The AI Backer That Made $13 Billion in a Year Without Training a Single Model

Deep thoughts on AI and aspirations —— ByteDeep Thought Circle

Let’s start with some numbers. Tether made $13 billion in 2024 with 150 employees. That same year, OpenAI had $3.7 billion in revenue, lost $5 billion, and employed over 3,000 people; Anthropic had $1 billion in revenue and also lost $5 billion. The combined losses of these two serious AI labs don’t even reach half of this stablecoin company’s annual profit. In terms of per-capita output, the gap is 60-fold.

This company then started pouring money into AI, investing nearly $1 billion over two years. Its position is delicate: one of the wealthiest backers in the entire AI industry, yet it doesn’t train a single line of models itself.

The Money Printer’s Mechanics: An Interest-Free Deposit

Tether’s business model is remarkably straightforward when broken down. You exchange $1 for 1 USDT, it takes your dollars to buy U.S. Treasury bonds, keeps the bond interest, and that’s none of your concern. Banks doing the same thing have to pay depositors interest; it doesn’t. With 150 people managing over $130 billion in Treasury bonds, interest alone brought in about $7 billion in 2024.

This structure isn’t novel in financial history—it’s float. Insurance premiums, in-transit credit card funds—all essentially “collect money first, deliver service later, use funds in between for free.” Float businesses share common traits: they don’t require innovation, just scale and avoiding disasters. Originally, this had zero connection to AI.

But float has an Achilles’ heel: returns follow interest rates. When the Fed cuts rates, Treasury yields drop, and this money printer slows down. Making $7 billion effortlessly in 2024 won’t necessarily continue. The money’s still there, but the growth story becomes harder to tell. This is its first motivation for pivoting to AI.

The second motivation is more straightforward: narrative positioning. Call yourself a stablecoin company and investors, media, and politicians barely glance your way; call yourself an AI, brain-computer interface, humanoid robotics player, and you’re a tech leader. Even money printers need new stories, and AI is the only story today with enough weight.

How It’s Placing Its Bets

Looking at the portfolio reveals not scattershot investing but a vertical chain.

One move deserves special mention: open-sourcing the dataset. While OpenAI and Anthropic’s training data are core secrets, Tether does the opposite, releasing what it claims is the world’s largest open-source training dataset. This isn’t charity—it’s trading zero marginal cost for ecosystem positioning. By occupying the “open” persona in the data layer while adding to its own narrative.

Funding Frontier Tech with Cash Flow: An Old Recipe

Seeing “a company unrelated to AI using profits to invest in AI,” don’t rush to treat it as novelty news. This structure has appeared repeatedly in tech history: Bell Labs, funded by telephone monopoly profits for decades, produced a host of inventions that shouldn’t have belonged to any single company; Xerox PARC, sustained by copier profits, invented the GUI and Ethernet but made no money from them. A more recent example: Tencent using gaming profits to sustain research labs and numerous unprofitable foundational products for years.

This leads to my assessment: to evaluate who can sustain bets on long-cycle technologies, look not at funding amounts but at cash flow structure. Whether there’s a cash inflow unrelated to R&D cycles with near-zero marginal cost determines a company’s patience ceiling.

Funding has cycles—when markets cool, you pitch stories and cut projects; float has no cycle, only interest rates. This is why the most composed bettors in the AI race are those holding money printers: OpenAI tells investors not to expect profitability until around 2029, Anthropic around 2028, both still pitching to capital; Tether doesn’t need to pitch—the money’s already in pocket. Betting right is foresight, betting wrong is tuition, either way it doesn’t hurt the core business.

The actionable insight for entrepreneurs and investors: when examining a company’s AI investment, first ask where the money comes from. Profit-funded, fundraising-funded, debt-funded—these are three entirely different species with vastly different patience levels and degrees of distortion under pressure. Teams burning fundraised capital must re-convince investors every 18 months; their AI strategy is inherently shortsighted—not a team problem but a capital structure mechanics problem.

Two Types of Money, and Their Respective Ways of Dying

Money for doing AI and money for investing in AI are two different things. The former demands closed-loop business models, hence loss pressure, fundraising rhythms, valuation games; the latter only requires principal preservation, so it can buy compute, buy data, buy options that won’t pay off for a decade. Tether stands in the latter camp; its composure comes from structure, not superior vision.

But even the best structure has boundaries. This money printer carries at least three risks. First, in ten years it’s never undergone complete audit; reserve transparency relies on self-disclosure—a soft spot regulators could close anytime. Second, falling rates erode precisely its core returns. Third, its nearly $1 billion AI investment is merely a fraction of its $13 billion annual profit; “wrong bets are just tuition” holds only while the printer keeps running. Should stablecoin regulation land and reserve rules tighten, the tuition budget would immediately recalculate.

Finally, the unavoidable irony: the most centralized company in crypto—it controls issuance, self-reports reserves, hasn’t audited in ten years—now champions “decentralized AI, local operation, returning intelligence to individuals,” rather like a casino owner teaching gambling addiction recovery. But irony doesn’t constitute negation. For funded technology assets, money has no smell; for observers, don’t learn its slogans, learn its structure.

What’s truly worth taking away is this sequence: first have a cash flow machine unrelated to hot trends, then use the machine’s spare change to buy the most expensive options. Reverse the sequence—do AI first then find a business model—and you get today’s top labs losing $10 billion annually. AI’s real business model challenge hasn’t vanished; it’s just been reframed by a company in no hurry: it hasn’t solved “how to profit from doing AI,” it’s solved “how to make AI losses not hurt.” These are two questions, and everyone entering the space should be clear which one they’re solving.