German semiconductor startup Sparxity has closed a Series A financing round, bringing its cumulative funding to $90 million as it develops specialised processors designed to make artificial intelligence workloads faster and more energy-efficient.
The Dresden-based company announced the financing on November 10, positioning its technology around sparse computing, an approach that avoids processing unnecessary information during AI calculations.
The company did not disclose the amount raised specifically in its latest Series A, according to The Next Web’s reporting.
The investment highlights a growing opportunity for European semiconductor startups: improving how efficiently AI models run, rather than simply expanding the computing capacity available to them.
Investors Back a Different Approach to AI Computing
The financing was led by ICM HPQC Fund, with participation from Lockheed Martin Ventures, GlobalFoundries, GS Futures, Cloudberry Ventures, TGFS and the European Innovation Council Fund.
Angel investors included Richard Socher, founder of Recursive Superintelligence, and Nancy Xu, whose startup Moonhub was acquired by Salesforce. Existing backers Atlantic and Silicon Catalyst also continued supporting Sparxity.
Founded in 2021 as SpiNNcloud, a spin-off from TU Dresden, the company developed technology based on the brain-inspired SpiNNaker2 processor before rebranding as Sparxity in October 2026.
Its current strategy centres on dynamic sparsity, which allows computing systems to activate only the operations needed for a particular calculation.
Unlike conventional approaches that allocate substantial processing resources across a model, sparse computing can reduce unnecessary operations and data movement.
Sparxity says its next-generation chips are targeting up to 44 times faster AI inference in ultra-sparse models.
That figure represents a development target for particular model architectures, not an independently established speed improvement across all AI workloads.
Commercial Deployments Strengthen the Investment Case
Sparxity is not relying exclusively on future chip designs.
According to its financing announcement, the company has completed three chip tapeouts, manufactured more than 70,000 chips and deployed 12 computing racks. It reports serving more than 30 customers worldwide.
The startup also said revenue quadrupled between 2024 and 2025, although it did not disclose the underlying revenue figures.
Its reported results on selected existing workloads include 10-fold faster sparse solvers, 18-fold lower energy consumption for event-based machine learning and 20-fold shorter optimisation runtimes.
Sparxity has secured a multimillion-dollar deployment at Germany’s Jülich Supercomputing Centre, home to Europe’s first exascale supercomputer. It has also signed a letter of intent with Deutsche Telekom’s T-Systems to introduce sparse-native technology into its data-centre infrastructure.
Co-founder and CEO Hector Gonzalez argues that the next phase of AI infrastructure will depend on using computing resources more intelligently, not merely increasing capacity.
Europe’s AI Infrastructure Opportunity
The financing reflects a wider challenge facing AI developers and cloud operators: higher performance does not automatically translate into economically sustainable inference.
Sparxity’s proposition is that specialised hardware can reduce the resources consumed by workloads that activate only portions of a model.
For European technology companies, that creates an opportunity to compete in a narrower but potentially valuable segment of the AI hardware market.
However, its success will depend on whether developers adopt sparse architectures widely enough to justify dedicated processors, and whether performance improvements translate into lower production costs.
The investment therefore represents more than another European semiconductor funding announcement. It is a bet that the future AI infrastructure market will reward efficiency and architectural specialisation alongside raw computing power.
