A growing group of AI companies is building substantial businesses by running models efficiently. From Fireworks and Modal to model developers expanding into inference, the opportunity is to turn increasingly accessible intelligence into reliable services. What makes these businesses defensible, and where will the value accumulate?
Neoclouds are becoming a critical part of AI infrastructure, but can their businesses outlast their GPUs? From rented data-centre space to managed inference, we explore the financing, hardware economics and software ambitions shaping the category, including Nebius’s connection to Israel.
AI capabilities are advancing faster than businesses can adopt them. For startups, the opportunity lies in turning that potential into dependable products, repeatable workflows and measurable results.
Choosing the right AI model matters. But when competitors can access the same technology, a repeatable way to reach customers and earn their trust can become the more durable advantage.
A new generation of frontier AI labs has raised approximately $11.6 billion across ten mega-rounds. Some have released usable models, products and APIs. Others remain bets on elite teams and unproven research theses. Here is what each neo-lab has actually shipped.
Jev does not generate text. It turns unstructured state into typed, probabilistic decisions in milliseconds and at a fraction of the cost of frontier LLMs. Here is why TypeSafe’s first “System One” model could become a new decision layer for AI agents, automation and data analysis.
Making AI more capable does not automatically make it more trustworthy. As investment and geopolitical competition accelerate development, can safety keep pace? From alignment challenges to proposals for a FINRA-style watchdog, this post explores who should hold AI accountable and how startups can help deliver its benefits while reducing the risks.
Stripe’s roughly $8 billion agreement to acquire OpenRouter connects model routing with billing and payments. Its shareholder letter reveals a broader ambition to serve an economy of AI agents, with implications for startup margins, pricing, distribution and platform dependence.
Not every AI task needs the largest or most expensive model. Model routing promises to match each request with the right model based on accuracy, cost, speed and privacy, but it also introduces a new control point with its own security risks, commercial conflicts and potential for lock-in.
After years of limited venture activity, Israeli EdTech may be entering a new chapter. The founders of Moovit and Outbrain are joining a growing group of entrepreneurs using AI to reimagine tutoring, personalised learning, reading and what students can create.
Weekly Firgun Newsletter
Join founders and investors reading the weekly Firgun newsletter by Eze Vidra.
