Anderson Bell 🇺🇸

inference is going to be the largest commodity on earth. once it's being spent at that scale, the only thing worth caring about is where the money lands. openai and anthropic route it to a cap table. neither is profitable. every dollar of revenue services that expectation. venice is profitable, and routes a fixed share of what it makes into buying its own token off the market and burning it. the company bittensor, render and akash have real usage. none of them have a consumer product. venice is the only crypto ai company with a mainstream app, four million users, nine figures of revenue and a changelog that ships every two weeks. • $70m annualized on july 1 • $100m on august 17 • profitable since q1 2026 • four millionth user announced august 12 the bear case, stated properly there's no legal claim. no dividend, no covenant, no obligation. venice could route 0% to burns tomorrow and no holder could do a thing about it. that's true. it's also the wrong place to look. what they've actually done • burned the unclaimed airdrop • bought back and burned team tokens after the launch controversy • started a monthly discretionary buy-and-burn in december 2025 • added programmatic burns on every new subscription • raised those to $2 / $5 / $10 per pro, pro+, max signup in the spring • extended the whole mechanism to api credits in july at $5 per $100 every channel was added. not one was ever removed or reduced. emissions over the same window: 10m a year at launch, then 8m, 6m, 5m in may, 3m in july, 2.5m on september 1, 2m on october 1. five cuts, down 80%. and when they needed $65m for gpus and data centers, sitting on 30m+ vvv worth several hundred million, they sold 8.98% of the company instead. company and team have sold zero tokens through a year where vvv is up 700%. voorhees says the goal is to burn every token in existence, and that series a investors had to accept it going in. none of that was required. all of it was chosen. two years, one direction, no exceptions. why it's rational, not charitable burning doesn't add tokens to the treasury. it raises every holder's share of what's left. venice holds 30m+ of an 80m supply, so its own position compounds faster than anyone else's with every buyback it runs. equity holders own that treasury. they capture the burn program indirectly, through a stake that gets proportionally larger each time. burning tokens makes venice richer. that's why voorhees can say it doesn't harm equity holders, and it's why the program doesn't need a covenant to keep running. incentives outlast contracts. a company with a legal obligation can renegotiate it. a company that owns 38% of its own token has no reason to want to. the series a proves it • 8.98% of the company • a 1.5m vvv vesting grant • warrants on 5m more at ~$13.30, locked a year then vesting over three (on exercise gives venice more than $60m of capital for more gpus and growth which in turn drives token up) that's roughly $110m of token exposure against roughly $90m of equity rn. dragonfly ran diligence on both sides of the capital structure and weighted toward the token. voorhees: "let's sell the equity, we'll keep the tokens." the raise matters for a second reason. the $65m is earmarked for the buildout, so equity financed the data centers and operating cash flow never has to compete with them. outside investors took dilution to pay the capex. the burn program runs underneath it, untouched. the supply math 2m tokens of annual issuance at $17 needs about $34m of annual buying to reach net deflation. at $300m of revenue that's 11% of the top line, and the credit channel alone already routes 5%. the plumbing that exists today gets there on growth, with nothing new announced. burns are only half of it. diem is the other half. each diem staked pays $1 a day of venice api credit forever, and the only way to mint one is locking staked vvv. that target sat at 38,000 from launch until this month. it's now stepping to 40,000 across four increments between august 3 and september 14, the first increase there has ever been. the mint curve is exponential, so those last 2,000 lock far more vvv than the first 2,000 did. well over a million tokens out of float on this expansion alone. burns take tokens out permanently. diem takes them out for as long as anyone wants the compute. both scale with usage. neither needs a decision. what's still coming minds hasn't shipped. creator marketplace, strings already sitting in production js. it runs on credits, and credits already burn $5 per $100, so every mind run burns vvv without venice changing a single policy. this is not including any additional burns and utility minds might add. x402 is what makes that marketplace work. cards floor out near thirty cents, which is why every creator marketplace defaults to subscriptions. x402 deletes the floor, so a mind can charge three cents a run, hold its own wallet, and pay for its own sub-calls. x402 also gives agents more utility, letting users create extremely diverse and powerful agents with access to a ton of tools. i hold a large portion of base:0xacfe6019ed1a7dc6f7b508c02d1b04ec88cc21bf and continue to accumulate more.

tldr for people who won't read 3,500 words: • venice is profitable, $70m+ arr, token throughput 10x'd in 6 months — and you can watch the revenue on-chain, every sub burns $vvv • the raise didn't sideline the token: dragonfly took 8-year warrants on 5m $vvv alongside equity. nobody takes 8yr exposure to something "structurally junior" or has no value • venice pays no dividend, no equity buyback, won't sell treasury. the only pipe surplus exits through is the burn. stated endgame, on the record: "buy and burn every last token." float: 15m • the treasury holds ~$480m of $vvv — 7.4x the raise. burning is venice investing in its own biggest asset • and while digging i found more details about "minds" — an unannounced agent product / marketplace sitting in their production javascript, localized into six languages — the bridge from selling inference to taxing an economy of agents selling outcomes i also read 1,900 tweets of fud and answer the three real ones inside. i'm long, not advice.