It seems every five minutes, another AI company announces a funding round with a valuation that would make your eyes water. Just when you think the numbers cannot get any more detached from reality, a new story breaks that pushes the boundaries of belief. The latest protagonist in this high-stakes drama? An infrastructure startup called Modal Labs.
According to a recent report from TechCrunch, Modal is in talks to raise a fresh round of capital that would value the company at a staggering $2.5 billion. What makes this particularly noteworthy is that its last funding round, just five months ago, pegged its valuation at a mere $1.1 billion. That’s more than a doubling in value in less time than it takes to get a new sofa delivered. All of this for a company with a reported annualised revenue run rate of around $50 million. So, what on earth is going on? To understand this, we need to look under the bonnet of the AI industry, at the engine that actually makes it run: inference.
The Not-So-Simple Act of Thinking
What Exactly is AI Inference?
Let’s get one thing straight. Building a massive AI model—what we call ‘training’—is the glamorous part. It’s like writing an encyclopaedia from scratch. It requires monumental amounts of data and unthinkable processing power, costing hundreds of millions of pounds. But once that encyclopaedia is written, its real value comes from people asking it questions and getting quick, intelligent answers.
That second part is AI inference. It’s the ‘live’ operational phase where a trained model uses its knowledge to make predictions, generate text, or create images. If training a model is like building a Formula 1 car, inference is the act of racing it, lap after lap. It has to be fast, reliable, and, critically, not bankrupt the team with every race. This is where the real challenge—and the huge business opportunity—now lies.
Why Compute Optimisation is the Name of the Game
Running inference at scale is ferociously expensive. Every time you ask a chatbot a question, it consumes costly processing power on specialised chips. For companies deploying AI-powered features to millions of users, these costs can spiral out of control. This has ignited a gold rush in the AI inference market for anyone who can make this process cheaper and more efficient.
This is where compute optimization comes in. It’s the art and science of squeezing every last drop of performance out of the hardware. Companies that master this are not just selling a service; they are selling savings. They are the efficiency consultants for the entire AI economy, and their services are in very high demand.
Modal Labs: The $2.5 Billion Bet
A Founder Who Knows Scale
Modal Labs was founded in 2021 by Erik Bernhardsson, a name that carries weight in engineering circles. As the former CTO of Better.com and a distinguished engineer at Spotify, Bernhardsson has a deep background in building and scaling complex, high-demand software infrastructure. This isn’t his first rodeo.
His new venture, Modal, focuses squarely on this model serving problem. It provides a serverless platform that allows developers to run their AI models without having to worry about managing the complicated and expensive infrastructure underneath. The goal is simple: make running AI models as easy and cost-effective as possible.
The Dizzying Valuation Leap
The news that venture capital giant General Catalyst is reportedly in talks to lead a new round at a $2.5 billion startup valuation has certainly turned heads. While Modal’s CEO has publicly stated the company isn’t actively fundraising, the conversations are clearly happening. This potential investment would more than double the company’s value from its $87 million Series B round a few months ago, a sign of just how heated the competition is among investors to back a winner in this space.
A Crowded and Expensive Playing Field
Modal is by no means alone. The AI inference market is becoming a battleground for well-funded startups, each promising to solve the efficiency puzzle. The valuations are, frankly, mind-boggling across the board.
– Baseten: Recently rumoured to be raising $300 million at a $5 billion valuation.
– Fireworks AI: Reportedly securing $250 million at a $4 billion valuation.
– Inferact: A newer player that raised a $150 million seed round at an $800 million valuation.
What we are witnessing is a classic “picks and shovels” investment strategy. During a gold rush, the people who consistently make money aren’t always the prospectors, but the ones selling them the tools to dig. In the AI boom, the foundation models are the gold, and companies like Modal, Baseten, and Fireworks AI are selling the high-tech shovels.
What Does This Valuation Madness Mean?
Are These Startup Valuations Sustainable?
The rapid inflation of these startup valuations raises an obvious question: is this a bubble? The numbers seem disconnected from current revenues. A $2.5 billion valuation on $50 million of revenue is a 50x multiple, a figure that would be considered aggressive even for a top-tier public software company.
However, investors are not valuing these companies based on today’s earnings. They are making a strategic bet on the future composition of the entire tech industry. The belief is that AI is not just a feature but the next fundamental computing platform, akin to the internet or the mobile phone. In that scenario, the companies that provide the essential infrastructure—the “operating system” for AI—stand to capture an immense amount of value. VCs are willing to pay almost any price for a shot at owning a piece of that foundational layer.
The Future of AI Infrastructure
This intense investment will accelerate innovation. The competition to provide the fastest and cheapest model serving will push the entire industry forward, ultimately benefiting developers and consumers by making AI applications more powerful and accessible.
However, it also raises the stakes immensely. With billions of dollars pouring in, the pressure to grow and capture market share will be enormous. We can expect aggressive market positioning, talent wars, and eventually, consolidation. Not all of these high-flying startups will survive. The winners will be those who can not only build superior technology but also execute a flawless go-to-market strategy in a brutally competitive environment.
The race is on, not just to build the smartest AI, but to build the most efficient engine to run it. Modal’s potential new funding is just the latest signal that in the world of AI, the plumbing is rapidly becoming more valuable than the water.
What do you think? Are these valuations a sign of a frothy, overheated market, or a rational bet on the future of technology? Let me know your thoughts in the comments below.
