The artificial intelligence industry is seeing a major shift in elite talent. Core researchers who built foundational models over the past five years are now launching their own ventures. They want to solve complex technical bottlenecks on their own terms.
The latest move comes from Shyamal Hitesh Anadkat. He is an Indian-origin researcher who spent nearly four years at OpenAI working across applied research and scaling systems. Now, he enters the startup scene with his new venture, Empirical Machines. This marks another high-profile spin-off from the industry’s most prominent lab.
The Strategic Mechanics of Empirical Machines
When an engineer with Anadkat’s background leaves a top lab, venture capital firms pay close attention. His deep scaling experience is rare. Because of this pedigree, investors treat his new venture as an urgent, high-priority bet.
According to people familiar with the developments speaking to Money control, Anadkat is actively negotiating to secure roughly $15 million in funding to get Empirical Machines off the ground. The active discussions involve heavyweight institutional investors, specifically Peak XV Partners and Bessemer Venture Partners.
This aggressive backing isn’t just a blind financial bet on a smart engineer. It represents a highly calculated, structural move by venture firms to capture critical AI infrastructure at the ground floor. Institutional investors are painfully aware that the next massive wave of enterprise software value will not come from building thin wrappers around existing large language models.
Instead, as enterprise priorities pivot, The Proof-of-Concept Era is Dead: Where 2026 AI Budgets Are Actually Going, and the real returns will flow from specialized, deeply empirical systems that are designed from scratch to handle complex business workflows and unique data processing tasks.
Backing a founder who already knows exactly how frontier models fail and succeed at scale gives these funds a massive statistical advantage.
The Gravity of a Heavy Initial Raise
Raising $15 million right away sounds steep for a young software company. However, AI startups follow different economic rules. Building proprietary AI systems demands heavy upfront capital. Founders must pay for cloud compute, secure GPU clusters, and hire top engineering talent.
Peak XV and Bessemer understand that underfunding an AI startup almost guarantees failure today. With this capital, Empirical Machines can skip the slow growth typical of older tech startups.
They can immediately begin securing the scarce GPU clusters and dedicated research minds necessary to understand The AI Stack Explained: Models, Vector Databases, Agents & Infrastructure in 2026 and actually build something novel.
This heavy initial capitalization is precisely how the current generation of machine learning innovators builds a defensive moat before they even launch a commercial product.
They are not merely writing standard application code; they are constructing complex mathematical engines that require immense raw power to train and execute. That fifteen million dollar figure acts as a necessary entry ticket into the infrastructure race, giving Empirical Machines the exact leverage needed to compete in a landscape where raw compute is the ultimate currency.
