Atomic Semi, the semiconductor startup founded by chip architect Jim Keller and engineer Sam Zeloof, has rebranded as Fab2 and relocated to Texas. Unlike traditional chipmakers building massive factories, Fab2 plans to mass-produce compact, modular semiconductor fabs. If successful, this approach could offer a new tool for rapid prototyping and small-batch chip production, rather than competing head-on with giants like TSMC.
The company calls this concept the ”fab fab”: a factory that designs and assembles its own equipment and major components for small-scale semiconductor manufacturing. Fab2 engineers nearly every essential part in-house, including pumps, valves, gas lines, vacuum chambers, and lithography systems. These elements are then assembled into machines, which form the building blocks of their mini fabs.
The key differentiator is producing chips on smaller wafers with simpler setups. Most semiconductor fabs revolve around expensive 300mm wafers and huge production lines. Fab2’s model shrinks the footprint, costs, and launch time dramatically. They estimate their modular fabs could churn out chips within hours, rather than the months-long queues typical at large foundries.
On top of hardware, Fab2 develops its own design and simulation software stack. This integration is vital: past attempts to cut chip manufacturing costs faltered due to reliance on manual tuning and inconsistent results. Fab2’s tightly coupled combination of equipment, software, and workflows aims to avoid those pitfalls.
Fab2’s origins and technical capabilities
The idea behind Fab2 is rooted in Sam Zeloof’s early experiments-back when he was a teenager, he demonstrated that chips around the 300nm node size could be manufactured in a garage-like setup. Later, he partnered with Jim Keller, a semiconductor luminary known for his work on AMD, Apple, Tesla, and Intel processor architectures. This pairing gave Fab2 both technical credibility and industry connections to attract investors and suppliers.
Currently, Fab2 operates three sites: its headquarters and R&D/manufacturing center in Austin, Texas; ”fab fab” assembly operations in Lockhart, Texas; and the original ”garage fab” still working out of San Francisco. The startup employs around 84 people and is actively expanding its Texas workforce.
Fab2’s modular fabs rely on electron beam lithography-writing chip patterns directly without masks, unlike traditional photolithography used in mass production. This makes for easy iteration and tiny production runs but results in slower throughput. It’s a niche solution unsuitable for matching the volume economies of TSMC, Samsung, or Intel’s mega fabs, which produce millions of identical chips.
That said, Fab2 is targeting a real market need. According to SEMI, global semiconductor equipment sales surpassed $100 billion in 2025, with significant demand not only for cutting-edge chip processes but also for mature tech nodes used in power electronics, sensors, microcontrollers, and specialized chips. In these areas, getting quick access to smaller production runs often matters more than pushing transistor density limits.
Fab2’s approach echoes broader trends in chipmaking equipment. Major players like Applied Materials, Lam Research, and ASML profit by selling tools rather than owning fabs. The US CHIPS Act is fueling investment in local, more flexible manufacturing, encouraging startups like Pragmatic Semiconductor to prove that simplified fabs can find their own segment-focusing on select chip classes instead of chasing advanced nodes.
The critical test will be whether Fab2 can turn impressive engineering demos into a reliable product. Over the next 12 to 24 months, the company needs to show it can deliver a mini fab to customers, deploy it without an army of contractors, and produce chips with consistent yields. Success would carve out a niche between academic prototyping labs and massive contract semiconductor factories.