The thing isn’t even out of simulation yet, and the first developer kits won’t land until late 2025. Full production is unlikely before 2026. So, while Zeus might sound like a Greek god in the making, right now, it’s just a theoretical deity scribbled onto PowerPoint slides.
Still, the emergence of a new discrete GPU maker outside of China is worth noting, considering that the number of Western companies in the game has dwindled to just three: Nvidia, AMD, and the ever-troubled Chipzilla.
Bolt claims that Zeus is more than just another gaming GPU—it’s gunning for the high-performance computing and rendering markets, targeting movie studios, scientists, and supercomputer simulations.
According to Tom's Hardware, unlike GPUs from AMD, Intel, and Nvidia that rely on proprietary instruction set architectures, Bolt's Zeus relies on the open-source RISC-V ISA, according to the published slides. The Zeus core relies on an open-source out-of-order general-purpose RVA23 scalar core mated with FP64 ALUs and the RVV 1.0 (RISC-V Vector Extension Version 1.0) that can handle 8-bit, 16-bit, 32-bit, and 64-bit data types as well as Bolt's additional proprietary extensions designed for acceleration of scientific workloads.
While Zeus cleans the clock of the RTX 5090 in path tracing and FP64 performance, it’s looking rather feeble in AI workloads—an area where Nvidia’s GPUs reign supreme. Despite supporting rasterisation, Zeus might struggle with traditional gaming graphics, meaning it’s unlikely to seriously threaten the established order in consumer GPUs.
More concerning is that Bolt is pitching Zeus as a data centre and scientific computing alternative, yet it’s still stuck in the simulation phase. By the time the hardware exists, Nvidia will have moved on to even more monstrous AI accelerators, and Bolt might find itself trying to compete with vapourware-level promises against an industry that doesn’t wait around for upstarts to catch up.
The company’s chipset-based design strategy is all about scalability, ranging from the modest Zeus 1c26-032 with a single processing unit and 32GB of LPDDR5X memory to the monstrous Zeus 4c26-256, a four-chipset server-grade beast with 256GB of LPDDR5X and up to 2TB of DDR5. That’s a lot of memory muscle, but memory alone doesn’t win GPU battles—ask Intel’s Arc division how that turned out.
The Zeus GPUs have high-speed connectivity, featuring 400GbE and 800GbE ports for fast data movement between networked GPUs. It clarifies that Bolt is gunning for data centre workloads rather than traditional gaming. Including CXL 3.0 support via PCIe Gen5 x16 slots further underscores the focus on memory pooling and high-performance computing. However, this means the cards will depend more on strong software integration—something that often trips up new GPU players.
While the Zeus design prioritises large memory capacities over raw bandwidth, Bolt’s real challenge is proving it can compete with Nvidia’s deeply entrenched dominance in AI, rendering, and supercomputing.
Given that Zeus is nothing more than a set of PowerPoint slides and simulations, Nvidia and AMD aren’t precisely sweating yet. If Bolt can deliver, it might carve out a niche in scientific computing and high-end rendering. Still, without rock-solid software support or a working product, Zeus risks becoming another footnote in the long history of GPU startups that flew too close to the sun.
