Fifteen months ago, Anthropic was a research lab doing about $1B in annualized revenue.

Today, it's doing $30B.

OpenAI is at $25B.

Anthropic didn't just catch up. It passed them.

The gap is $5B. The story is much bigger than that.

There is no comparable curve in B2B software history. Salesforce took 20 years to reach $30B in annual revenue. Microsoft Azure took 12. Anthropic did it in 15 months.

Here's the part most people miss:

OpenAI has 900M weekly users, the largest consumer software base ever assembled. Anthropic has a fraction of that.

And still makes more money.

This isn't about models. It's about business models.

OpenAI is primarily consumer:

Anthropic is primarily enterprise:

Two companies. Same technology. Completely different revenue engines.

OpenAI's mix is roughly 60% consumer, 40% enterprise. Anthropic's is the inverse: 80% enterprise, 20% consumer.

Anthropic now has 1,000+ customers spending over $1M per year.

That number was 500 in February. It doubled in under two months.

Eight of the Fortune 10 are customers.

Once these systems get embedded into workflows, they don't get ripped out. They expand.

That's not usage growth. That's revenue compounding.

The market is already pricing the difference.

Anthropic raised at $380B in February. Unsolicited offers above $800B since. An IPO is targeted for October.

OpenAI's last private round priced it at $500B. The numbers will move.

The question being asked in board rooms right now is which of these companies is actually the more valuable business.

The answer is no longer obvious.

That's what Alphabet, Amazon, Microsoft, and Meta will spend on infrastructure in 2026.

Up 77% in a year. Bigger than Sweden's GDP. Bigger than the entire Apollo program in inflation-adjusted dollars, spent in a single year.

Each of these four companies is now individually exceeding $125B in capex. Microsoft's CFO told investors $25B of the increase came from rising memory chip costs alone.

This is what belief looks like when it hits the balance sheet.

The bottleneck isn't chips.

It's power.

Microsoft is about to spend $190B on infrastructure in 2026. Amy Hood, the CFO, told investors something striking: it won't be enough. Microsoft will remain capacity-constrained through 2026 even at that spend.

Roughly 11 gigawatts of announced data center capacity for 2026 is currently stalled at the announcement stage, waiting for grid interconnection. Some projects have been waiting since 2023. PJM, the grid operator covering the largest data center cluster in the U.S., has a multi-year backlog.

You can buy GPUs in months. You can't build a grid that fast.

Power is now the limiting factor. The companies that have it are sitting on the most valuable real estate in the world.

Robotics. Figure raised $1.5B at a $39.5B valuation last week, up from $2.6B in 2024. A 15x markup in 18 months for a company that has shipped fewer humanoid robots than Tesla has Cybertrucks. The valuation isn't about today's deployments. It's a bet that BMW's pilot becomes the auto industry's standard.

Global. The UAE committed $50B to AI infrastructure across G42 and MGX in Q1 2026. Saudi Arabia's HUMAIN announced a $40B compute fund. Sovereign wealth is now a tier-one buyer of GPUs, sitting alongside the hyperscalers in NVIDIA's allocation queue.

US Economy. Data center electricity demand will hit roughly 1,100 TWh in 2026, per the IEA. About 4% of all U.S. electricity consumption. By 2028, the EIA projects data centers alone will consume more electricity than the entire U.S. manufacturing sector.

Markets. 4.2% of all U.S. job postings now mention AI tools or skills, per Indeed's January Hiring Lab data. AI-mentioning postings are 45% above pre-pandemic baseline. Total tech postings are 34% below. AI is creating jobs while the rest of tech is shedding them.

Should Know. NVIDIA reported $62.3B in Q1 data center revenue, up 73% year-over-year. That single business unit now generates more quarterly revenue than the entire 2025 GDP of Iceland.

This isn't a cycle. It's a replatforming.

A recurring snapshot of the numbers driving the AI economy.

You're reading the first issue of Machines & Money.

A weekly newsletter about what happens when AI meets capital. No hype. No demos. Just the economics.

The AI economy is the largest capital reallocation of our lifetimes. Hundreds of billions are being deployed. Trillions will follow.

Most coverage is either too technical to act on, too political to trust, or too optimistic to believe.

Machines & Money focuses on one thing: what the numbers actually mean.

Every Sunday, we'll cover the AI economy: who's making money, who's burning it, and what it means for operators, investors, and anyone building in this market.

See you next time.

- Machines & Money

Like this? Forward it to someone who's going to need it. They'll thank you.

Sources: Anthropic public disclosures and Series G filings, OpenAI Q1 2026 revenue reporting, Counterpoint Research Q1 2026 LLM tracker, FT Q1 2026 hyperscaler earnings analysis, Microsoft Q3 FY26 earnings call, NVIDIA Q1 FY26 earnings, IEA Energy and AI Report April 2026, EIA Annual Energy Outlook 2026, Indeed Hiring Lab January 2026 update, Figure Series F press, G42/MGX/HUMAIN public announcements.