Unlock Editor's Digest for free
FT editor Roula Khalaf has chosen her favorite stories in this weekly newsletter.
It's a truism in the technology industry that each new computing platform opens the door to a whole new generation of software companies. The client-server era that began in the 1990s gave us Oracle and SAP, and cloud computing gave rise to Salesforce and a slew of “software-as-a-service” companies.
Large-scale language models are shaping up to be the next platform that will make the dreams of thousands of entrepreneurs come true. Generative AI is now readily available from companies like OpenAI and Anthropic, and “smart apps” designed to make work easier have proliferated. The speed with which some of these have gained users and their valuations have skyrocketed is setting new records in the software world.
Most notable is the rise of coding assistants such as Cursor. Its owners are reportedly nearing the completion of an investment round valuing it at $10 billion, just three months after raising $2.5 billion in funding.
Coding assistance tools and other AI-powered tools for tech-savvy users are leading the way, but many other startups are tapping into nearly every aspect of white-collar work. These range from tools used to create or edit any form of content or digital media to tools that can handle in-depth research. Tomasz Tunguz, a software investor at Theory Ventures, says this is fueled by the fear on the part of many workers that if they don't learn how to use the tools, they won't be able to quickly acquire the skills expected of their jobs.
Some apps register results surprisingly quickly. Mercor, which uses AI-powered agents to screen and interview job applicants, announced in January that it had reached $50 million in annual recurring revenue less than two years after it was founded. For comparison, it took Salesforce four years to reach $50 million in annual revenue.
Other companies' revenues appear to be exploding even faster. Loveable.dev, a Swedish company that helps non-technical users build websites and more, announced last month that it had reached $17 million in ARR just three months after launching. A similar company, Bolt.new, said it went from zero to $20 million in profit in two months.
As these companies grow rapidly, they face some of the same problems as new software applications as well as previous generations of new software applications.
One challenge is turning an AI-powered tool designed for a single task into a core part of a customer's software. This means automating more aspects of the process in question until agents can digest the entire workflow. In this regard, they are up against software giants such as Microsoft, Salesforce, and Adobe, which have their own AI agents and already have strong connections with many companies.
In the early days of the cloud, startups had built-in advantages over established companies that had technology and business models associated with different delivery methods. But the AI era of software is actually more of an extension of the cloud than an entirely new computing platform, notes Byron Deeter, a veteran software investor at Bessemer Venture Partners. This reduces destructive potential.
Another difference is rapid growth. As a result, the most successful newcomers have come to look more like consumer apps than traditional enterprise software, Dieter says. It is not yet clear whether these will maintain consumer-like characteristics as they mature, which could lead to higher churn rates than typically seen in the business software world, for example.
The financial situation also looks very different. AI app companies face significant costs in the form of fees paid to LLM companies each time their services are used. Many people are choosing to accept these costs for now in the hope that LLM fees will continue to fall. For example, Cursor allows customers to make 500 calls per month for a $20 subscription fee, which may leave little gross profit if they pay for their usage in full.
Futurist and venture investor Peter Diamandis compares the massive investment in LLM to the over-investment in new communications networks during the early years of the Internet in the late 1990s. Just as then, companies building new infrastructure are now forced to cut prices and struggle to turn a profit, he said, paving the way for application makers to profit.
The soaring valuations of high-tech companies in the late 1990s ended with the dot-com bust. While some app makers are at least making a lot of money this time around, there's no guarantee that another bubble won't form as expectations for AI grow.
richard.waters@ft.com
