On-Demand GPU Startup Andromeda Raises Funding at $1.5B Valuation

On-demand GPU startup Andromeda has closed funding at a $1.5B valuation. The marketplace model aggregates idle GPU capacity from 100+ providers, allowing AI companies to rent compute without long-term contracts. ARR grew from $50M to $100M between 2024-2025, profitable since launch.

Andromeda Raises Funding at $1.5B Valuation: Why GPU-as-a-Service Is Booming

The Core Event

On-demand GPU rental startup Andromeda AI Inc. has closed a new funding round at a $1.5 billion valuation. The capital was provided by Paradigm, a venture capital firm focused on crypto and emerging technologies, which has reportedly invested a total of $60 million in the startup to date.

Andromeda's Business Model

Andromeda operates a GPU compute marketplace that helps companies rent unused GPU resources from the data centers of internet providers, cryptocurrency mine operators, and various other organizations. Unlike traditional cloud providers (AWS, Azure, GCP), Andromeda allows customers to rent on-demand without signing multi-year contracts.

For AI startups that only need short-term GPU compute — for example, training a small language model might only require a GPU cluster for a few days or weeks — this flexibility is extremely attractive.

Core Platform Capabilities

Hardware Quality Auditing: Before making a provider's infrastructure available, Andromeda checks that hardware meets performance and security requirements. This includes evaluating not only GPUs but also supporting equipment such as storage arrays. Malfunctioning drives and network issues can interrupt AI training runs.

Unified Procurement and Billing: Customers can purchase capacity from multiple providers through a single interface, with hardware bills consolidated into a single invoice.

Andromeda Pricing Index: Provides real-time data about current GPU prices with region-specific variations, helping customers negotiate better rates from infrastructure providers.

Operations Monitoring: Andromeda's engineers continuously monitor infrastructure for reliability issues and make adjustments when necessary.

Market Metrics

Andromeda currently provides access to infrastructure from more than 100 providers and has processed over 1,000 GPU transactions since launching about two years ago. Its annualized revenue run rate grew from $50 million in 2024 to $100 million in 2025, and the company has operated profitably since launch.

Its customer base reportedly includes multiple "notable AI startups and labs" that typically spend $250 million to $500 million per year on infrastructure.

Competitive Landscape

The GPU-as-a-Service space is heating up rapidly, with major players including:

  • **CoreWeave**: Valued at tens of billions, the benchmark company in this space, focused on large-scale GPU cloud services
  • **Lambda**: Offers GPU cloud and on-premise AI clusters targeting research institutions and enterprises
  • **Together AI**: Focuses on inference services and open-source model hosting
  • **Vast.ai / RunPod**: Low-cost GPU rental for individual developers and small teams

Andromeda's differentiation is its "GPU marketplace" model — rather than building its own data centers, it aggregates third-party idle compute capacity, similar to an Airbnb for GPUs.

Industry Trends

The $1.5 billion valuation reflects several deep trends:

1. **GPU compute is the oil of the AI era**: February 2026 saw a record $189 billion in global VC funding, with a substantial portion flowing to AI infrastructure

2. **Long-term contracts don't work for everyone**: Many small and mid-size AI companies don't need and can't afford multi-year GPU contracts

3. **GPU utilization problem**: As crypto mining has cooled, significant GPU capacity sits idle and needs new allocation mechanisms

4. **Profitability validation**: Andromeda has been profitable since launch, proving the viability of this business model

For AI entrepreneurs, the rise of platforms like Andromeda means the barrier to accessing GPU compute is lowering. However, performance consistency, network latency, and security of third-party GPUs remain risk factors that need careful evaluation.