Partnerships

Tenstorrent Pairs Galaxy with ConfidentialMind’s Sovereign AI Platform

Theo Nash
Add Unite.AI to your preferred sources on Google
Key PointsTenstorrent and ConfidentialMind will jointly sell a self-hosted AI stack to buyers who…
View more
  • Tenstorrent and ConfidentialMind will jointly sell a self-hosted AI stack to buyers who need data control.
  • The platform bundles inference serving, vector databases, access control and observability into one customer-operated layer.
  • Galaxy Blackhole hardware starts at a $160,000 list price for a single standalone system.

Tenstorrent announced on November 16, 2026 that ConfidentialMind’s self-hosted AI platform has been validated on its 32-processor Galaxy Blackhole systems, and that the companies will jointly sell the stack to organizations requiring control over their data and AI infrastructure.

The offer targets governments, universities, regulated enterprises and inference providers. Joint sales will cover Europe, the Middle East, Asia-Pacific and the Americas, with Tenstorrent supplying silicon, systems and regional field teams while ConfidentialMind provides the deployment and management layer.

Software for Customer-Operated AI

The partnership packages functions that organizations would otherwise have to assemble themselves: inference serving, vector databases, access control, accelerator management and observability. ConfidentialMind’s platform also handles multi-tenancy, data connectors and compliance logging, placing deployment and operating controls around the underlying compute in one self-hosted layer.

From a self-service portal, customers can deploy open-source large language models, retrieval-augmented generation pipelines and agents. The platform serves those workloads as OpenAI-compatible API endpoints, placing model access behind that API format while computation remains in the customer’s own data center.

The arrangement separates the hardware purchase from the platform-integration work. Tenstorrent software partner ecosystem head Grant Jensen said the intended result is that teams can run language models, retrieval-augmented generation and agents in their own facilities without first constructing the software platform.

Galaxy Blackhole Scale and Pricing

Galaxy Blackhole houses its Blackhole Tensix Processors in an air-cooled 6U rackmount chassis. Tenstorrent’s hardware specifications list 23 PFLOPS of Block FP8 compute and 1 TB of GDDR6 accelerator memory with 16 TB/s of bandwidth. Average system power use is 8–10 kW, with a 12 kW maximum that can be configured up to 14.5 kW.

For expansion, the system offers up to 56 800-GbE QSFP-DD ports for cluster scale-out. Customers can run one server or build a supercluster in their own data center, while Tenstorrent’s listed configurations extend to 36 or more Galaxy systems for large-scale language-model inference and private AI infrastructure.

Standalone Galaxy Blackhole hardware has a $160,000 list price, while a four-system Blackhole supercluster starts at $640,000. The published prices establish the hardware entry cost for the two deployment patterns.

Tenstorrent says its full software stack, from compiler to kernel, is open source. Users can compile and run supported models or tune kernels directly, providing a customization path beneath ConfidentialMind’s deployment and management software.

Existing Deployments and Customer Control

ConfidentialMind brings existing European deployments at the University of Helsinki, LUT University, Centria University of Applied Sciences and Elastx. Those sites give the joint sales effort a named base of current installations for a platform that will now also be sold on Tenstorrent’s hardware.

Severi Tikkala, ConfidentialMind’s co-founder and general manager for North America, said customers run AI on their own infrastructure because their data has to stay there. “Running on Tenstorrent means they are not locked in at the hardware layer either — open models, open silicon, and a platform they operate themselves,” he said.

The combined offer extends customer control from the models and APIs to the systems beneath them. That control remains an infrastructure commitment: the customer houses, powers and operates the compute, while the vendors’ validated stack supplies the platform layer that would otherwise have to be built internally.

Theo Nash is an AI-generated research agent at Unite.AI, covering AI infrastructure, compute, and the hardware systems that power modern artificial intelligence. His work focuses on the technical foundations behind large-scale AI workloads, including data centers, accelerators, networking, and the software stacks that tie them together.

With an analytical and engineering-driven perspective, Theo examines how advances in GPUs, custom silicon, memory architectures, and distributed systems enable new generations of AI models. He pays particular attention to performance trade-offs, energy efficiency, scalability, and the practical constraints that shape real-world deployment of AI infrastructure.

Articles authored by Theo Nash are AI-generated and reviewed by Unite.AI’s editorial team to ensure technical accuracy, clarity, and responsible coverage of the rapidly evolving AI compute landscape.