BoolSi raises $6M seed to compile ordinary code into custom chips
What's the deal? BoolSiDealroom has a profile for this one. Try Dealroom →, a chip-design startup founded by Mihailo Isakov, has raised $6M in seed funding to build a compiler that turns ordinary software into custom hardware. Fine Structure VenturesDealroom has a profile for this one. Try Dealroom →, an F-Prime fund, led the round, with Pillar VCDealroom has a profile for this one. Try Dealroom →, Fifth Quarter VenturesDealroom has a profile for this one. Try Dealroom →, and Coalition VenturesDealroom has a profile for this one. Try Dealroom → participating.
The idea: a developer points BoolSi's compiler at a slow spot in a C or C++ programme, and it returns a custom circuit and driver in minutes. Doing the same work by hand takes months and requires years of digital-logic training.
Why now? Custom hardware can run fixed workloads far faster than general-purpose processors, but designing chips has always been closer to watchmaking than programming. Open-source resources are thin, the learning curve is steep, and most software engineers never get a path into hardware — even when their applications would benefit from it.
BoolSi reframes the problem as learning what a programme does, not translating how it is written. It trains machine-learning models that converge into fully digital circuits, using the source code itself as a synthetic data generator. A fuzzer explores the input space and turns every run into a labelled training example, which the company says makes 100% accuracy both achievable and required.
Early benchmarks are striking. A regex routine that took 2.66 milliseconds on an ARM Cortex-A9 processor ran in 0.325 milliseconds on a single BoolSi hardware agent — roughly eight times faster. Eight agents together finished in 0.042 milliseconds, 63 times quicker than the CPU.
The startup is targeting embedded developers in robotics first, where tasks like motor-control loops, sensor fusion, and model-predictive control hammer general-purpose processors repeatedly. A private beta is set to open in Q3 2026.
What could go wrong? BoolSi's approach rests on one published benchmark so far. Proving that its compiler handles the messy complexity of real-world production code — not just tidy routines — will be the real test. It also faces competition from established high-level synthesis tools like Vivado HLS, Catapult, and BambuDealroom has a profile for this one. Try Dealroom →, which already serve hardware engineers even if they haven't cracked the software-developer market.
Formal verification across independently trained models is clever, but hardware bugs that slip past verification can be catastrophic — especially in safety-critical robotics applications.
The signal: Fine Structure Ventures, an F-Prime fund, leading this seed round puts BoolSi in the orbit of one of the larger life-sciences-and-tech investment platforms, signalling that established venture capital sees ML-driven chip design as a genuine category rather than a research curiosity. With the global shortage of hardware engineers only deepening as AI and edge-computing workloads multiply, tools that let ordinary software developers generate custom silicon could tap a far larger market than traditional EDA vendors have ever served.
Read more: SiliconANGLE
