Analysis
Runable has raised a $21 million Series A co-led by Susquehanna Venture Capital and Nexus Venture Partners, with Together Fund and Array VC participating, at a $65 million post-money valuation, TechCrunch reported. Founded in 2025 in Bengaluru by CEO Umesh Kumar and Saksham Sarda, the company crossed a $2 million annualized run rate within three weeks of launching payments in March.
The positioning is a direct swipe at the code-generation category. "In the end, a business doesn't require Codex or Claude Code or anything. They require real outcomes," Kumar told TechCrunch. Runable's pitch is that the agent should find customers, not just ship the product -- moving from building a business to growing one.
The metrics worth noting
Runable reports 1.7 million registered users, with the United States, United Kingdom and Japan as its biggest markets. The disclosure that matters more is this: 60% to 70% of the more than one trillion tokens it consumed over the past 90 days came from paying customers. Almost nobody in the agent category publishes that ratio, and it is the single most useful number for judging whether an AI product has real demand or a free-tier bonfire. Most consumer AI apps are inverted -- the overwhelming majority of inference spend goes to users who will never pay.
The competitive set
Runable names Cursor, Lovable, Replit, Manus and Genspark as reference points, which spans two distinct markets. Cursor and Replit sell to developers and are priced accordingly -- Cursor reportedly cleared $500 million in annualized revenue in 2025. Lovable, out of Stockholm, went from launch to roughly $100 million ARR in under a year on non-technical builders. Manus and Genspark are general agent products. Runable's claim is that the build layer commoditizes and the growth layer -- customer acquisition, outreach, operations -- does not.
The context on price
A $65 million post-money on $2 million ARR is roughly 32x, which in 2026 is well inside the normal band for an AI company growing this fast; comparable agent companies are clearing 50x to 100x. The modest valuation relative to peers likely reflects geography as much as fundamentals. That gap is itself the opportunity for LPs paying attention to Indian venture: the same metrics in San Francisco would have cleared at three times the price.
The number to check in six months is whether that $2 million run rate compounds or plateaus once the launch cohort cycles.
Where the risk sits
Runable's gross margin depends entirely on token economics it does not control. A trillion tokens in 90 days at frontier-model pricing is a meaningful cost base, and the company's margin improves or collapses based on decisions made in San Francisco. That is the structural fragility of every agent company built on hosted models, and it cuts both ways -- inference prices have fallen consistently, but a single pricing change from a primary provider can move a startup's gross margin ten points overnight. The counter is that open-weight models keep improving; running Qwen or DeepSeek weights on rented hardware is now a credible fallback for a large share of agent workloads, which is precisely why Nvidia is willing to pay for the repository where those weights live.
