Allbirds built its name on merino-wool sneakers. Now the public company behind the brand is trying to build something very different: AI infrastructure. After selling its footwear brand and assets, the company has renamed itself Smartbird as it builds out the team needed to run the new business.
Smartbird is chasing demand for private AI clusters at a time when companies are weighing where sensitive workloads should run, how much control they need over infrastructure, and whether newer compute providers can match hyperscalers on reliability. The company still needs GPUs, power, colocation capacity, operations talent, and customers.
How Allbirds became Smartbird
The pivot began on April 15, when Allbirds announced plans to become NewBird AI and use a $50 million convertible financing facility to acquire high-performance GPU assets for customers seeking dedicated AI compute. By the June 17 announcement, the company had completed the footwear sale, changed its name to Smartbird, and named Nadia Carlsten to lead the new business.
Smartbird announced Carlsten as president and CEO on June 17, with the SEC filing listing her appointment as effective June 18. The Guardian and other outlets reported the sale price at $39 million, marking a sharp fall for a consumer brand that went public in 2021.
Smartbird did not announce a conventional equity round. In a June 15 SEC filing, the company said it increased the amount of senior secured convertible notes it may issue and sell from $50 million to up to $100 million.
Smartbird is not pitching AI shoes or consumer AI software. Its stated goal is managed, dedicated AI infrastructure for enterprise customers that want more control over where workloads run and how infrastructure is operated.
Carlsten brings relevant infrastructure credentials: she previously led DCAI, worked on advanced computing platforms at SandboxAQ and AWS, and holds an engineering doctorate from UC Berkeley.
The infrastructure gaps behind Smartbird’s AI bet
Carlsten told Business Insider the company will focus on customers such as midmarket enterprises, pharmaceutical companies, financial-services firms, and sovereign AI users that want persistent AI infrastructure without putting proprietary data into shared, multi-tenant systems.
Some enterprises may prefer dedicated clusters for proprietary models, data residency, security, or infrastructure-control requirements, especially as AI projects run into data readiness and workload-placement problems. But hyperscalers, neocloud providers, colocation firms, and hardware vendors are already chasing the same demand, while Amazon’s reported talks to sell custom AI chips show how quickly cloud providers are trying to control more of the hardware layer.
Smartbird’s plan depends on physical capacity as much as investor interest. AI data center projects face tight power availability, crowded colocation markets, long equipment lead times, and growing scrutiny over who pays for AI data center expansion. The company says its service would cover the lifecycle of private AI clusters, from procurement and deployment to operations and hardware refreshes.
Smartbird also has to prove that customers will trust a newly rebuilt company with critical AI workloads. The company says it is in active discussions with prospective customers and is designing its first cluster deployments, but it has not yet announced named customers, a GPU supply agreement, a colocation or power partner, or a head of infrastructure operations.
The next disclosures to watch are a GPU supply deal, a colocation or power partner, named customers, and senior infrastructure hires.
For enterprise technology leaders, Smartbird’s next test is concrete: securing the hardware, power, operations talent, and customer contracts needed to turn an AI pivot into a working infrastructure business.
Also read: Google DeepMind’s AI agent roadmap shows why enterprise AI systems increasingly need monitoring, access controls, audit logs, and real-time safeguards.