Tenstorrent, the innovative AI chip startup led by legendary processor architect Jim Keller, is undergoing a strategic realignment that includes a workforce reduction of about 7.5%. Following the layoffs, the company’s headcount will stand at roughly 1,000 employees. The move reflects a targeted reallocation of resources rather than a broad retrenchment, as Tenstorrent pivots from a wide-ranging commercial approach to a business model focused directly on software developers.

The company believes developers play a pivotal role in the deployment and integration of AI technologies. By selling directly to this group, Tenstorrent aims to simplify its operating structure, improve efficiency, and strengthen its positioning in a market dominated by industry heavyweight Nvidia. The developer-first strategy is designed to lower barriers to adoption through tighter hardware–software integration, while fostering a stronger ecosystem to support long-term growth.

According to EE Times, the restructuring is closely tied to intensifying competition in the AI hardware sector. Concentrating on developers allows Tenstorrent to differentiate itself beyond traditional enterprise sales channels and accelerate the real-world adoption of its AI solutions. The layoffs also reflect broader pressures facing technology startups, where maintaining financial discipline increasingly requires difficult strategic decisions.

Founded in 2016 in Toronto by former AMD and Nvidia architects Ljubisa Bajic, Ivan Hamer, and Milos Trajkovic, Tenstorrent focuses on high-performance, low-power AI chips and RISC-V processors. Jim Keller joined the company as chief technology officer in December 2020 and was promoted to chief executive officer in January 2023, when the company formally shifted its strategy toward developers and open-source ecosystems. That same year, Tenstorrent relocated its headquarters from Toronto to the United States to support global expansion.

On the technology front, Tenstorrent’s Grayskull AI accelerator integrates 120 “Tensix” cores and 120 MB of on-chip SRAM, delivering up to 315 TOPS of INT8 performance at 200 watts, targeting AI inference workloads. Its newer BlackHole processor, released in 2024, features a 768-core RISC-V architecture with performance reaching 745 TOPS, which the company claims surpasses Nvidia’s A100 in overall AI throughput. In December 2025, Tenstorrent plans to launch TT-Ascalon, a high-performance RISC-V CPU aimed at servers, AI infrastructure, and automotive high-performance computing, with the Ascalon S positioned against Arm’s Cortex-A78.

Tenstorrent is also promoting its Open Chiplet Architecture, which enables plug-and-play integration of chiplets from different vendors, reducing development costs and supporting broader adoption of AI technologies. Through its renewed focus on developers, the company hopes to carve out a sustainable niche as AI hardware competition continues to intensify.