Tau Ventures has launched its third fund on a contrarian wagerAs venture capital crowds into a handful of AI giants, Tau Ventures has launched its third fund on a contrarian wager — that the best returns in AI are still made at the seed stageIn July 2023, a small Silicon Valley firm wrote a small seed cheque into Assort Health, a startup using AI to handle the patient scheduling and call-centre operations that medical practices drown in. Three years later, Assort is valued at $1.2 billion after a $120 million raise and that early stake is worth about low 8-figure on paper, a roughly twenty-fold gain in a company most investors never got the chance to see.The firm was Tau Ventures. The deal is a fair summary of its method: get in earliest, in workflows others overlook, and let conviction compound. Now the firm has launched its third fund — and expanded its investment team — on the belief that the biggest opportunity in AI investing sits precisely where the biggest money is not looking.Small by designFounded in 2019 by Amit Garg and Sanjay Rao, who first worked together at Norwest Venture Partners in Silicon Valley more than fifteen years ago, Tau manages over $100 million across its funds.Tau reviews roughly 6,000 companies a year and invests in about one a month, an acceptance rate of 0.2 per cent, a bar far higher than top universities like Harvard. Of the more than 80 companies that have made it through, 95%+ have gone on to raise their next round of funding.Fund III continues the playbook: initial cheques of $500,000 to $1 million into seed-stage companies, with capital held in reserve to concentrate into the breakout winners. The focus is digital health, enterprise AI, particularly fintech infrastructure and cybersecurity, and automation and robotics, the arena the industry has taken to calling physical AI. Tau invests in companies led by US-based chief executives, though teams can be, and often are, distributed globally, a structure familiar to the many founders who build engineering teams across the world while selling into American markets."The question in AI is no longer whether the technology works," Garg says. "It's who captures the value. We believe it will be the companies that own a specific workflow, where the return on investment is obvious, the labour intensity is high and the customer pain is real. Our job is to find them before the crowd does."The contrarian arithmeticTau's timing is pointed. Over the last three decades, early-stage venture funds have returned an average of 21.3 per cent annually against 12.6 per cent for late-stage funds, according to Cambridge Associates benchmarks, a gap of nearly nine percentage points a year, compounding.Yet institutional money keeps flowing the other way. In the first half of 2026, by industry estimates, just a small group of elite venture capital and private equity mega-firms captured over 73 per cent of all new limited-partner commitments, while OpenAI and Anthropic alone accounted for 43% of all global venture funding in the first half of 2026, according to Crunchbase data."When most of the new money concentrates in same AI conglomerates, everyone ends up fishing in the same pond," Rao says. "The mathematics of venture hasn't changed; the biggest multiples are still made in the smallest rounds. What has changed is how few investors are actually positioned there."When a $10 billion fund writes the first cheque into a two-person team with a working prototype, they rarely devote time and attention to the investment. A smaller fund though specializes exactly at that seed stage and, where pricing has remained comparatively rational, that access is the differentiation.Proof in the portfolioAssort is the headline, but not the outlier. Tau backed Labelbox, the data infrastructure company whose platform underpins AI development at large enterprises, and has kept adding to the position through follow-on rounds, a conviction bet on the picks-and-shovels layer of the AI build-out. In cybersecurity, its early cheque into ArmorCode, which helps enterprises manage application security as AI multiplies the volume of code being shipped with Tau contributing investor introductions, market positioning and a board observer seat along the way. Other holdings stretch across the firm's verticals: Chef Robotics, whose robots assemble meals for food companies; Infinitus, whose AI voice agents automate the phone calls that clog American healthcare; and Signos, which pairs continuous glucose monitoring with AI to help people manage metabolic health.Behind the wins is an unusually involved operating model: customer and strategic introductions, market intelligence, investor mapping for the next round, go-to-market positioning and senior hiring. When a portfolio company drops below six months of runway, the firm treats it as an emergency, finding customers, connecting investors and, when necessary, finding the right acquirer to protect the downside."We look for founder-market fit above everything," Garg says of his filter. "People building in a domain they understand deeply, with an AI-native product and customers already leaning in. Credentials are a plus. Conviction is the requirement."New faces at the tableFund III also brings new investors to the team. Currently they have two Principals, two people part-time, and three Venture Partners.Insoo Chang is a Principal leading enterprise AI and fintech investing. Previously a senior vice-president on Citigroup's strategic investment team, where he backed early-stage wealth-tech startups, Chang also co-founded NotedSource, an AI platform for research collaboration. He holds an MBA from Harvard Business School and an economics degree from Cornell University."At Citi, I saw how much of global finance still runs on manual workflows and legacy plumbing," Chang says. "The AI companies that win this cycle won't be the ones with the flashiest demos; they'll be the ones that quietly remove that friction. Those companies get built at the seed stage, and that's where we want to meet them."Sam Bogrov is a Principal focused on healthcare. Formerly at Pediatric Associates, where he worked on mergers and acquisitions, Bogrov cut his teeth in private equity at Investindustrial and Platinum Equity after starting his career in M&A at Oppenheimer & Co."I've spent my career on the deal side of healthcare, but I've also spent years inside the system as a patient," Bogrov says. "I know first-hand how broken it can feel. Backing founders who are using AI to fix it isn't abstract for me, it's personal." A widening circleTau has also built a following unusual for a firm its size: a community of more than 5,000 investors on its mailing list, with posts regularly crossing 100,000 views, and limited partners who co-invest directly alongside the fund. The firm treats that ecosystem as part of the product - a source of pilots, senior hires and follow-on capital for its founders, not merely a distribution list.What comes next, Garg argues, is the decade in which AI stops being a demonstration and becomes infrastructure; agents doing real work inside hospitals, banks and factories. "Every platform shift has a window where discipline beats scale," he says. "We built Tau for that window."For a market transfixed by giants, the firm's wager is that the future of AI will be decided not only in billion-dollar rounds, but in the two-in-a-thousand seed cheques written before the rest of the world catches on.