Conviction is an AI-only venture firm founded in 2022 by Sarah Guo, backing the AI-native 'Software 3.0' era. It writes US$1–25M checks from pre-seed to Series A and is often the first institutional investor. Its portfolio spans Mistral, Harvey, Sierra, Cognition, HeyGen, Baseten and OpenEvidence.
Conviction is purpose-built to serve AI-native, 'Software 3.0' companies, believing we are extremely early in the translation of powerful AI models into powerful products that transform industries. It invests only where the product itself is built on frontier AI rather than in AI bolted onto traditional software.
Conviction partners with what it calls the most important companies of tomorrow with US$1-25M checks, likes to invest early and is often the first investor a founder takes. It deliberately keeps a small, concentrated portfolio so it can be hands-on and serve founders deeply rather than spread thinly.
The firm says it has a wide and deep network it puts to work for portfolio companies, understands the domain, and gravitates toward technical teams that build with velocity and pragmatism. Programs such as Embed grants, the Commit fellowship, Mixture of Experts recruiting events, a portfolio jobs board, and the No Priors and Good Company podcasts feed that ecosystem.
Conviction was founded in 2022 by Sarah Guo after her time as a Greylock general partner. Fund I raised US$101M in 2022; Fund II closed at US$230M in January 2025, when former Sequoia partner Mike Vernal joined as General Partner.
Companies listed on conviction.com's Portfolio section (homepage, Sep 2026). The firm says it backs engineers, researchers, product folks, domain experts and dropouts, gravitating toward technical teams that build with velocity and pragmatism - plus others still in stealth.
Serve and scale open-source and custom AI models on a fast, reliable inference platform.
Real-time voice and audio AI — text-to-speech via Sonic-3.6 with models that learn and interact like humans.
Operates Devin, the first autonomous software engineer — planning, writing, testing and shipping production code.
Secure AI coding at the source — prevents security flaws before they are even written.
Real-time insights from the radio spectrum using small, cheap and efficient sensors.
AI software for legal and professional services — contract analysis, due diligence, compliance and litigation.
AI video platform with lifelike avatars, text-to-video, video translation and AI studio tools.
A personal assistant that understands your work and takes action across your applications and devices.
Independent web search infrastructure for AI — its own crawler, index and ranking that improves with agent usage.
Enterprise pharmacy intelligence for leading health systems, powered by a clinical agentic engine.
Trusted AI research for leading brands — customer research conducted, analyzed and summarized by AI.
European frontier AI lab offering LLMs, assistants, agents and an enterprise platform.
The AI omnicloud — orchestrates multi-cloud GPUs for training and inference.
Secure AI control plane that oversees how AI agents are used, keeps them secure and compliant, and measures effectiveness.
AI medical knowledge platform, free for verified U.S. clinicians, with official content partnerships incl. NEJM and JAMA.
Autonomous offensive security solution that finds exploitable vulnerabilities before a breach.
Conversational AI agents that help businesses build more human customer experiences.
Agentic process automation — automates data entry, scraping and processing flows using LLMs and computer vision.
Sunday Robotics — helpful home robots; Memo works 24/7 to lighten household chores (launching 2026).
No rounds led by Conviction are in the tracker yet. Deals will appear here as they are recorded.
Concentration is the service model, not a constraint. A US$230M Fund II spread across 19 names implies roughly US$12M of capacity per company — enough for US$1–25M checks from pre-seed into Series A while staying hands-on. Compare Y Combinator's hundreds per batch-year or Gradient's 134 disclosed: Conviction sits at the opposite pole, and its 'first institutional investor' pitch only works thin.
Because Fund II was the moment to scale the thesis: the US$230M close more than doubled Fund I's US$101M, and Mike Vernal arrived as General Partner from Sequoia at the same time. The traffic ran both directions that cycle — specialist firms gained big-firm partners just as those firms pivoted aggressively into AI — and Conviction's concentrated model gave a departing partner a focused book to lead.
AI bolted onto traditional software. Conviction invests only where the product itself is built on frontier AI — which is why the 19-company book spans a European lab (Mistral), inference (Baseten) and vertical applications (Harvey, OpenEvidence) without a single 'AI features added to legacy SaaS' story. The filter is about how the product is constructed, not which sector it sells into.
By entering years earlier. Mistral, Harvey, Sierra and Cognition appear in Conviction's 19-company book and also in a16z's and Sequoia's — but a16z co-led Cognition's US$2B Series E at US$48B in September 2026, while Conviction holds it from the early rounds its model targets. Early US$1–25M checks plus follow-on rights buy the same exposure as late nine-figure checks, at a different entry price.
As sourcing infrastructure for a concentrated book. No Priors, the Embed grants, the Commit fellowship and Mixture of Experts recruiting events keep Conviction inside the researcher-and-founder community where its next 19 must come from — an intensity a 134-company portfolio would not need. a16z builds platform teams for hundreds of companies; Conviction builds media and grants to see 19 coming.