AI infrastructure startup Baseten secures $1.5 billion in funding, reaching a valuation of up to $13 billion.

      The AI infrastructure startup Baseten has secured $1.5 billion in a Series F funding round, with Australia's Blackbird VC contributing what could be the largest single investment to date by an Australian firm. This round values the company at approximately $13 billion, occurring just 18 months and four fundraising rounds into its growth phase.

      The funding was spearheaded by US investors Sands Capital and Wellington Management, with a noteworthy detail from Australia: Blackbird VC has reported its largest investment to date in this deal. Although the company is based in California, it was co-founded by Australians, adding a local dimension to Blackbird's financial interest.

      While specific investment figures from Blackbird have not been disclosed, reports suggest that its contribution may surpass all previous investments by an Australian VC firm, although this claim comes with necessary caution due to the lack of specific numbers.

      Baseten specializes in the foundational technology essential for the AI boom, offering software and infrastructure that enables businesses to customize and manage their own AI models, presenting a more affordable alternative to relying on providers like OpenAI and Anthropic. Rather than renting AI capabilities from advanced labs, clients utilize Baseten to deploy and manage models independently, which is where the ongoing costs of AI begin to accrue.

      The driving force behind Baseten's growth is inference, the phase where a trained model ceases to learn and starts generating outputs, such as answers, images, and predictions that customers are willing to pay for. The company reported a 20-fold increase in revenue over the past year, spurred by demand for this operational workload. As models transition from training to daily production use, the cost of running them at scale has evolved into a significant market, and a competitive one at that.

      The economic rationale for this surge in investment is clear. Training a model is a one-time, resource-intensive process, while inference constitutes an ongoing expense that increases with each query processed by a deployed model; thus, costs rise alongside adoption.

      For businesses employing AI in production, reducing the price of each inference call leads to considerable savings, which is the value proposition Baseten offers. This is also why investors are viewing infrastructure as a more stable investment, believing that the need for model serving will outlast the excitement surrounding any individual model.

      Baseten is vying for attention in a space that has attracted substantial capital this year. The same demand for more affordable, accelerated inference has fueled competitors focused on inference optimization, and new GPU-cloud startups have also garnered funding aimed at the same market. Some companies are even tackling the issue at the chip level, with at least one startup developing a specialized inference chip designed with Anthropic in mind. The common thread is that managing models, rather than merely creating them, is now where financial and engineering interests are aligning.

      This shift has altered the landscape of who is willing to invest, as infrastructure providers have become some of the most desirable names in the field, drawing sizable equity investments that were previously reserved for model development companies.

      Baseten's rise to a $13 billion valuation, based on inference revenue rather than a flagship chatbot, exemplifies this trend. For the startup, this is its fourth fundraising round in 18 months, a frequency that reflects both investor enthusiasm and the company's requirements.

      The upcoming challenge will be to maintain its 20-fold revenue growth as inference workloads expand and competition escalates, and whether the $13 billion valuation will prove insightful or excessive once rival companies publicly share their performance data.