Hippius (SN75)

In short

• The price advantage is real and narrow. On storage alone Hippius is 14% below Backblaze and 60% below Cloudflare R2 at list price. On a download-heavy workload it is far cheaper than AWS and Backblaze, and $18 a month cheaper than R2, which also gives downloads away.
• The low price is paid for mainly by emission. Miners receive about $5.4M a year of new alpha; customers paid about $27K a year. No customer money is routed to miners today. Free downloads also rest on miners' flat-rate internet links and on the chain's bandwidth price being set to zero, which the team can change.
• Nine of the ten organisations named as users are Bittensor subnets or ecosystem firms. The one external company is Zapme Social, a 6,000-member community app. One subnet user, Vidaio, lists the Hippius co-founder as an advisor.
• The owner account sells most of its emission. It receives $2.4M a year and sold the TAO equivalent of 65% of that in the 30 days to 11 Oct. A public buyback wallet is claimed; no address has been published. A permanent lock pallet (Engram) is announced and not in the code.
• Market size does not constrain Hippius; winning customers does. Backblaze's B2 storage line alone runs at $113M a year. The closest decentralised competitor, Storj, reached $13M a year after eight years and its company filed for Chapter 11 in July 2026.

The claims Hippius is known for, and what the checks show

The headline claims are accurate at list price for the workload they describe. Each depends on the buyer's download pattern, and none speaks to how much paid demand exists.

Hippius sells three storage products and has opened a compute beta

Demand found so far is inside the Bittensor network. That is cheap to win, since subnets already hold TAO and know the team, and it is funded by those subnets' own emission. External routes the team is pursuing are AI model and dataset hosting, migration offers aimed at AWS and at Storj customers after Storj's bankruptcy, and consumer Drive; none has a named paying external customer above Zapme's size yet.

Who earns what: all participant pay is new alpha; the company also collects about $27K a year from customers

Customers pay the company for credits, and the credits are used up as data is stored. Miners are paid from emission whatever customers buy. The co-founder put the team's costs at about $25K a month, funded by selling 60–150 TAO (post, 7 Aug); customer purchases cover roughly a tenth of that.

Usage is growing faster than paid demand

Storage grew 3.2× in three months while customer purchases grew 2.3× in 30 days from a $12K-a-year base. Billing per terabyte reconciles once the 3× piece overhead is removed: the last week's billing ran at $24.5K a year, about $7 per TB of customer data a month against the $6 list price (our estimate, since Hippius's per-account totals were not read). A further $23K of usage in 8–15 September was paid with credits the team issued itself; it is shown on the dashboard and excluded above.

Quality: no outage on record, no service-level promise, and no independent speed test yet

Hippius is unrated against its competitors rather than rated and losing. The test that would settle it is a sustained multi-terabyte download benchmark against R2 and Backblaze from several regions.

Against its competitors, Hippius is the cheapest at list price, and the free-download niche is already occupied

The edge: the lowest list price with no download meter, which matters most for files read many times. Every machine that runs a model downloads its weights, again on each restart or update, and every training node pulls the dataset for each run, so a model or dataset is downloaded many times over. Minos's genome fan-out to every miner is that workload.

The offsetting weakness: the price is financed by emission, and the niche is contested by firms with balance sheets. AWS's download fee is mostly margin: Cloudflare estimated it at about 80 times wholesale transit cost (Cloudflare, 2021), which is why R2 can give downloads away profitably. Hippius gives them away because miners are paid from emission and rent flat-rate 1 Gbps links. Against R2 the saving is $9 per TB stored; against Backblaze it is $0.95. Durability promises, support and speed under load are where those competitors are ahead.

The market is large; disclosed revenue in this layer starts at about $131M a year

Market size is not the binding constraint. Hippius's purchases are about 0.02% of the disclosed floor. The constraint is turning a list-price lead into paying customers outside Bittensor, which the decentralised precedents show is slow: Storj took eight years and about $35M to reach $13M a year. Against that, Hippius has issued about $23.7M of alpha in 19 months, a financing cost comparable to Storj's whole raise.

How the miner contest works: one Hippius account takes 99% of miner emission and pays storage operators on its own chain

Miners withdraw about half of what the bridge receives. Of 271K α deposited since August, 134K has been withdrawn back to Bittensor, rising to 80% of deposits in September (team ledger, hipstats.com); the two largest withdrawers no longer hold SN75. This is the miner sell flow.

Team, openness and plans

Compute matters most to the case: it is the one plan that adds a product with its own revenue line, and the co-founder links the move to full miner payouts to it.

The token: owner emission runs at $2.4M a year, about two-thirds is sold, and no buyback or permanent lock exists yet

Net insider flow is the full owner cut, $6,500 a day, since no buyback is verified; realised net sales are about $4,000 a day (11 Oct). The token's TAO inflow was $5,200 a day on 6 Oct (1.52% of all TAO sent to subnets, 21st of 90). Holding SN75 is a claim on the emission schedule and on any future mechanism that routes Hippius's revenue into alpha; at $27K a year of customer purchases, that revenue would not move the price even if all of it bought alpha.

What would change our mind

Open items

• Who placed 82% of the owner pledge, and who the 43 accounts withdrawing from the bridge are.
• Customer data held per account on the Hippius chain (the read failed); the 280 TB figure is an inference from the 3× piece overhead.
• Whether the 600-group cap is closing the network to new storage operators; ask the team.
• The team's "almost 6 figs" revenue, its $46,600 figure and the claimed buyback wallet: ask for definitions and the address.
• A clean download-speed test against R2 and Backblaze; this run's connection failed for every provider.
• The team's X posts between 9 Jul and 3 Aug and after 3 Oct were not found by search; they may exist.