Andromeda Hits $1.5B Valuation With $100M+ Revenue

Paradigm just wrote a $60 million check to value an AI compute startup at $1.5 billion. That's not the interesting part. The interesting part is that Andromeda — the company receiving the money — actually generates revenue. Over $100 million in annual run rate, to be precise. In an AI ecosystem drowning in pre-revenue unicorns, that distinction matters enormously.

From Side Project to Billion-Dollar Business

Andromeda didn't start as a venture-backed startup. It started as a service spun up by Nat Friedman and Daniel Gross — two of the most prolific AI investors in Silicon Valley — through their investment vehicle NFDG. The idea was straightforward: help AI companies get access to GPUs without the operational nightmare of managing them directly.

That side project is now its own company. It manages tens of thousands of GPUs, operates profitably, and plans to scale to 100,000 GPUs this year. It doubled revenue from over $50 million in 2024 to over $100 million in 2025. And it just opened a San Francisco office to widen its customer base.

The $60 million funding round from Paradigm — equivalent to a Series A — values Andromeda at $1.5 billion. That's a 15x revenue multiple on a fast-growing, profitable infrastructure business. In 2026's AI market, that actually looks reasonable.

Why Paradigm Is Here

Let's address the obvious question: what is a crypto-native venture firm doing leading an AI compute deal?

Paradigm has been telegraphing this move for months. In February 2026, the firm was reportedly raising a $1.5 billion fund targeting AI, robotics, and other frontier technologies — a significant expansion beyond its crypto roots. Co-founder Matt Huang has been vocal about the overlap between AI and crypto ecosystems, pointing to applications like autonomous agentic payments that sit at the intersection of both worlds.

The firm also partnered with OpenAI in early March to launch EVMbench, a tool for testing machine-learning models' ability to identify flaws in smart contracts. Paradigm isn't dabbling in AI. It's building a thesis.

And that thesis starts with infrastructure. Huang called compute "a massive new market requiring substantial infrastructure." He's right. But the deeper insight is that compute infrastructure looks a lot like the kind of market-making and exchange-building that crypto VCs understand intuitively.

The Commodities Market for GPUs

Andromeda's stated ambition is to create a commodity market for GPUs — analogous to futures markets for oil or wheat. This is the detail that explains everything about why Paradigm is the right investor here.

Think about it:

• GPUs are a scarce, fungible resource with volatile supply-demand dynamics
• AI companies need predictable access to compute the way manufacturers need predictable access to raw materials
• Intermediaries who can aggregate, allocate, and manage that resource capture enormous value
• The market needs price discovery, liquidity, and standardized access — all things Paradigm understands from building crypto market infrastructure

Andromeda isn't just renting GPUs. It's positioning itself as the clearinghouse for AI compute. That's a fundamentally different business than a cloud provider, and it's one that could scale to be much larger than its current $100 million run rate suggests.

The Infrastructure Layer Is Where the Money Is

There's a narrative in AI that all the value will accrue to the model layer — to the OpenAIs and Anthropics of the world. The Andromeda story suggests something different.

Consider the current landscape:

• AI startups raised $258.7 billion in 2025, according to OECD data
• Every one of those companies needs compute — lots of it
• NVIDIA can't build chips fast enough, and hyperscalers can't deploy data centers fast enough
• The gap between compute demand and compute supply is widening, not narrowing

Andromeda sits in that gap. It doesn't train models. It doesn't build applications. It makes sure the companies that do those things can actually get the GPUs they need, when they need them, at a price they can stomach. That's an unsexy business that prints money.

The most valuable companies in a gold rush sell picks and shovels. The most valuable companies in an AI rush sell GPU access and manage compute clusters.

Andromeda's profitability from inception is telling. This isn't a business burning cash to acquire customers with below-market pricing. The demand is so acute that customers are willing to pay real margins for reliable compute access. That's a structural advantage, not a temporary one.

The Stealth Factor

One detail from the Upstarts Media report stands out: Andromeda operates with unusual discretion because of the strategic nature of its clients' work. AI labs don't want competitors knowing how much compute they're procuring, when they're scaling up training runs, or what capacity they have reserved.

This creates a moat that's hard to replicate. Andromeda isn't just a vendor — it's a trusted intermediary handling sensitive competitive intelligence. Once you're embedded in that role, switching costs are enormous. No AI lab wants to risk operational disruption or information leakage by moving to a new compute broker.

What This Means

The Andromeda deal signals three things:

First, AI infrastructure is producing real, profitable businesses at scale — not just hype. A $100M+ revenue run rate with profitability is the kind of traction most AI startups can only dream of.

Second, the smart money is moving down the stack. Paradigm's bet isn't on any particular model or application. It's on the physical reality that all AI depends on compute, and someone needs to manage the market for it.

Third, the crypto-to-AI pipeline is real. Paradigm brings expertise in building market infrastructure, managing liquidity, and creating standardized access to scarce digital resources. Those skills transfer directly to the GPU compute market.

Andromeda may not have the name recognition of OpenAI or the hype of the latest reasoning model. But it has something most AI companies don't: a profitable business solving a problem that's getting worse, not better. At $1.5 billion, Paradigm is betting this is still early innings.

Given the trajectory of AI compute demand, that's probably the right bet.