Tenstorrent Secures $1.5 B Funding, Valuation Rises to $6.2 B for Open‑Weight AI Chip Expansion

Tenstorrent, a Toronto‑based chip startup that specializes in inference workloads, announced a $1.5 billion financing round.

The new funding values the company at $6.2 billion, more than double its $2.6 billion valuation recorded in late 2024.

Funding and Valuation

The round was led by Trace Capital, a firm founded by Quiet Capital co‑founder Ben Mahdavi, together with Innovation X.

Additional investors included Cadence Ventures, Eclipse, Fidelity, SanDisk and T. Rowe Price.

Jim Keller, Tenstorrent’s chief executive and a veteran microprocessor architect, told Bloomberg the company chose to stay independent.

Keller added that the decision followed successful tests showing Tenstorrent’s Galaxy servers could run DeepSeek models.

Strategic Focus on Open‑Weight Models

The fresh capital will be used to scale chips designed for serving open‑weight AI models and for video‑generation workloads.

Open‑weight models refer to AI systems whose parameters are publicly accessible, allowing broader deployment across diverse hardware.

By targeting this segment, Tenstorrent hopes to provide an alternative to Nvidia’s proprietary GPU ecosystem.

The company’s focus on inference distinguishes it from firms that prioritize training‑centric accelerators.

Tenstorrent’s valuation jump follows a wave of interest in AI hardware startups that emerged after 2022.

Earlier this year, Etched, another AI‑chip maker, was reported to be fielding acquisition offers valuing it at $40 billion.

Tenstorrent itself attracted acquisition interest from Intel and Qualcomm five months earlier, with potential deals cited at up to $5 billion.

The recent round therefore signals the firm’s intention to grow organically rather than sell to a larger silicon player.

Implications for the AI‑Chip Landscape

Investors such as Fidelity and T. Rowe Price are betting on a diversified AI‑chip market beyond Nvidia’s dominance.

Nvidia currently controls a large share of the data‑center GPU market, a position that many startups aim to erode.

Tenstorrent’s Galaxy server platform demonstrates that specialized ASICs can achieve comparable performance on certain open‑weight workloads.

The ability to run DeepSeek models, which are known for large parameter counts, showcases the chip’s efficiency.

Industry analysts view this as evidence that the hardware stack for open‑weight AI is maturing.

Scaling production will require significant engineering resources, which the $1.5 billion injection is intended to support.

The funding also enables Tenstorrent to expand its video‑generation chip line, a market segment gaining traction with generative media.

Video generation places distinct demands on latency and bandwidth, areas where custom silicon can provide advantages.

Tenstorrent’s strategy aligns with broader trends where companies tailor processors to specific AI tasks rather than offering generic solutions.

This task‑specific approach may lower total cost of ownership for enterprises deploying large language models at scale.

As more open‑weight models become commercially viable, demand for compatible inference hardware is expected to rise.

Tenstorrent’s progress therefore serves as a barometer for how quickly the ecosystem can diversify.

The company’s next milestones include volume‑shipments of its next‑generation inference chips and broader software integration.

Observers will watch how Tenstorrent competes with Nvidia’s upcoming Hopper and Grace CPUs in the inference space.

The outcome will influence whether data‑center operators can source performance‑critical workloads from multiple vendors.

Why This Matters