Smartbird partners with WEKA on managed AI infrastructure

Smartbird has partnered with WEKA to provide managed AI infrastructure services for enterprise AI workloads. The agreement is Smartbird's first strategic infrastructure partnership.

The deal combines WEKA NeuralMesh software with Smartbird's dedicated AI clusters to provide enterprises with a managed infrastructure platform for production AI work. The service is aimed at organisations that want private AI infrastructure in environments they control, including on-premises sites and dedicated colocation facilities.

Smartbird, which launched in June 2026, designs, builds and operates dedicated AI clusters for enterprises as a managed service. Each cluster is configured for a customer's specific workloads and assigned to a single customer rather than shared across multiple users.

The company is targeting organisations in regulated sectors such as government, healthcare and financial services, where data cannot easily be moved into shared public cloud systems. It also points to a shortage of specialist staff in many businesses to build and run advanced AI systems.

Managed stack

Under the partnership, Smartbird will use WEKA's software as the data and memory layer in its clusters. The combined offering covers facilities, compute, networking, data infrastructure and orchestration under a single managed service model.

The service is designed to help customers avoid integrating multiple suppliers and managing hardware release cycles across different parts of the stack. It is also intended to support enterprise AI workloads from initial deployment.

The companies framed the partnership around the changing economics of AI in business. As AI systems shift from model training to ongoing inference and agentic applications, factors such as GPU use, power consumption and cost per token are becoming more important to whether projects can operate at scale.

Smartbird said adding WEKA NeuralMesh is intended to improve GPU data supply and memory use, and help lower inference costs by reducing recomputation and improving time to first token. Customer data remains within customer-controlled infrastructure, with isolation, encryption and auditability features aimed at regulated industries.

Executive comments

Dr Nadia Carlsten, President and CEO of Smartbird, described the deal as part of a broader partner strategy.

"Enterprises don't want to assemble AI infrastructure from individual parts. They want a complete solution that works, operated by people who have done this before," Carlsten said.

"Every Smartbird AI cluster is designed around the customer's specific requirements, whether they're a bank, a hospital, a manufacturer, or a government agency. By incorporating WEKA NeuralMesh, Smartbird gives customers a data and memory layer proven to perform at scale. This partnership reflects our approach to building infrastructure: experienced partners, trusted technologies, and disciplined execution," she said.

WEKA positioned the partnership around the needs of businesses running continuous AI services rather than periodic training jobs. Those workloads place more pressure on infrastructure efficiency because idle GPUs and stalled data pipelines directly affect operating costs.

Liran Zvibel, Co-Founder and CEO of WEKA, said the market has often treated sovereignty and performance as competing goals.

"Agentic AI is unforgiving: idle GPUs and stalled pipelines land straight on the P&L statement. Enterprises have been told the fix is a tradeoff, performance or sovereignty, but that was always a failure of infrastructure, not a law of physics," Zvibel said.

"WEKA NeuralMesh ends the tradeoff. Smartbird delivers it as dedicated clusters customers don't have to build or operate. Together, we're giving regulated industries production AI that runs at full speed, on infrastructure only they touch, with more output per GPU, per watt and per dollar," he said.

WEKA said its NeuralMesh platform combines data storage with extended GPU memory for training, inference and agentic workloads. Smartbird said the deal is the first part of a broader ecosystem strategy as it builds out a technology partner network around its managed AI infrastructure offering.