Compute access · Fundamental research
Pleroma Works is building a distribution business designed to lower the delivered price of local AI systems—and publish the evidence behind every savings claim.
One governed customer view
Before launch, the same card shows what remains unverified instead of manufacturing a savings claim. Once an offer and benchmark clear their evidence gates, this becomes the single authoritative customer-facing Dividend record.
Capability Dividend
NVIDIA DGX Spark · GB10 · 128GB · 4TB
Observed market reference
$4,699
Pleroma price
Pending supplier economics
Capability Dividend
Pending verified benchmark
No savings amount or percentage is claimed before the benchmark and offer are verified.
Captured Aug 26, 2026 · Out of stock at capture
NVIDIA MarketplaceS1 reference class
Your model files stay local. There is no mandatory cloud token bill. Standard tools remain available, and Pleroma services stay optional.
128GB
Coherent or unified memory class
4TB
Local model-storage target
Linux
Standard NVIDIA and development tooling
You own it
Optional services, not forced dependence
Open operating hypothesis
Lower relative price may increase adoption; adoption may create purchasing scale; scale may reduce structural cost. Those causal links are not treated as established results.
H1
Publish the customer price and its reference.
H2
Observe whether more customers choose local ownership.
H3
Measure realized volume and supplier economics.
H4
Test whether delivery becomes sustainably cheaper.
Commercial test
Revenuet+1 > Revenuet
Customer-value test
$/Capabilityt+1 < $/Capabilityt
One company-level method
Both publish the boundary between what is established and what is not. The shared link is method—not a claim that Thermogeonesis already powers S1.
Flagship research program
Foundational work on information, physical events, records, states, geometry, and computation. Every claim carries an explicit evidence status.
Product, benchmark, research, and company updates are dated so suppliers, customers, lenders, and future collaborators can see what changed and when.