The most expensive part of running an AI model isn't the software. It's the electricity. That simple, physical reality is the wedge for Valu, a London-based startup quietly building a search and retrieval tool for AI infrastructure [Valu website, retrieved 2024]. Its pitch is direct: help data centers get more capacity and lower the cost per watt for GPU inference [Valu website, retrieved 2024]. In an industry racing for more chips, Valu is betting the real race is for the cheapest kilowatt-hour.
The Search for Efficiency
Valu's product remains in an invitation-only phase, a common tactic for deep-tech firms targeting complex enterprise buyers [Valu website, retrieved 2024]. The core concept, however, is a market-making one. It aims to function as a discovery layer, helping users locate and access underutilized or more power-efficient GPU capacity across data centers. For operators drowning in power bills and for AI teams with constrained budgets, the value proposition is operational margin. If a model can be run for 10% less per query by finding a more efficient location, the savings at scale are material.
The Team and the Ticket
The company is led by co-founders Hirsh Pithadia, CEO, and Alexander Ng, who heads AI [LinkedIn, retrieved 2024]. They founded the company in 2023 and have since secured a $4.57 million Series A round [PitchBook, retrieved 2026]. The investor list includes Japanese venture firm Global Brain Corporation and early-stage fund Comma Capital, with angel investor Kotaro Chiba also participating [PitchBook, retrieved 2026]. The round is a standard ticket for a capital-intensive infrastructure play, giving the team runway to build and validate its technology with early design partners.
| Founder | Role | Source |
|---|---|---|
| Hirsh Pithadia | Co-Founder & CEO | [LinkedIn, retrieved 2024] |
| Alexander Ng | Co-Founder & Head of AI | [LinkedIn, retrieved 2026] |
Where the Wheels Could Come Off
The bet is clear, but the path is lined with execution risks. Valu is entering a field where hyperscalers like AWS, Google, and Microsoft are aggressively optimizing their own data centers and offering discounted spot instances. Competing on pure price discovery against their scale and integrated stacks is a formidable challenge. Furthermore, the product's success hinges on creating a dense, liquid network of supply,data center operators willing to list spare capacity,and demand from AI teams. Building that two-sided marketplace from zero is a classic chicken-and-egg problem.
The company's technical differentiation also remains unproven in public. While the concept of a search layer is sound, its accuracy, latency, and the breadth of its sourcing will determine if it offers a meaningful advantage over existing brokerages or direct procurement. The invitation-only model suggests a focus on high-touch, early validation, but it delays the public traction metrics that typically build market confidence.
The Next Twelve Months
For Valu, the coming year is about moving from concept to contracted capacity. The key signals to watch will be a transition out of invitation-only access, the announcement of initial data center partners, and any published metrics on cost savings for early users. The $4.57 million from Global Brain and Comma Capital provides a foundation, but the real test is whether AI engineering teams will change their procurement workflow for the promise of a lower power bill.
The round values the company at an undisclosed level, but the investor mix suggests a thesis around cross-border efficiency gains, linking Asian capital with European and North American infrastructure. For data center operators staring down flatlining power utilization effectiveness (PUE) ratios, the question is whether a third-party tool can find optimizations their own teams cannot. If Valu's search engine can answer that affirmatively, the cost per watt for the entire AI stack might just have a new variable.
Sources
- [Valu website, retrieved 2024] Valu homepage | https://getvalu.io/
- [LinkedIn, retrieved 2024] Valyu | LinkedIn | https://www.linkedin.com/company/valyu-ai
- [PitchBook, retrieved 2026] Valu Company Profile | https://pitchbook.com/profiles/company/264888-10
- [LinkedIn, retrieved 2026] Alexander Ng - Founder & Head of AI at Valyu | https://uk.linkedin.com/in/alexander-ng-yew-shen
- [Encord, retrieved 2026] AI Data Chats article | https://encord.com/ai-data-chats/from-context-stuffing-to-context-grounding-for-ai-systems/
- [LinkedIn, retrieved 2026] David Ongchoco - Comma Capital | LinkedIn | https://www.linkedin.com/in/davidongchoco