US-Based Bolt Graphics Tests Zeus Chip in One Day, Opens Taiwan Office Ahead of 2028 GPU Launch

Sunnyvale startup Bolt Graphics reported on October 13, 2026, that its Zeus test chip came alive on the first power-on and passed testing within 24 hours, with no functional issues found across a temperature span of -40°C to 85°C.

The same announcement included the opening of a Taipei office positioned to support volume production of the company's first fully commercial GPU, slated for general availability in 2028.

First Silicon Confirms Fixed-Function Physics

The 6 sq mm test chip was fabricated on TSMC's 12 nm process, with its initial tape-out finished in April.

Measured consumption came in at 1.4 watts, a sharp drop from the 35 watts drawn by the FPGA prototyping platform it replaces, and Bolt says the results beat its own internal performance and efficiency estimates.

Beyond raw efficiency, the company frames the chip as the first commercial GPU architecture to carry dedicated electromagnetic simulation hardware for both software and hardware acceleration of path tracing and electromagnetic workloads.

Bolt contends that CPU- and GPU-based software has historically been blocked by performance, accuracy, and memory limits from simulating complete products rather than individual components, and says its silicon removes the need for physical prototypes.

The company also states it has demonstrated that highly complex physics algorithms can run on fixed-function hardware, which it links to faster training of advanced robotics and world models.

"Electromagnetics is our starting point, and the goal is to let users accurately simulate anything in the world by building a complete physics simulation solution," said Darwesh Singh, Founder and CEO of Bolt Graphics.

"For years, engineers have had to choose between simulating individual components or building a physical prototype of complete products. Running Maxwell's equations on our silicon eliminates that tradeoff."

Looking Past Electromagnetics to Broader Workloads

According to Bolt, the constraints of general-purpose CPUs and GPUs also hold back computational fluid dynamics and molecular dynamics, and the company intends to accelerate all three domains.

The ambition extends to letting engineers and scientists simulate and render highly complex parts and products, train robots and world models, and support applications the company lists as silicon photonics development and research into new materials, optics, and stealth geometries.

To recognize the test chip achievement, Bolt is hosting invited guests at its Sunnyvale, California, headquarters on October 13.

Taipei Presence Tied to Supplier Coordination

The new office at Taipei Nanshan Plaza places Bolt's team in proximity to its suppliers as the flagship Zeus GPU moves toward volume production. Staff are already stationed there, and the company plans a rapid scale-up of headcount to deliver production silicon.

"Building production silicon takes close coordination with our partners," said Lawrence Lee, VP, Manufacturing Operations at Bolt Graphics.

"Having our team in Taipei puts us beside our key suppliers, allowing us to keep a pulse on the progress and move faster while ensuring quality is maintained throughout development."

Process Node Choice and the Path to Volume

Bolt deliberately selected TSMC's N5P as its production node, saying the choice was made to avoid competing for wafer allocation. With the architecture now validated in silicon, the company reports it is deepening partner relationships while customers await access to Zeus.

The rollout schedule laid out by the company begins with developer access in the first quarter of 2027, followed by initial Zeus GPU availability to Early Access Program customers in late 2027, and volume production of PCIe cards and servers in 2028.

Company Positioning

Bolt Graphics positions itself as the builder of a next-generation compute platform aimed at substantially improving cost efficiency across high-performance computing, rendering, and emerging workloads.

The company says its optimization for performance-per-dollar enables new applications and that it intends to become a foundational layer for the future of compute.