Skip to content

VVV

6 Apr 20268 min

Venice is a privacy-first AI platform that provides uncensorable access to leading open-source and frontier models without storing prompts or responses. Recently, they reached over 2M users and have processed over 1M daily API requests across integrations with OpenRouter, Brave Leo, Cursor, and VSCode among others.

Inference already accounts for roughly two-thirds of AI compute in 2026, and McKinsey projects inference power demand growing at 35% CAGR to 93 GW by 2030, materially outpacing training at 22%. Venice doesn't need to prove AI demand is large, or that privacy matters. It needs a slice of recurring users who would rather own capacity over renting it, and the onchain data shows this conversion is already happening.

Our view is that the market still treats VVV as a generic token in the AI crypto narrative. Such categorization overlooks an intrinsic mechanism design that increases the value of VVV as Venice’s total AI capacity growth and future DIEM issuance start scaling. We think we are getting closer to that inflection point, making the timing of the opportunity particularly appealing from a risk-reward perspective.

Most AI products are still sold the same way: pay by subscription if you are a consumer, or pay by usage if you are a developer. On Venice, there are 3 ways to pay for AI models: i) a similar subscription model going for $180/year for Venice Pro, ii) USD deposits in exchange of token credits or pay-per-request (like any other API), or iii) use Venice’s token system (VVV and DIEM).

VVV is the base token. It is staked in return of sVVV, which earns emission yield (currently 18% APR) and grants Pro-level platform access when you have at least 100 VVV staked. Holding sVVV is the only way to mint DIEM, Venice’s fixed compute token explained below. There is a 7-day cooldown period to unstake sVVV back into liquid VVV.

DIEM is the fixed position token, turning AI inference capacity into a tradeable unit. Each DIEM, once staked, produces a standardized $1/day API credit that never changes regardless of how many other people stake or how much capacity Venice adds. DIEM supply is capped by Venice's actual compute capacity (current target 38,000) and backed 1:1 by locked sVVV. You lock sVVV to mint DIEM and burn DIEM to unlock sVVV at any time. Think of DIEM less like a token and more like buying a permanent API subscription at today's price.

There are two ways to think about these as investable assets. Suppose a user who wants fixed compute at today's prices. They would look at DIEM, which has traded around the mid-$700s to $800s, and purchase. Once staked, each DIEM produces $1/day in API credits. Thus, a buyer breaks even on the purchase cost in roughly 2 years. This break-even point, helps gauge when cumulative demand and usage can fully cover the purchase price, or effectively yield ~45% annually in compute credits. Besides, this isn’t a sunk cost. Even after this period of time passes, the asset is still owned by the user. Assuming price stays flat, usage during 3 years plus a sale would mean that they essentially got 3 years of compute for free.

Ending Q1 2026, 77% of all DIEM is currently staked, which tells us that most buyers are activating it for compute credits, rather than holding it to trade or LP.

Altogether, DIEM behaves more like a perpetual API subscription with a resale value than a speculative token. In a rational market, its price should stay anchored to whatever breakeven period users are willing to accept on $1/day of API credit.

The speculative upside, however, plays in VVV’s favor, as it is the claim on future capacity expansion. When Venice expands capacity and raises the DIEM target supply, new DIEM enters circulation only by minting through locked sVVV.

Current market dynamics show that the minting cost has climbed to roughly $3,800 worth of sVVV per DIEM (on an exponential mint curve), while DIEM trades at ~$800 on secondary markets.

That ~5x gap tells us two things: i) capacity is effectively full, and ii) the mint curve is disincentivizing new supply at the current target of 38,000 DIEM. Anyone who wants fixed compute today would rationally buy existing DIEM on secondary markets like Aerodrome rather than minting fresh supply through VVV.

The VVV bid shows up in a different regime. When Venice adds enough real capacity to raise the target, the mint curve resets and new DIEM becomes viable to create again. At that point, the only way to produce the next unit of fixed compute is to lock sVVV, and the demand for that locking flows directly into VVV. In other words, VVV is not bought when people merely want compute, but when the market needs more compute issuance, which can only be created upstream through sVVV.

DIEM repricing upward from yield compression has a second-order effect on VVV. As the market price of DIEM climbs toward the mint cost, the spread compresses and minting becomes economical again. That is what reactivates sVVV locking and turns DIEM demand into VVV demand.

That spread between mint cost and market price is an important ratio to track. It tells users which asset absorbs the compute demand. At current prices, the mint cost has a 310% premium over the price of DIEM. When it compressed toward 100% in Nov/Dec 2025, the mint path became economical and sVVV started getting locked. Looking at the price chart, that was when VVV bottomed.

Locking sVVV to mint DIEM shows that people are willing to convert flexible claims into fixed compute issuance, and this share is now at ~25%. The mechanical bid in VVV shows up when secondary DIEM gets bought up enough that its price pushes breakeven periods beyond what users are willing to accept, and the only way to get more DIEM into the market is to mint it by locking sVVV.

However, there is a scenario where DIEM keeps proving demand without forcing a VVV bid. If users expect compute to remain scarce and are willing to accept longer breakeven periods, they bid secondary DIEM directly at higher prices rather than routing through VVV. In that regime, DIEM reprices upward while locked sVVV stays flat as a share of total staked supply.

A future increase in the DIEM target supply is not bearish the way a normal token supply expansion would be. Since each DIEM is hardcoded to provide $1/day of utility, doubling the supply from 38,000 to 76,000 does not halve the value of existing tokens. The price is still anchored to whatever people are willing to pay for API credits in perpetuity even beyond the breakeven period.

On VVV supply, two forces work against circulating tokens:

  1. Emissions: Venice has lowered annual emissions from 10M at launch to 6M as of Feb 2026, with further discretionary cuts from 1 May reducing emissions by 1M/month to 3M by 1 July. 33M VVV was already burnt in a one-off event in Mar 2025.
  2. Burns: Venice uses an unspecified portion of monthly platform revenue (from Pro subscriptions, USD API credit purchases, 20% of staking yields on the locked portion of sVVV) to buy VVV on the open market and permanently remove it from circulation. This currently runs at ~$100K/month but because VVV’s price has risen, the quantity burnt has actually been declining. For buybacks alone to match the 500K tokens emitted monthly this month, revenue would need to grow 20x.

DIEM is the proof that users want to own compute while VVV is the bet that Venice can keep growing capacity and that future compute demand will eventually need to be capitalized through VVV rather than merely traded in the secondary DIEM market.

As long as compute remains cheap enough on the secondary market, demand should rationally show up in DIEM. The bid in VVV appears when that inventory thins out, when Venice has expanded enough real capacity to raise the DIEM target, and when the next unit of compute has to be minted by locking sVVV.

For users who want fixed compute in perpetuity, currently trading at a ~2 year breakeven and still have residual value, DIEM is the expression. For holders who want exposure to Venice's capacity growth and believe DIEM price will eventually compress the mint spread, VVV is the claim on every future unit of fixed compute.

Alea Research · 6 Apr 2026 · 8 min

Back to top