Startup profile · funding coverage
Sparxity funding, valuation and investors
Builds sparse-native computing chips, boards and racks for HPC and future AI inference.
Verified company research · funding, investors and latest developments
Keep track of Sparxity
Save this profile to your VCT watchlist for a quick return.
Funding, valuation & investors
Answer-first snapshotLatest funding
Series A
Undisclosed · November 10, 2026
Latest known valuation
Not publicly disclosed
Total disclosed equity funding
Not publicly disclosed
Excludes debt, grants, acquisitions, secondaries, and IPO proceeds.
Current status
private
series a · Dresden, Germany
Investors in latest funding
Lead: ICM HPQC Fund
Other: Lockheed Martin Ventures , GlobalFoundries , GS Futures , Cloudberry Ventures , TGFS , EIC Fund
Overview
Sparxity, formerly SpiNNcloud and based in Dresden, announced a Series A on November 10, 2026 led by ICM HPQC Fund, with Lockheed Martin Ventures, GlobalFoundries, GS Futures, Cloudberry Ventures, TGFS, EIC Fund and other investors. The company's exact Series A proceeds were NOT disclosed: $90M is cumulative financing across its history.
Sparxity develops sparse-native chips and full-stack systems that skip unnecessary computation for event-driven HPC and sparse AI. Its own site reports three tape-outs, over 70,000 chips produced, more than 12 racks deployed and 30-plus customers, including a Jülich research deployment; these are vendor figures, not independently audited.
A Deutsche Telekom/T-Systems partnership is announced, but future deployment commitments must not be described as already completed. Claims of 10x faster sparse solvers, 18x lower energy for selected event-based ML and 20x shorter optimization runs are workload-specific. Up to 44x faster next-generation AI inference is a development target, not a general proven GPU comparison.
Source: https://www.businesswire.com/news/home/20261110433588/en/ ; https://www.sparxity.ai/
Why Sparxity is interesting
The November 10 Series A brought cumulative funding to $90M, but its standalone round amount was not disclosed. Unlike many chip concepts, Sparxity claims shipped silicon and installed racks; the 44x faster inference figure remains a next-generation target.
Product & use cases
A prospective HPC or cloud operator would identify workloads with meaningful sparsity, compare model structure and compute/data movement against existing GPU or CPU clusters, request Sparxity hardware/software integration details, run representative workload and power benchmarks on a pilot rack, and compare throughput, engineering effort and total cost before scaling. This is an illustrative enterprise procurement workflow, not VCT hands-on testing. Commercial list pricing, general availability and reproducible cross-vendor benchmark suites were not found.
- Sparse linear solvers and event-based high-performance computing
- Energy-efficient specialized AI inference for fine-grained dynamic sparsity (next-generation goal)
- Research supercomputing installations such as the company-reported Jülich system
Key facts
- Series A announced Nov 10; ICM HPQC Fund lead; standalone round amount undisclosed
- $90M is total funding to date, NOT the Series A amount
- Vendor reports 3 tape-outs, 70,000+ chips, 12+ racks and 30+ customers
- Current selected workload benchmarks are company-reported and not universal
- Up to 44x faster inference is a goal for next-generation silicon, not a demonstrated all-model result
Funding history (newest first)
Series A
2026-11-10 Undisclosed- ICM HPQC Fund (lead)
- Lockheed Martin Ventures (participant)
- GlobalFoundries (participant)
- GS Futures (participant)
- Cloudberry Ventures (participant)
- TGFS (participant)
- EIC Fund (participant)
Source: Read source · businesswire.com
Competitive landscape
Edge: Full-stack sparse computing with some existing silicon and racks provides evidence of hardware engineering maturity. Commercial performance versus mainstream accelerators remains workload-dependent and needs reproducible benchmarks.
Sparxity's historical tape-outs and deployed racks should be separated from the announced future dynamic-sparsity chip. Published vendor comparisons do not establish standardized throughput/$ or joules/token across equivalent models. The Series A's undisclosed size prevents reliable post-money or cash-runway math.
-
NVIDIA GPU platforms incumbent
Broad AI software and deployment ecosystem; sparse workloads may still incur routing overhead.
-
Cerebras adjacent
Specialized AI hardware and systems with a different architectural emphasis.
-
Groq adjacent
Dedicated low-latency inference hardware, not identical sparse-native architecture.
-
Conventional HPC CPU clusters alternative
Mature general-purpose systems for sparse solvers, with different energy and development trade-offs.
Market / IPO context
Editorial / static context — not a live quote.
FAQs about Sparxity
Practical answers founders, operators, and investors typically search for.
Last updated:
Editorial note: AI tools assisted with research, structure, or drafting. Venture Capital Tracker retains human editorial responsibility for factual accuracy, relevance, and source quality before publication.