Vast: $400 Million Raised, 20 Million Users, and a World Model That Doesn't Exist Yet

Vast, the Beijing-based startup behind Tripo AI, is weighing a Hong Kong listing, Bloomberg reports.

The headline plugs into a familiar pattern. Manycore Tech - a 3D-design AI startup - listed in Hong Kong in April 2026. MiniMax, co-founded by Vast's own Simon Song before he departed in 2022, raised $619 million in its Hong Kong IPO in February. PixVerse, the Alibaba-backed video AI startup, is also considering an exit there. The route is clear, the timing is opportunistic, and the narrative around Vast is polished.

That is precisely where the analysis should begin: with skepticism.

The Product Is Real. The Pivot Is Not.

Tripo AI converts text and image prompts into 3D mesh models. It does this faster and with higher quality than most competitors. Independent community testing published in April and June 2026 placed Tripo alongside Meshy and Rodin as top-tier image-to-3D tools, with direct-import success rates in the 90-plus percent range when feeding output into engines like Unity.

The product works. That is not a trivial claim in a category where most AI-generated 3D assets still require hours of manual cleanup before they are usable in production.

What does not yet exist is Project Eden, the "world model" that Vast unveiled alongside its latest funding round. A world model generates fully explorable 3D environments with simulated physics - not just static objects. Fei-Fei Li's World Labs, which raised $1 billion earlier in 2026, is still shipping a preview. Vast, a 150-person company founded in 2023, says Eden will launch "as soon as this month" in May 2026 press materials.

A company that generates 3D chairs and avatars from images is not the same company that simulates interactive 3D worlds. The engineering gap between object generation and world simulation is not incremental. It is structural.

The Revenue Math Does Not Support the Timeline

Vast claims 10 million users, mostly from the U.S., Europe, Japan, and South Korea. Individual subscriptions run from $20 to $140 per month, plus token-based usage fees. The company also lists enterprise clients including Sony and NetEase, charging on a project basis.

Even at the highest subscription tier, and assuming 20 million users all pay $140 per month, that is a theoretical ceiling of $33.6 billion in annual recurring revenue. Obviously, that is not realistic. The actual distribution across free tiers, $20 tiers, and enterprise contracts is undisclosed. But at $20 per month averaged across 20 million users, annual revenue is $4.8 billion - still far above what a 150-person startup has actually collected.

Here is the more likely scenario: a small fraction of those 20 million are paying at meaningful tiers, with the vast majority on free or minimal plans. That is how AI startups report user counts. The metric is headline-friendly and economically hollow.

Vast has raised roughly $400 million across its Series A, A+, A++, and A3 rounds since March 2026 alone. Total funding since founding in 2023 likely exceeds that figure. The company has existed for three years, has not reported audited revenue, and has not demonstrated a path from $400 million burned to profitability.

Founder Simon Song told Bloomberg Vast could turn a profit within two years. That statement should be read as marketing asset, not earnings guidance. Any astute investor would ask what revenue run rate a 150-person AI model company needs to reach profitability, and whether its current user economics can scale to that number before the next capital call.

The Funding Pattern Tells the Real Story

The capital raises accelerate, not decelerate:

• March 2026: Series A - $50 million, led by Alibaba and Hengxu Capital
• May 2026: Series A+ and A++ - ~$200 million, led by Ince Capital and a China Life Insurance-backed fund, with Genesis Capital and Primavera Capital Group
• July 2026: Series A3 - raised $150 million, with Geely Capital, gaming companies (4399 Network, Tanwan, Giant Network), Fosun Capital, Orinno Capital, and existing shareholders INCE Capital and Genesis Capital increasing positions

Three rounds in five months. $400 million in less than six months. The investors are deep-pocketed and strategic, and that is real value - Alibaba's backing provides cloud infrastructure and distribution, China Life Insurance provides patient capital, and gaming companies provide early customers. But the pattern of ever-larger rounds, ever-sooner rounds, and ever-bigger product claims is the fingerprint of a company burning through its runway to justify its valuation.

A startup that is two years from profitability and raises $400 million in six months is not being rewarded for efficiency. It is being funded to survive until the IPO window closes.

The Founder's Track Record Is a Warning Sign, Not a Guarantee

Simon Song's pedigree is real. He worked at SenseTime, then co-founded MiniMax - which later became one of China's most prominent AI companies and went public in Hong Kong in February 2026. But Song left MiniMax in 2022, years before its IPO. He founded Vast in 2023 and has now pushed the company from 3D object generation into world models and "interactive TikTok" apps in the same breath.

The pattern of leaving one company before its payoff and pivoting the next company into increasingly ambitious territory is not necessarily bad faith. It is, however, a risk factor. It means the founder's execution record on completion - shipping a product through to maturity and profitability - is untested.

The Cross-Currents

Vast is not a scam. The product works, the user base is genuine, the enterprise relationships with Sony and NetEase are credible, and the capital behind it is institutional rather than speculative. The question is whether the IPO narrative matches the economic reality.

• Directionally bullish: The 3D generation market is real and underserved. Game studios and VFX houses have acute cost pressures, and Tripo AI's 90,000 studio clients represent genuine pipeline integration, not hobbyist signups. If enterprise contracts scale faster than the consumer subscription model, the revenue mix shifts toward higher-margin project work.
• Directionally bearish: World models are years away from production readiness for a company of this size and age. The user count is likely inflated by free-tier accounts. The funding pattern shows acceleration, not deceleration. Two years to profitability assumes revenue triples or quadruples while costs are contained - a scenario that requires every assumption to go right simultaneously.
• Directionally neutral: The Hong Kong listing itself is neither a validation nor a red flag. It is a capital-raising event in a market hungry for AI listings. The pricing will tell investors whether the company believes in its own valuation or is seeking exit liquidity for early shareholders.

The bottom line: Vast has built something useful. It has raised enough money to keep building. But "weighs a Hong Kong listing" is not the same as "has a business that justifies a public market." The product is real. The revenue is not yet.