Phinity Labs raises $5.2M Seed for autonomous chip design. Sonya Jin and Aadi Nashikkar arrived at chip design through the work of teaching AI models. At Phinity Labs, the San Francisco cofounders are turning that background toward a costly engineering decision: how many possible architectures a team can explore before committing to silicon. Phinity announced a $5.2M Seed on October 8, 2026, led by Uncork Capital, with Moxxie Ventures and angel investors including Jeff Dean participating. The company previously received pre-seed backing led by Pear through PearX. The funding supports an applied research business pursuing autonomous chip design, a field where the value of an additional experiment depends on whether its result survives physical implementation. For semiconductor teams and foundation-model laboratories, Phinity's approach puts more attention on the search preceding a chip. Faster exploration could make specialized architectures easier to evaluate while workloads continue changing. Phinity's homepage places its broader ambition on a future timetable: by 2028, a requested chip would become a GDS layout file in weeks. What Phinity's $5.2M Seed establishes. The announced transaction is a Seed financing, with Uncork Capital's original investment account corroborating its lead role. Phinity names Moxxie and Jeff Dean in the round; Uncork also names Pear among the backers. Pear's earlier involvement and the new financing belong to the same company history, but the undisclosed pre-seed amount leaves cumulative funding unknown. That distinction matters when assessing the resources behind a hardware ambition. A $5.2M announcement measures an investment in Phinity, while the economics of any eventual chip depend on the design, tools, fabrication and customer requirements involved. Neither a valuation nor a detailed allocation of the proceeds was disclosed in the original announcement. The founders' route into the field helps explain the company's research orientation. Jin's disclosed background includes NVIDIA model-training data work and graph machine learning at AWS AI Lab. Nashikkar has worked in synthetic data and post-training and previously built a business that was acquired. Uncork identifies both as cofounders who met at Stanford; the public sources reviewed do not establish current CEO or CTO titles. The product argument is about exploration. Phinity describes agents that design, evaluate and revise computational systems using physical feedback. That sequence gives the research a concrete engineering purpose. A candidate architecture has to be judged against power consumption, performance and area, and the outcome of that evaluation has to inform the next attempt. Consider the practical choice this creates for a design manager. An engineering team can have several plausible ways to serve a workload while possessing enough time to examine only a few carefully. Making each examination less costly would change the number of alternatives worth considering. The commercial significance lies in the team's ability to compare useful options before its commitments become harder to revise. Phinity's 2028 target concerns a GDS file, the layout artifact associated with the move toward manufacturing. Producing a layout remains part of a larger physical undertaking, with fabrication and the customer's requirements still shaping the finished device. The roadmap should be read as the company's intended destination, alongside its current research and commercial work, rather than a statement that arbitrary chips can already be ordered and delivered autonomously. AI-assisted chip design has an established market. Synopsys reported its first 100 commercial DSO.ai tapeouts in February 2023. That historical milestone makes AI-assisted design a customer practice with a record of use. It also gives semiconductor buyers a baseline for evaluating a new approach: optimization has to fit into the engineering process they already depend on. Cadence Cerebrus offers AI-based design-flow optimization around power, performance and area, including concurrent work across multiple blocks. Cadence's product description retains an engineer's ability to inspect results and control the workflow. Phinity therefore enters a field where customers already have substantial automation to assess and integrate. The original Phinity and Uncork accounts report favorable task-level results, but the material reviewed does not provide a public, comparable evaluation against those commercial platforms. Buyers will need to examine the workload, tools, constraints and validation supporting any performance claim. An attractive result on a defined design task can open a technical conversation; the purchasing decision extends across the workflow that must carry it into a usable chip. A research business with customers in the loop. Phinity reports work with major foundation-model laboratories and an eight-figure annualized run rate. Customer identities and audited financial statements were not disclosed, so that figure should remain attributed to the company. It describes an annualized measure rather than a published account of revenue collected over a completed year. Uncork describes commercial partnerships developing alongside Phinity's own chip-design agents. That arrangement connects the startup's research agenda with practical engineering demands. Work for model laboratories can expose the systems to difficult problems while giving the company a business to operate as its broader autonomous-design ambition develops. For an operator evaluating Phinity, the useful questions concern which parts of a design the agents can handle, which results can be reproduced and how the work enters existing verification and implementation tools. Phinity's official careers destination also shows where it is seeking people to build the business. The cofounders are bringing model-training experience into an industry where the experiment eventually has to occupy space, consume power and justify the engineering team's decision to build it. DevCuration Data AI Infrastructure funding, last 30 days. DevCuration's funding database tracked 41 AI Infrastructure rounds totaling $15.3B in disclosed capital over the past 30 days. Recent deals DevCuration covered:

AI agents promise to compress chip development from years to weeks, bringing custom silicon design to foundation model labs with eight-figure early...