Bessemer Closes $5.75 Billion to Back AI Startups From Seed to Growth
Bessemer Venture Partners has closed $5.75 billion across two new funds to invest across the AI stack. The firm announced the fundraise on September 23, 2026. TechCrunch
Of that total, $1.75 billion is for seed and early-stage investing, the first institutional money a startup raises to build a product. The other $4 billion is for growth startups, the larger rounds used to scale. TechCrunch The two-fund setup keeps both stages, from first check to scale capital, inside one fundraising cycle.
Bessemer connects the funds to work already underway. Since 2022, it has invested in more than 260 AI-native companies, startups built around machine learning from the start. TechCrunch It puts its total investment in AI-related startups to date at $3 billion. TechCrunch
The announcement was distributed via Business Wire on September 23, 2026. Business Wire It carries the headline "Bessemer Venture Partners Closes $5.75 Billion to Back Founders From Seed Through Growth". Yahoo Finance Bessemer's own news page lists the item as "$5.75 billion to back the founders building what's next", described as new capital to continue fueling AI innovation. Bessemer Venture Partners
The broader context here is structure as much as size. Having seed and growth funds under one manager lets Bessemer take early product risk and still write larger checks later for hiring, infrastructure, and go-to-market expansion. For founders, that continuity helps. It lowers the handoff problem, where an early backer cannot follow on, and it gives the growth team direct knowledge of company history that outside late-stage investors would need to rebuild.
Looking at what this means for builders, the language is worth parsing. Bessemer uses two different phrases. AI-native companies are systems designed around machine learning from inception, where models shape architecture, data pipelines, and product behavior. AI-related startups is broader, covering tools, infrastructure, and applied products where AI is central but not the original premise. Keeping both terms makes sense while value is still shifting between foundation capabilities, the base models and platforms, and vertical deployment, or AI put to work in specific industries, and neither layer has settled.
In my view, the more telling figure is the balance between money spent and money available. Three billion dollars already invested means an existing operating portfolio. Five point seven five billion in new commitments adds both reserve capacity and new-deal capacity. That ratio gives Bessemer room to defend pro rata, or its right to maintain ownership in follow-on rounds, in current AI positions while opening new ones at seed, where ownership is built, and at growth, where ownership is protected. For enterprise and infrastructure engineers judging vendors, that follow-through funding is practical. It supports hiring stability, roadmap funding, and the ability to carry inference and training costs, the day-to-day expense of running models and the one-time expense of teaching them, through uneven adoption.
The long arc here remains constructive. More dedicated AI capital will fund more experiments. Most will fail or consolidate. That is normal. What lasts are the interfaces, workflows, and cost structures that survive contact with production. If even a fraction of those 260-plus early bets grows into lasting platforms, the next group of founders will start with better building blocks.
