Nikshep argues $VVV trades at the cheapest multiple in AI while being the only profitable one, capturing surplus through an automated buy-and-burn mechanism that's already destroyed ~33.8m tokens (42% of remaining supply). Venice's $70m+ ARR grows profitably with subscription burns scaled by tier ($2–$10) firing ~1,250 times daily, and the Dragonfly warrant—denominated in the asset Venice plans to incinerate—signals institutional confidence in the burn thesis rather than betrayal. The next inflection arrives when Venice launches its "minds" marketplace (already flagged in production code), moving from selling inference tokens to outcome-priced agents that crypto enables through identity, payment, and ownership primitives incumbents lack.

Yan argues Grass is a real AI data infrastructure business hiding inside a token, not a typical DePIN project. With 8M+ users sharing internet connections, the network generated $2.75M revenue in Q2 2025, accelerating to ~$50M ARR by Q4 with 197% QoQ growth (verified under NDA by Messari and Grayscale), positioning it to capture share in a $1B web scraping market where competitors like Bright Data disclose >$300M ARR. The token captures all value since there's no equity company above it—a structural advantage the market has overlooked due to thin public information rather than weak fundamentals.

Mike Zajko argues GRASS has achieved a $50M annualized revenue run rate as of Q4 with QoQ growth accelerating from 56% to 197%, verified independently by Messari and Grayscale, by monetizing idle bandwidth from 8M nodes to sell cleaned web data to AI labs at scale. The Foundation structure ensures revenue flows to token holders rather than the operating entity, comparable to Jito's model, while the company has processed 250 petabytes of multimodal data—roughly the entire indexed web—in under 24 months, creating a defensible moat through its filtering and processing infrastructure that frontier labs require for model training.

Nillion's $NIL token has shifted from optional to required: it's the currency for compute credits on the Blind Computer, the stake securing verifiers, and the settlement layer for partnerships like the Vodafone fraud-detection tower project—all scaling with usage rather than linear user growth. The critical risk is that despite rising demand, supply remains net-inflationary at 0.5% yearly minting plus ~2% monthly unlocks through 2029, with no verified on-chain burn mechanism yet; the incoming tokenomics redesign must deliver credible usage-driven burn to flip the math from "usage grows but so does supply" to structural re-rating.

G.W. Jackston argues DIEM, Venice's perpetual $1/day inference credit token, trades at a 410% premium because the market expects the obligation to hold for decades. The real value emerges if Venice builds a monetization stack around DIEM—DIEM lending marketplaces, DeFi collateral integration, premium tiers, and capacity expansion—that generates 95% of revenue without touching the core promise, potentially justifying $50-70 per token versus today's $14.56.

Yan argues Venice's $200M annualized subscription and API revenue is underappreciated, with latest weekly subscriber additions annualizing to $268M run rate. At current trajectory, 12-month forward ARR reaches $260M, compressing multiples to 2.5x revenue versus 15-25x for comparable AI-infrastructure companies. Beyond subscription, VVV's token mechanics—staking, DIEM minting, buy-and-burn, and lock-up demand—capture broader value as revenue growth outpaces the token's market cap expansion.

David argues Cerebras' IPO signals AI's structural shift from training to inference, where demand scales with compute rather than users. Venice's dual-token ecosystem—VVV for staking and platform access, DIEM for direct API credit exposure—is positioned to capture this market, which JP Morgan sizes at 10 to 50 times training's scale. POD, the token behind Dolphin's distributed inference network, offers complementary exposure through its buyback mechanism and staker allocations.

G.W. argues Venice.ai represents crypto's next primitive by offering private, uncensored AI inference requiring no crypto literacy—a grandmother can use it for basic queries. VVV functions as an access key; staking it earns pro-rata API capacity and mints DIEM, a $1/day inference credit token enabling a rental marketplace where idle capacity becomes productive capital, eventually attracting external DeFi protocols and positioning AI agents as blockchain's next billion users.

Nikshep argues Venice's economic structure inverts the AI incumbents' cap-table trap. While OpenAI projects $85B in losses by 2028 and needs $200–280B annual revenue by 2030, open-source models (GLM-5.1, Kimi K2.6, DeepSeek V4-Pro) now match frontier performance at 5-15x cheaper pricing for the 80% of workloads already saturated to "good enough." Venice's zero training costs, token burn mechanics (42% of genesis supply destroyed), and agent-native architecture position it to capture inference demand that agents structurally cannot route through surveillance-based labs requiring KYC, with the agent economy projected at $3-5 trillion by 2030.

Kevin compares VeniceAI's $VVV token valuation against similar infrastructure plays, finding it expensive on most metrics. OpenRouter handles 60x VVV's inference volume at 2x the valuation, while TogetherAI does 45x the volume at 11x the valuation. Inferring Venice's ~$8.4m ARR from recent token burns yields an 80x revenue multiple versus 26x for OpenRouter and 7.5x for TogetherAI, though this depends on unconfirmed burn rate assumptions.