AI-native accounting startup Rillet raised a $100M Series C led by Iconiq at a $1B valuation, just two years after coming out of stealth. Backed by a16z and Sequoia, the round closed in under 48 hours and follows a doubled ARR in three months, an EY alliance, and 600+ customers. A signal of how fast AI can displace legacy ERP incumbents like NetSuite.
Silicon Data closed a $30M Series A to become the reference price for GPU compute, with an index that Wall Street futures contracts could settle against. It plans to launch compute futures trading on the CME October 5th, pending approval, giving the AI buildout a way to hedge its single biggest cost.
Sachin Bansal's eight-year-old, self-funded fintech Navi has taken its first institutional capital: a $100M Prosus investment valuing it near $1.3B. The round lands as Navi preps a ~$314M IPO, reports ~$323M FY26 revenue, and runs India's fourth-largest UPI app.
Still raw over being ousted from Uber, Kalanick says just 1% of VCs meet the bar of being "helpful," yet he just took $1.7B from Andreessen Horowitz for his robotics company Atoms (Ben Horowitz joins the board). His advice: hone a pitch that starts a bidding war, and don't fall into a "victim mentality."
An audit of how 50 top SaaS companies position against the foundation models their own buyers ask about found 34 shipped MCP servers or connectors so models route through the product instead of around it. Yet most vendors still ignore the build-versus-buy threat entirely.
RevenueCat crunched 3,519 AI subscription apps and found they churn 30% harder than non-AI, yet the best retain 13.9% of paid subs a year in versus 1.4% at the bottom. The gap opens at first renewal. Subscription-only, 7-day-trial, freemium apps with modest pricing retain best.
Dan Hock, Faire's outgoing Chief Strategy Officer, distills nine years of 100x team growth into 9 operating lessons: speed over polished strategy, constant application of force, and ruthless owner focus. A rare inside look at how a wholesale marketplace scaled from a whiteboard sketch to 20M products.
Shopify locking down its checkout killed Ruslan Leteyski's $8M ARR business in a day. He rebuilt with Zipchat, an AI sales agent for e-commerce, now handling 700k conversations a month and nearing $2M ARR. The cautionary takeaway: don't build a business on a platform that can change its mind.
A solo builder spent two years letting AI agents run product development, 128 npm/Homebrew releases, then a second 442-test product, and ended with zero external users. His frank post-mortem names the failure: guardrails written as polite prose that agents ignored, and polish substituting for distribution. A sharp warning for agent-era founders.
Stripe paid $7.5 billion for OpenRouter, a jump from its $1.3 billion May valuation. The founders' "singularity" line is tongue-in-cheek; the real play is embedding Stripe into both sides of AI spend, adding model-expense management to its cash-collection business.
The era of the single AI-powered "full-stack builder" is a trap. As building gets cheaper, the advantage moves to teams that tune output tightly to customer reality, not generalists. The handoffs between product, design, and engineering were checks and balances that AI-native workflows now quietly sacrifice.
Rippling's President & CPO published a real production test of 15 AI models doing scored payroll agent work, 2,100 graded runs each, with accuracy and cost recorded. The cheapest model, GLM 5.2, tied the most expensive on accuracy at half the price, while quality flattened at the top and cost diverged sharply.
15 growth-stage VCs explain how AI-era reporting warped once-standard SaaS metrics: ARR now usually means annualized run-rate, contracted ARR mixes signed bookings with unproven capacity, and NRR gets inflated by flipping small pilots into contracts. A sharp look at why headline numbers no longer mean what they say.
You’re caught up.
That was all 13 stories in the 20 August issue.