Baseten, a startup providing infrastructure to run AI applications in production, has closed a $1.5 billion Series F financing round at valuations of up to $13 billion, marking a 160 percent jump from the $5 billion valuation it commanded just five months earlier in its Series E round. The round was led by Altimeter Capital, Conviction and Spark Capital, with Sands Capital and Wellington Management as co-leads, alongside a roster of existing backers including IVP, Greylock, Battery Ventures and D. E. Shaw Ventures — underscoring the intensity of investor appetite for companies positioned at the “inference” layer of the AI stack, where trained models are actually run to serve live customer traffic.
The company disclosed that its platform now processes more than one billion inference requests daily, spread across 87 compute clusters operating on 18 different cloud environments worldwide, and that revenue has grown roughly twenty-fold year-over-year. That scale of growth places Baseten among the fastest-expanding infrastructure companies in the current AI cycle, and reflects a broader shift in enterprise AI spending: as more companies move from experimenting with large language models to deploying them at scale in production applications, the computational and engineering burden of serving those models cost-effectively and reliably has become a distinct, high-value market of its own, separate from the foundation model training race dominated by OpenAI, Anthropic, Google and Meta.
Baseten’s rapid re-rating comes amid an extraordinary period for AI infrastructure fundraising more broadly. The same week saw Shield AI close more than $1.5 billion as part of a $2.25 billion package at a $12.7 billion valuation, and Crusoe raise $1.38 billion in Series E funding at a $10 billion valuation, part of what Crunchbase described as a “spree of megadeals” driving venture funding to record levels. Globally, AI-related startup funding pushed total venture investment to roughly $300 billion in the first quarter of 2026 alone, with US-based companies capturing an estimated 88 percent of that capital — a concentration that continues to raise questions about the geographic imbalance of the AI investment boom even as global demand for AI infrastructure, including in India, expands rapidly.
The scale of Baseten’s raise and the speed of its valuation increase highlight how investors are increasingly differentiating between “model layer” companies, whose long-term defensibility remains contested as foundation models converge in capability, and “infrastructure layer” companies like Baseten that profit from usage growth regardless of which specific model ultimately wins in the market. Baseten’s own commentary pointed to this dynamic directly, noting that demand for inference at the application layer is surging precisely because closed-source and open-source models are converging in cost, capability and customisation options, making the deployment and serving layer — rather than model selection alone — the critical bottleneck for enterprises operationalising AI.
Baseten said it plans to triple its headcount over the course of 2026 to support growth across engineering, research, operations and commercial functions, a hiring pace that mirrors the aggressive scaling seen at other infrastructure-focused AI companies riding similar demand curves. The company’s rise also illustrates how quickly capital is now moving in the sector: the jump from a $5 billion to a $13 billion valuation in under five months represents one of the fastest markups among venture-backed AI infrastructure companies this year.
For enterprise technology buyers and India-based Global Capability Centres increasingly building AI-driven products, Baseten’s growth signals that the market for reliable, cost-efficient model-serving infrastructure is maturing rapidly alongside the foundation models themselves. With inference costs now a material line item for any company running AI at scale, investors are betting that companies solving the deployment and reliability problem — rather than those building models from scratch — will capture a disproportionate share of value as the broader AI infrastructure buildout continues through the rest of 2026 and beyond.
Leave a comment