Most tech companies spent 2025 trimming headcount. Meta, Google, Amazon — all went through layoff waves. The industry settled on a quiet consensus: do more with fewer people, let AI handle the rest.

OpenAI read the room and did the opposite.

According to the Financial Times on March 25, OpenAI plans to nearly double its workforce — from 4,500 to 8,000 by end of 2026. That is roughly 12 new hires per day for the rest of the year. New roles span product development, engineering, research, and sales. One title stands out: "technical ambassadors" — enterprise sales specialists who help large companies deploy OpenAI's tools. That is Salesforce's playbook, repackaged in a hoodie.

The fuel behind this: OpenAI closed the largest private funding round in history on March 19. $110 billion at an $840 billion post-money valuation — the price investors agree the company is worth after the investment lands. Amazon led with $50 billion, SoftBank and NVIDIA contributed $30 billion each. SoftBank financed its share by selling NVIDIA stock to invest in a company that pays NVIDIA for compute. Venture capital logic at its finest.

Part of that capital feeds Stargate — a joint venture with SoftBank, updated on March 22, to build $500 billion in AI data center infrastructure over four years. Half a trillion dollars in server farms. That is not a product bet. That is an infrastructure empire.

Why the rush?

One name: Anthropic. A Ramp spending report published March 17 found that Anthropic captures 73% of first-time enterprise AI budgets — meaning when a company buys AI tools for the first time, it picks Claude, not GPT. First enterprise contracts tend to stick. Companies rarely switch after deep integration. OpenAI is not just losing deals. It is losing lock-in — the kind that compounds for years.

The math worth sitting with ⚙️

8,000 employees at an average total compensation of $400K — standard for San Francisco AI talent — puts annual payroll at $3.2 billion. The company generates an estimated $5–6 billion in ARR (annual recurring revenue — total yearly subscription income). Revenue grows, but so does the burn rate — the speed at which a company spends cash beyond what it earns.

Stargate is the piece I watch most. OpenAI and SoftBank building the AI equivalent of a transcontinental railroad. If it works, they control the infrastructure layer of the entire AI economy. If it stalls, it becomes the most expensive construction project nobody needed.

The quiet contradiction

OpenAI is an AI company that believes AI will automate most knowledge work. It is also hiring 3,500 humans to do knowledge work. At some point, these two stories collide. If your own product cannot replace the roles you are filling, what does that say about the product? If it can — why are you hiring?

Every problem is a process problem. Throwing 12 bodies per day at a market share gap is not a system. It is a reaction. The outcome depends on the processes they build around those people — not the headcount itself. 🧘

The calmest move in a hiring war is knowing exactly which 500 people you actually need. Everything else is noise.