Announcing Our $10.5M Series A to Power the Future of Decentralized AI Infrastructure

At Manifold, we believe the future of AI infrastructure will not be owned by a few centralized labs, but rather powered by a network of independent and economically aligned contributors. Built on Bittensor, Targon represents a fundamentally new category: a decentralized, verifiably confidential, and trustless AI cloud.

Targon is currently serving over 20 billion paid inference tokens a day, backed by 1,500+ H200s and growing. More than $60 million in annual compute incentives now flow through the Targon network.

Today, we're proud to announce that Manifold has raised a $10.5M Series A, led by OSS Capital, with participation from:

• Digital Currency Group
• Tobi Lutke (Co-Founder and CEO, Shopify)
• Ram Shriram (Founding board member (longest standing) and angel investor in Google)
• Zachary Smith (Former co-founder of Packet, acquired by Equinix, board of Servers.com)
• Jacob Steeves (Co-founder of Bittensor, and Founder of Affine)
• Ala Shaabana (Co-founder of Bittensor and co-founder of Crucible Labs)
• Logan Kilpatrick (Group PM of Google DeepMind AI)

The Problem with Centralized AI Infrastructure

We believe compute is the defining commodity of the AI era, no different from the importance of electricity during the Second Industrial Revolution. Yet today's market is broken. Most infrastructure providers are heavily VC subsidized, capital inefficient, and increasingly reliant on opaque financing arrangements. This is not a scalable foundation for the future of AI.

Targon is playing an important role in reducing many of the inefficiencies that exist in today's market. We incentivize a globally distributed network of compute operators and compensate them via a pool of 24/7, 365 distributed incentives. In the same way Uber subsidized driver supply in the early days to attract rider demand, Targon is subsidizing compute supply to attract inference demand. Over the long run, we will not need the subsidy as we will have created the world's deepest and most efficient compute market, with many of the world's biggest companies plugged in.

What Comes Next

Manifold is focused on creating a commercially sustainable cloud that enterprises can rely on. As we continue to scale supply, the focus will be on increasing compute efficiency (via Targon Inference Engine), while simultaneously bringing in more enterprise demand. Today, Targon uses Nvidia PPCIE and Confidential Compute to run a CVM on the compute operator's machine. In the coming weeks Targon will become even more secure. We will be releasing a new upgrade that incorporates Intel TDX, which ensures all CVMs are fully encrypted - and most importantly, that all enterprise data remains confidential and secure, without relying on trust in the operator or host environment.

Currently, Targon has a market for on-demand H200s. Soon, we will be implementing collateralized compute markets which can be applied to any type of hardware (i.e., B200, H100, L40s). Operators will post collateral to commit to multi-month compute contracts (i.e., 8 H200s for 6 months), paired with uptime and performance guarantees. This structure creates durable supply commitments in a trustless manner, and ensures participants in the Targon market are getting best-in-class compute at best-in-class pricing.

Conclusion

To builders, enterprises, and researchers: if you're ready to adopt AI infrastructure that is secure, verifiable, and economically grounded, we invite you to join us.

The future of AI will not be gated. It will be incentivized.