Axiom Partners: 35 AI Investments, Half of Them Acknowledged Failures

Sandhya Venkatachalam is structuring Axiom Partners for AI applications rooted in industry, construction, or insurance. The $52 million fund plans 35 investments and assumes that about half will fail, betting on a rare success to make the fund profitable. The organization combines active practitioners and an internal tool, Axiom Brain, to accelerate decision-making.

Portfolio: 35 targeted investments and about half expected to fail

Axiom Partners manages a $52 million fund and plans to make 35 investments. Sandhya Venkatachalam expects that about half of these bets will fail, whether through closures or companies that do not take the desired trajectory. The fund's model relies on seeking an exceptional outcome and the idea that one extraordinary investment can alone make the fund profitable. Investing early enough and then seeing a company grow significantly can offset many failures. Accepting losses is an integral part of their upstream investment approach, before the opportunity becomes evident. This framework is embraced and planned in the fund's strategy, which targets AI startups in sectors like construction, industry, and insurance.

Business criteria: labor budgets and six-figure contracts

Among Axiom's criteria, clients must be willing to pay for AI within their labor budgets. Even at the alpha stage or in design partnerships, the team conducts due diligence to establish this willingness to buy. Sandhya Venkatachalam notes that she often sees contract values in the hundreds of thousands of dollars, rather than the much more modest amounts associated with intermediate software tools. This purchasing mode has manifested in the majority of the companies in the portfolio.

Organization: part-time practitioners and Axiom Brain for quick decisions

Axiom has built its organization around active AI practitioners, involved part-time with dedicated time, who receive a share of the fund's profits and are treated as partners rather than façade advisors. Their other jobs are central to the model, and they would be less valuable if they detached from them. The fund also deploys AI internally with Axiom Brain, a tool designed to read trends, identify people and companies, and accelerate due diligence like other workflows. For Sandhya Venkatachalam, the value of these means is the ability to act quickly. She argues that it is difficult to keep pace with the market without building, marketing, or deploying AI, and that these operators update the analysis grid and attract founders due to their experience with real-world challenges.

Product thesis: delivering an outcome and addressing the "last mile"

Axiom aims for "AI for the real world," with the belief that it should benefit beyond early adopters. The fund prioritizes underserved sectors where AI delivers an outcome, rather than yet another software tool, particularly in construction, industry, and insurance. The supported companies can combine hardware, sensors, or robotics, or rely solely on software, with the common goal of executing work that matters to the client. The fund generally steers clear of products that resemble traditional enterprise tools and seeks systems that effectively deliver an outcome. Sandhya Venkatachalam believes that a product becomes difficult to replace when it performs significant work that is highly valuable to the client, which she describes as the "last mile." In industry, this implies deep integration into clients' systems, understanding and training on their data, learning key workflows, and ensuring the outcome. This grounding goes beyond simply placing an interface on a model, creating relationships and capabilities that are hard for another startup to replicate, and involves work that large AI model providers are often reluctant to undertake themselves.

Origins and influence: from Groq to openness to atypical profiles

Before Axiom, Sandhya Venkatachalam supported Groq at a time when inference was far from an obvious investment. Coming from both hardware and software backgrounds, she examined the reasons behind Google building its own network switches, later discovering that it was also developing its chips. This exploration led her to Jonathan Ross, involved in these efforts and later the founder of Groq, who advocated that inference would be the largest future market. In 2016, she barely understood inference, but her intuition that the diffusion of models would create a demand for infrastructure guided her investment. This experience today fuels her preference for being slightly ahead and patient, without seeking contrarian views for their own sake, while investing before the consensus on what will be built on infrastructure and AI models. In terms of sourcing, she draws from Khosla Ventures the openness to the backgrounds of great founders and seeks less obvious profiles, as she believes Silicon Valley often narrows this pool to a few typical paths. Her previous journey—building companies, product leadership in a hardware company sold to Cisco, a position at Skype before its sale to Microsoft, a stint at Social Capital where she led the initial investment in Groq, and then Khosla Ventures—preceded her current role as founder and managing partner of Axiom Partners and shaped her early interest in data and machine learning.