- Sparxity, formerly SpiNNcloud, has completed its Series A round and now has $90 million in total funding.
- ICM HPQC Fund led the Series A, with Lockheed Martin Ventures and GlobalFoundries also participating.
- Sparxity and Deutsche Telekom have signed a Letter of Intent, and the first deployment is due in 2028.
Hector Gonzalez has spent his career designing energy-efficient chips, starting with biomedical devices, then moving to automotive applications, and now focusing on data centres.
“The common thread is chip design. The first one was biomedical applications… the next one was low-power chips for automotive, and the next one is low-power chips for data centres,” he told Tech Funding News.
Sparxity, formerly SpiNNcloud, has closed its Series A round, bringing total funding to $90 million. ICM HPQC Fund led the round, with participation from Lockheed Martin Ventures, GlobalFoundries, GS Futures, Cloudberry Ventures, TGFS, and the European Innovation Council Fund.
Angel investors in the round include Richard Socher, co-founder of Recursive Superintelligence, and Nancy Xu, whose startup, Moonhub, joined Salesforce in 2025. Existing investors Atlantic and Silicon Catalyst are also participating.
“The next era of AI will be defined not just by how much computing power we can deliver, but by how intelligently and efficiently we use it. This financing is a powerful validation of our vision for sparse-native computing and the complementary positioning of our systems. With the backing of leading, strategic investors, we’re accelerating the transition from dense, energy-intensive architectures to heterogeneous datacenters with dedicated silicon for ultra-sparse workloads to unlock greater AI performance and lower costs,” said Gonzalez, co-founder and CEO of Sparxity.
From brain-inspired to sparse computing
Gonzalez studied electronics and then joined a joint MIT/Masdar Institute programme in Abu Dhabi. In 2021 he co-founded SpiNNcloud as a spin-out from Prof Christian Mayr’s group at TU Dresden, with Christian Eichhorn and Matthias Lohrmann.
The SpiNNaker2 chip builds on Steve Furber’s work, as he co-designed the original ARM processor. GlobalFoundries did the chip’s first tapeout with the Dresden team in 2021, and that’s when the hardware collaboration began.
Most of the early funding came from public sources: a €2.5 million EIC Transition grant in February 2023 and €10 million in combined grant and equity from the EIC Accelerator, announced in June 2025.
Gonzalez says the original market was very niche: “That was the original market, where the technology started, as it was so sparse in time and space with the neuromorphic models. We then went on to develop the technology further for a wider value proposition and repositioned it into the mainstream AI ecosystem.”
He is very clear about competitors who have stayed firmly in the biology camp, with Intel’s neuromorphic system being “way too biologically inspired.”
Sparse computing means skipping the irrelevant calculations in a model. Many current AI models already do this by activating only a small subset of parameters for each input. Gonzalez cites two examples: Google DeepMind’s PEER, which uses a million experts, and Meta’s large-scale memory layers.
Gonzalez says Sparxity isn’t in competition with Nvidia: “Nvidia systems are absolutely excellent and they’re excellent for dense workloads. That’s where all the machine learning workloads started.”
For him, the competition is Moffett AI, a Shenzhen company founded in Silicon Valley in 2018 that makes sparse computing processors. “They’re effectively the first mover in this category of hardware which is still very much open,” Gonzalez adds.
Public funding helped build Sparxity’s track record
Sparxity says it has completed 3 chip designs, made over 70k chips, has 12 racks in place, and has signed over 30 customers. The customers are in the US and Europe, and revenue has increased 4x from 2024 to 2025.
So far, all of Sparxity’s named deployments are at research centres or universities: Sandia National Labs in the US, Leipzig University, the University of Texas at San Antonio on a US NSF project and now the Jülich Supercomputing Centre, which hosts Europe’s first exascale machine. The Jülich machine is JUBIS, and it’s funded by the European Regional Development Fund and the state of North Rhine-Westphalia.
Sparxity has its own performance figures: 10x faster sparse solvers, 18x lower energy use for event-based machine learning, and 20x shorter runtimes on optimisation problems. The primary aim of the new funding is to get up to 44x faster inference on ultra-sparse models with the next-gen chip, which is still in development.
Sparxity now has 72 employees and is hiring more. The staff are based in Germany, France, the UK, New Zealand and the US. The company’s head office is in Dresden, with a branch in San Francisco. Gonzalez splits his time between the two offices.
Market overview and what comes next
Even with all its grants, Sparxity’s total is still less than what some inference-chip startups have raised in a single round. Axelera AI raised over $250 million in February. Fractile closed a $220 million Series B in May. Euclyd raised over €200 million in a Series A on 15th September. Naveen Rao’s Unconventional AI, which also makes brain-inspired chips, raised a $475 million seed round.
Some companies reduce the cost of running AI models by moving memory closer to the processor or running inference at the edge. Sparxity takes a different tack by making models sparser and providing hardware designed for that. The market is already in motion, with the AI inference market set to grow from $106.15bn in 2025 to $254.98bn by 2030.
Gonzalez said the funding will fund the launch of the second product, support a new tapeout for sparse AI workloads, and cover production to meet current orders.
Sparxity’s success depends on model builders adopting sparsity faster than GPU makers can adapt, and on developing a new chip still in progress. Series A funding will help create this chip. If sparse models become the norm, Sparxity could gain a first-mover advantage, with only Moffett in China as a direct competitor.
