Sparxity Closes Series A, Reaching $90M Total Funding for Sparse AI Chips

Sparxity has closed a Series A round that lifts total funding secured by its team to $90 million. ICM HPQC Fund led the round. Lockheed Martin Ventures, GlobalFoundries, GS Futures, Cloudberry Ventures, TGFS and the EIC Fund joined, alongside angel investments from Richard Socher and Nancy Xu.

The company, formerly known as SpiNNcloud, builds sparse-native computing systems. It plans to spend the new capital on chips for fine-grained dynamic sparsity. The target: up to 44 times faster AI inference in ultra-sparse models, with greater energy efficiency than today's hardware delivers.

What does sparsity buy in practice?

Most AI chips compute everything, dense and indiscriminate. Sparxity's full-stack systems compute only what matters. That is the core mechanism. Sparse models carry large fractions of zeros or irrelevant activations, and dedicated silicon can skip those portions instead of processing them.

Co-founder and CEO Hector Gonzalez framed the round as a vote on where computing goes next. He said the next era of AI will be defined by how efficiently computing power is used, and described the financing as validation of a shift from dense, energy-intensive architectures to datacentres with dedicated silicon for sparse workloads.

Xu Owen He, a former Google DeepMind researcher and strategic advisor to Sparxity, pointed at the constraint holding the field back. He said current hardware limits how far dynamic sparsity can go. In other words, the algorithmic techniques exist; the silicon does not yet match them.

How much traction does the company already have?

The funding rests on shipped product, not just a thesis. Sparxity quadrupled revenue between 2024 and 2025. It has completed three tapeouts, produced more than 70,000 chips and serves over 30 customers.

That manufacturing record mattered to the lead investor. Matthew Gould of ICM HPQC Fund cited Sparxity's proven manufacturing and deployments as the basis for the investment.

Two deployments stand out. The company's hardware runs at the Jülich Supercomputing Centre. It also powers work through a Deutsche Telekom partnership, which T-Systems CEO Ferri Abolhassan said will bring energy-efficient AI computing to its AI factory.

Why are defense and foundry investors in the round?

The investor list reads like a hardware supply chain assembled end to end. Lockheed Martin Ventures brings defense-sector demand for efficient edge computing. GlobalFoundries brings manufacturing depth. The EIC Fund brings European public backing. GS Futures, Cloudberry Ventures and TGFS round out the syndicate.

For GlobalFoundries, the bet is straightforward: sparse-native chips need a foundry partner able to produce them at volume. Sparxity's 70,000-plus chips already produced suggest that relationship has moved past the prototype stage.

What comes next?

The money goes into silicon development. Sparxity aims to turn fine-grained dynamic sparsity into shipping chips that deliver the 44x inference speedup target in ultra-sparse models while cutting energy draw.

The bet ahead is whether datacentre operators move from dense accelerators to mixed fleets with dedicated sparse silicon, as Gonzalez predicts. With revenue already quadrupled year over year and deployments at Jülich and Deutsche Telekom's AI factory in place, Sparxity now has the capital to test that thesis at chip scale.