
This startup raised $15M to break Nvidia’s tightest grip on AI, its software
THE SO WHAT
A $15 million seed at a $100 million valuation to run AI inference on “any chip” shows how much appetite there is to loosen CUDA’s hold on the stack. If you’re planning multi-year AI infra, keep a close eye on these abstraction layers — they’re your best shot at future-proofing against single-vendor lock-in.
READ THE SOURCE
MORE FROM THE WIRE
Startups & VentureSource: Stripe revenue jumped by a third to $6.8B in 2025, its fastest revenue growth since 2021, with free cash flow up 52% to $3.2B
Stripe growing revenue ~33% to $6.8B and free cash flow 52% to $3.2B — driven partly by AI customers — means the AI boom is already reshaping payments economics. If you’re building infra around AI-native businesses, assume they can support premium take rates today but will demand volume discounts and deeper integration quickly.
Startups & VentureTravis Kalanick’s robotics company raises $1.7B, led by a16z
$1.7B into Atoms is a bet that “industrial AI” plus robotics can unlock a new logistics or fulfillment layer at massive scale. If you run physical operations, assume a step-change in automation options in the next 24–36 months and start mapping which workflows you’d hand to a well-capitalized new entrant.
Startups & VentureYope raises $12.3M to build a private social network without algorithms or ads
A $12.3M seed into an ad-free, non-algorithmic social app is a bet that messaging-first, small-group networks can peel off time from the big feeds. If your product relies on paid social distribution, budget for a world where more of your users’ attention is locked inside private, non-monetized spaces.
Startups & VentureSources: Cathedral, launched by ex-DOGE staffers to use AI to expand US military cyber capabilities, raised $160M led by a16z and Sequoia at a $1.4B valuation
A $160M round at a $1.4B valuation for an AI-first military cyber startup says defense and intelligence are becoming primary AI buyers, not edge cases. If you build security, infra, or model tooling, assume dual-use requirements and classified-environment constraints will shape your roadmap sooner than you think.