Venice AI completes a $65 million Series A—can this privacy AI from the crypto world actually make an end-to-end revenue loop work?

TL;DR

· Venice AI closes $65M Series A at $1B valuation, media says it was profitable in Q1.

· Market focus: Can the privacy-first multi-model inference gateway sustain paying users and gross margins?

· Related assets: VVV, DIEM, crypto-AI, AI inference infrastructure.

Venice AI announced on July 1 that it had closed a $65 million Series A round led by Dragonfly, with participation from Coinbase Ventures, F-Prime, North Island, and others, at a $1 billion valuation. This is the company’s first external funding round, and it brings a familiar crypto narrative before the AI capital market: Can privacy and less censorship sustain a high-growth AI company?

According to The Block, Venice told the media that the company was already profitable in Q1 of this year, with an annualized revenue run rate exceeding $70 million. This claim still needs to be taken with a grain of salt. The annualized revenue run rate extrapolates current revenue over a year, but it is not confirmed full-year revenue, nor is it a fully audited long-term proof of profitability.

But it’s enough to explain why the market is focused on Venice. Over the past two years, AI companies have generally faced the same problem: user growth is fast, but the costs of training models and processing requests are also rising. What makes Venice unusual is that it hasn’t put itself in the race to train the most expensive frontier models. Instead, it is building a privacy-first multi-model inference gateway.

Founder Erik Voorhees previously founded ShapeShift and is a long-time Bitcoin supporter in the crypto space. He set a straightforward product philosophy for Venice: AI should be as neutral as Bitcoin, and platforms should not record everyone’s identity for the sake of a few risks. Now the market needs to verify whether this philosophy can translate into sustainable revenue.

Venice sidesteps the training arms race first

Venice is not building "the next OpenAI." It is more like an AI access gateway: users ask questions on Venice, the platform routes the request to different models, and returns the results. Users can switch models and access both open and some closed-source models.

This determines its cost structure. Training involves teaching models new capabilities and requires ongoing investment in research teams, data, and massive computing power. Inference happens each time a user asks a question, with the model generating an answer in real time. Venice primarily does the latter.

That doesn’t mean Venice has no compute costs. The more user requests, the higher the inference costs. The difference is that it does not have to bear the full cost of training the most powerful models, nor stake the company’s fate on a single model’s capabilities. For investors, this is the prerequisite for being able to tell a profitability story.

The official announcement disclosed that Venice has 3.5 million registered users and processes 1.3 trillion tokens (text units processed by the model) per month. There are discrepancies in API call figures: TechCrunch reported an average of about 1.7 million per day, while Venice’s official figure is about 2 million per day, with a peak of 2.1 million. Regardless of which figure you use, it is no longer just a conceptual project.

Privacy premium must become a reason to pay

The core question Venice must validate is whether privacy can become a reason to pay for an AI product. When ordinary users interact with a chatbot, they’re not worried about technical jargon—they worry about whether their questions, files, ideas, code, and identity information will be recorded, trained on, or censored by the platform.

Venice’s solution is privacy layering. The platform promises not to store user questions on its servers, routing requests to different models through encryption and routing. Routing acts as an intermediary: users are not directly exposed to model providers; instead, Venice forwards the requests.

Boundaries must be acknowledged here. Venice’s privacy notice mentions that in Anonymous mode, third-party model providers may see and store prompts. End-to-end encryption (unreadable by anyone except the two parties) and secure execution environments (hardware-isolated zones) are Pro features, and end-to-end encryption does not support some functions like web search and memory.

So Venice is better understood as a "privacy-layered gateway" rather than an absolutely unmonitorable black box. It can reduce the risk of the platform storing and reading user content, but when requests are sent to closed-source models, data boundaries still depend on specific handling methods and partnership terms.

No censorship is the other side. Voorhees believes that users are adults, and AI platforms should not default to heavy filtering to decide what users can ask and what can be answered. This positioning can attract users dissatisfied with mainstream AI safety filters, but it also brings regulatory and platform distribution risks.

VVV puts revenue into the asset loop

The biggest difference between Venice and ordinary AI applications is that it embeds token design into its business model. VVV is not just a brand asset—it is placed into a cycle of payment, credit, and buyback-and-burn.

In April, the company said it had cumulatively burned over 33.7 million VVV, about 42%. Burning refers to permanently destroying tokens to reduce circulating supply. Venice also said it would use revenue to buy back and burn VVV, allowing investors to link platform revenue to token value capture.

DIEM is another layer. More precisely, users stake VVV to obtain sVVV, then lock sVVV to mint DIEM, which can correspond to or generate API credits. In plain language, VVV is more like an asset gateway, while DIEM is more like a usage quota. Together, they put token holding, AI service usage, and platform revenue into the same closed loop.

That is also why this funding round is not simply seen as equity financing. According to The Block, investors received about 8.98% equity, 1.5 million VVV in vested grants, and warrants to purchase 5 million VVV over the next 8 years. The related tokens are subject to a 1-year lockup, then vest linearly over 3 years.

For VVV holders, both positives and pressures exist. The positive is that revenue buyback and burn provide a clearer asset narrative, and the company’s choice of equity financing may reduce short-term selling pressure from directly selling treasury tokens. The pressure is that if future warrants are exercised and investor tokens gradually unlock, the market will still need to absorb new supply.

A $1B valuation bets on improving gross margins

Venice’s current valuation bets not only on the "privacy AI" narrative, but also on its ability to continue improving unit economics as it scales. The company said the funding will be used to purchase GPUs, build proprietary data centers, and expand its marketing and team.

This step is critical. Current profitability may come from a lighter structure: no frontier model training, partial routing of requests to third-party models, and strong early expense control. If Venice succeeds in building its own infrastructure, it could theoretically lower per-inference costs, improve gross margins, and allocate more revenue to VVV buybacks and burns.

The risks are also here. Data centers and GPU purchases will bring capital expenditures, potentially squeezing profits in the short term. If user growth comes mainly from free or low-cost traffic, inference costs could also erode revenue again. The company has disclosed early growth metrics and profitability information relayed by the media, not yet proof of long-term financial stability.

This funding round turns Venice from a conceptual project into a company that must deliver results. Whether the $1 billion valuation holds depends on whether gross margins, paying user retention, regulatory pressure, and changes in VVV supply can all withstand scrutiny after building its own infrastructure. For investors, the next stronger signal will not be in the narrative, but in revenue quality and the token supply schedule.

Click to learn about BlockBeats job openings

Welcome to join the BlockBeats official community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Twitter Official Account: https://twitter.com/BlockBeatsAsia