, and it becomes the most important question in the cycle when it does.

A defensible counter-argument: this is the early phase of a permanent productivity reshaping, and the pattern-recognition I am applying to the dotcom era does not transfer. AI is a different kind of input than fiber-optic capacity, and the demand for it is structurally different from the speculative demand that funded the dotcom build-out. Under that view, Nvidia's $81B run-rate is the leading edge of a much larger durable revenue base. The peak is further out. Meta's layoffs are routine rebalancing after pandemic-era overhiring. Samsung's strike is a routine labor action with limited industry-level implication. All three readings are coherent. I take the prior-cycle-pattern view as the higher-probability default, but I take the case for permanence seriously enough to avoid betting against it heavily.

  • See Nvidia's SEC 8-K filing for the official Q1 numbers
  • Meta's full layoff disclosure is expected in next week's 10-Q amendment
  • The Samsung union's stated demands center on wage parity with semiconductor competitors in Taiwan, a multi-year structural issue

For operators inside any company with substantial AI-related capex commitments: model both a capacity-extension scenario and a capacity-overshoot scenario, with separate hurdle rates. For investors: the dispersion between top-of-stack vendors and integrators is going to widen before it narrows, and that is the trade. For workers in supplier-tier supply chains: this is the moment of leverage. Use it now.

— Hank