Vast

1.1 Identity, product, and operating model

Vast was founded in 2023 by Simon Song after his work at SenseTime and MiniMax, according to multiple independent profiles. The company operates the Tripo AI/Tripo Studio platform, which turns text or image prompts into 3D objects and is positioned for game, film, industrial-design, and XR workflows. Public descriptions identify Beijing as the operating base, with reporting also describing staff split between Beijing and Hangzhou and a Cayman Islands registration for Tripo AI. The commercial model is not fully disclosed, but reporting describes creator subscriptions, token-based usage, and project-based enterprise charges. The product is therefore best understood as a hybrid creator software and enterprise/API platform rather than a pure consumer application. The available evidence supports a clear product identity, while the legal-entity structure, board composition, and precise recurring-revenue mix remain diligence items.[CO001, CO002, CO003, CO004, CO006, CO007]

1.2 Leadership, team, and adoption signals

Simon Song is the only named executive consistently identified in the assigned evidence. His prior SenseTime and MiniMax experience is relevant founder-market-fit evidence for an AI platform, but the pool does not establish a complete executive team, board, or succession plan. Scale signals are substantial but inconsistent: one June 2026 profile reports roughly 20 million users and around 150 staff, while an earlier company announcement reported more than 6.5 million creators, 90,000 developers, and nearly 100 million generated assets. These figures may reflect different dates, product surfaces, or user definitions; they should not be combined into a single funnel. Enterprise references to NetEase, Sony, Tencent, Microsoft, and others indicate distribution and workflow validation, but the evidence does not disclose contract size, retention, or customer concentration. Key-person dependence should remain a material diligence risk until governance and leadership depth are documented.[CO008, CO009, CO010, CO011, CO012, CO013]

1.3 Financing, milestones, and evidence limits

The financing record shows a rapid sequence of disclosed rounds and a valuation threshold, but it is not yet a clean audited capitalisation history. Bloomberg and other coverage report nearly $200 million raised in the financing that took Vast above a $1 billion valuation; Forbes reports a March 2026 $50 million Series A and says two people familiar with the later deal placed the valuation at about $1.5 billion, while the company declined to comment on an exact figure. A Dealroom News item later reports a $412.5 million Series B and approximately $687.5 million raised in under six months, whereas a private-company database reports $644 million across three rounds and includes clearly unrelated VAST Data entries. These discrepancies require round documents and entity-level reconciliation before using total raised or valuation as precise cover metrics. Product milestones include Project Eden, platform upgrades, and a stated ambition to make 3D creation broadly accessible; no assigned source establishes recognized revenue, profitability, debt, liquidation preferences, or board control.[CO014, CO015, CO016, CO017]

1.4 Exhibits

2.1 Market boundary and workflow

The relevant market is software and usage spend for generating, editing, and integrating production-ready 3D assets from text or images. It includes creator subscriptions, enterprise/API usage, plugins, and workflow integrations serving games, digital content, ecommerce, XR, and design. It excludes GPU hardware, model-training infrastructure, general purpose AI applications, and downstream game or media revenue. This boundary matters because broad China AI figures combine chips, infrastructure, models, and services, while private-market estimates may cover only software and applications. Status-quo substitutes remain internal 3D artists, outsourced studios, procedural tools, and generic image/video models that do not deliver editable geometry. The economic wedge is reduced modeling and cleanup time, but production acceptance still depends on topology, UVs, materials, and editability rather than visual plausibility alone.[CM001, CM002, CM003, CM004]

2.2 Sizing lenses and adoption signals

No retained source isolates a defensible Vast-specific SAM or SOM. A bounded sizing approach therefore separates three lenses: the broad China AI stack, the narrower software/application market, and the company-relevant generative-3D workflow. Axis Intelligence reports that China’s core AI industry exceeded 1.2 trillion yuan in 2025, while it separately cites a $28.18 billion 2025 China AI software and applications estimate and $12.4 billion of private AI investment. These values are not interchangeable market revenues. The same source reports 602 million generative-AI users in December 2025, but that is a usage population, not a paid 3D buyer count. Vast-related reporting describes roughly 10 million individual users and 90,000 studios or companies, while other reporting says Tripo Studio serves tens of millions globally; the discrepancy makes an addressable paid-user calculation premature. The market case is therefore directionally attractive but valuation-sensitive to conversion, retention, enterprise ACV, and the share of workloads that require production-grade 3D output.[CM005, CM006, CM007, CM008, CM009]

2.3 Buyers, drivers, and constraints

The buyer map has at least three routes. Individual creators and small studios are likely self-serve subscription users; larger game, media, design, and technology organizations are API or integration buyers; and strategic investors or platform partners can provide distribution while also shaping requirements. Dealroom reports Tripo integration into NetEase, Tencent, ByteDance, and Microsoft workflows, and other reporting names NetEase and Sony as clients, but the public record does not disclose contract values, budget owners, renewal rates, or deployment volumes. Adoption drivers are labor-cost compression, faster prototyping, and expanding demand for interactive content. Constraints include quality assurance, workflow switching and integration costs, compute intensity, content and safety compliance, open-source competition, platform bundling, and cross-border technology restrictions. Chinese guidance issued by the Cyberspace Administration of China, NDRC, and MIIT emphasizes safety, security, standardization, and application-driven development for AI agents; while not a 3D-specific rule, it is a relevant policy signal that regulated deployment and controllability can affect enterprise adoption timing.[CM010, CM011, CM012, CM013, CM014]

2.4 Sizing and adoption gaps

The central diligence gap is not whether AI adoption is growing, but how much of that growth becomes recurring spend on Vast’s production-grade 3D workflow. Public sources do not provide a consistent paid-user denominator, cohort conversion, enterprise ACV, usage-based gross margin, or customer concentration. They also do not reconcile whether reported user counts are registered, active, or paying, nor whether “studios and companies” are customers, prospects, or platform accounts. Broad industry figures should not be used to back-solve Vast revenue. A buyer diligence request should obtain plan-level active users, free-to-paid conversion, retention, API volume, named-customer production use, compute cost per generated asset, and the percentage of output accepted without substantial human rework. Until those measures are available, market momentum supports a positive adoption thesis but not a precise revenue multiple or SOM claim.[CM015, CM016]

2.5 Exhibits

3.1 Competitive landscape and strategic position

Vast's relevant competitive set spans specialized text-to-3D products, multimodal creative platforms, and the incumbent workflow of manual modeling or internal studio teams. Independent coverage describes Tripo as a 3D foundation-model and content-platform effort. Meshy is the closest specialized public comparator: its own product surface emphasizes text/image-to-3D, PBR texturing, rigging, animation, game-engine exports, and 3D-printing workflows. Luma is a broader creative-intelligence platform spanning video, image, audio, and text, and Microsoft's Sora documentation shows how general-purpose video generation can absorb adjacent creative budgets without proving direct 3D parity. The practical buyer choice is therefore not only which 3D generator to use, but whether to buy a specialist, use a broader creative suite, or retain manual and internal production. Public evidence does not enumerate all likely entrants, so platform expansion remains a scenario rather than a complete competitor census.[CP001, CP002, CP003, CP004, CP005, CP013]

3.2 Capability, pricing, and buyer trade-offs

Public evidence supports a capability comparison but leaves important commercial cells unresolved. Vast's reported Tripo Studio and enterprise API address game studios and content producers, while Meshy publishes a wider visible workflow spanning creation, texturing, rigging, animation, game engines, and print exports. A Meshy-issued release distributed by PRNewswire claims $15 million ARR, 30% monthly revenue growth in 2025, 85% gross margin, and six million users; each metric remains a competitor claim rather than an independent benchmark of Vast. Publicly visible comparable pricing is absent from the retained Vast, Meshy, Luma, and Microsoft evidence. Accordingly, the matrix marks price, realized discounts, SLA, and switching costs as unknown rather than inferring parity from feature lists or treating funding as operating scale.[CP006, CP007, CP008, CP009, CP012, CP014, CP015]

3.3 Moat durability and displacement risk

Vast has credible technical and ecosystem signals: a Hamidun report says its Tripo P1.0 model generated a topologically clean model in about two seconds and that the company had five SIGGRAPH Technical Papers in 2026. Dealroom reports Tripo P2.0 Preview raised the single-object cap from 20,000 to 50,000 triangles and added native quad topology, while Dealroom and KuCoin name integrations or clients including NetEase, Tencent, ByteDance, Microsoft, and Sony. Those signals can support product quality and distribution, but they do not establish retention, paid conversion, gross profit, or exclusive supply. Meshy's visible workflow breadth and claimed commercial scale raise commoditization pressure, while Luma and Microsoft illustrate substitution by broader creative or multimodal platforms. Multiples' reported funding total shows capital capacity, not customer lock-in. The moat is therefore conditional on continued model quality, workflow completion, enterprise trust, and measurable usage conversion.[CP008, CP009, CP010, CP011, CP012, CP013, CP016]

3.4 Exhibits

4.1 Monetization and public traction

Vast monetizes Tripo through subscription tools, creator software, and developer APIs, while public reporting also describes a B2B and B2C customer focus. The product is distributed globally: SCMP reported that more than 85% of users were outside China and identified Europe and the United States as the largest markets. The same report said monthly income grew 2.5 times within one month and fivefold within three months after the beta launch, but this is a founder quotation rather than audited revenue. Auganix reported more than 6.5 million creators, 90,000 developers, and nearly 100 million generated assets, while Dealroom described tens of millions of Tripo Studio users and integrations with NetEase, Tencent, ByteDance, and Microsoft. These are useful demand and distribution indicators, but they do not establish paid conversion, recognized revenue, average revenue per account, retention, or enterprise contract value. The strongest commercial diligence question is whether usage converts into durable paid cohorts rather than promotional or free activity.[CI001, CI002, CI003, CI004]

4.2 Pricing, delivery economics, and unit economics

Public sources identify the monetization mechanisms but not list prices, realized prices, discounting, API consumption rates, or customer-level contract terms. That prevents a reliable calculation of gross profit per generated asset, contribution margin by plan, CAC payback, or sales efficiency. SCMP described a very lean commercial motion in which only two salespeople were focused on service and the company relied on organic growth and word of mouth; this supports a low reported sales-and-marketing burden at that point in time, but it is not a complete CAC measure and may not persist as enterprise selling scales. The technical and service-cost side is also unresolved: generating production-ready meshes in seconds and operating a large spatial-asset corpus imply compute, storage, inference, and support costs, but no public source supplies a cost-per-generation or gross-margin figure. The table therefore distinguishes observed mechanisms from missing underwriting inputs. An independent review adds pricing risk: Tencent's open-source Hunyuan3D could push the category toward commoditization before Vast proves retention in production workflows.[CI006, CI007, CI009, CI019]

4.3 Capital adequacy and financing dependency

Vast has unusually strong public financing signals for a private company. Independent coverage described nearly $200 million raised around June 2026 and a valuation above $1 billion; Multiples subsequently listed a September Series B of about $446 million and total financing of $644 million across three rounds, while Dealroom's article text said the company reported roughly $687.5 million raised in under six months. These figures are not fully reconciled: sources use different round labels, currencies, and aggregation boundaries, and the company-level financing announcement in Cyzone describes more than RMB1 billion for an A3 strategic round without publishing cash-on-hand or obligations. Capital raised is not cash available. No assigned source reports cash balance, monthly burn, runway, debt, project-finance commitments, or the next-round trigger. The financing pace reduces near-term funding risk only conditionally; it also raises the diligence need for cap-table terms, use of proceeds, investor concentration, and whether infrastructure investment will consume capital faster than monetization develops.[CI010, CI011, CI012, CI013, CI014, CI015]

4.4 Financial verdict and diligence blockers

The public record supports a thesis of strong demand signals and financing access, not a verified profitability or valuation case. Usage growth, international reach, named enterprise integrations, and a broad investor base are positive indicators for revenue potential. Conversely, the evidence does not establish ARR, recognized revenue, gross margin, net revenue retention, free cash flow, customer concentration, or the realized price paid by enterprise users. Open-source competition adds a credible risk to pricing power before paid-cohort durability is known. The reported valuation and financing totals are particularly difficult to underwrite because they are headline private-market figures rather than a disclosed common-equity value or a cash-flow multiple. A decision-ready model should be built from monthly paid accounts, API and subscription revenue by cohort, generation volume, inference and storage cost, enterprise ACV, renewal and expansion data, headcount, cash, burn, debt, and financing preferences. Until those inputs are supplied, Vast merits continued diligence rather than a conclusion based on revenue quality or margin path.[CI016, CI017, CI019]

4.5 Exhibits

5.1 Product workflow and user jobs

Vast delivers Tripo as a creator and developer workflow rather than only a model endpoint. Public descriptions cover text-to-3D, image-to-3D and multi-view generation, with exports into common 3D formats and a REST API for application integration. A review describes the product as spanning modeling, texturing, retopology, rigging and animation, while another source describes an integrated workspace, creator applications and developer APIs. This makes the immediate customer job clear: create a usable starting asset quickly, then move it into an existing game, visualization or content pipeline. The strongest use cases are rapid prototyping, props, hard-surface objects and budget-constrained teams; final hero assets and complex organic forms still require specialist intervention. Public user and generation totals are company-reported through coverage that does not expose denominators, cohort definitions or production conversion, so they are adoption signals rather than proof of durable commercial usage.[CE001, CE002, CE003, CE004]

5.2 Technical architecture and maturity

The product architecture has three observable layers. First, Tripo exposes generation modes and model families, including H3.1 and P1.0, with the company describing P1.0 as a native 3D diffusion approach that models spatial structure holistically. Second, post-processing addresses topology, texturing, low-poly optimization and rigging. Third, plugins, standard exports and an asynchronous API connect outputs to Unity, Unreal, Godot, ComfyUI and other tools. This is a credible operating model for reducing workflow friction, but several maturity claims remain issuer-led: production-ready output, two-second generation, and platform scale are reported by publications relaying company claims. An independent review is more qualified: it reports strong results for simple props while finding hollow areas, organic shapes, UVs and rigging still need manual work. The architecture therefore appears commercially usable for bounded workflows, not yet a zero-touch replacement for 3D production.[CE005, CE006, CE007, CE008, CE009]

5.3 Differentiation, dependencies and trust risks

Differentiation is concentrated in the combination of native 3D model development, workflow integrations and a growing creator/developer surface, rather than in a publicly audited benchmark. The independent review's comparison credits Tripo with speed, broad plugin coverage and value, while giving competitors an advantage on some complex geometry or editing workflows. The API guide shows practical dependencies on API authentication, asynchronous task polling, credit limits, rate limits and downloadable result URLs; these are implementation and service-continuity considerations for enterprise adoption. Public evidence does not establish uptime, security certifications, privacy controls, data-retention terms, model-evaluation methodology, or the legal status of generated asset rights across jurisdictions. Project Eden and world-model descriptions are roadmap material, not shipped capability. Buyers should validate asset-license terms, reproducibility, failure rates, queue behavior, incident history and support commitments before embedding Tripo in production.[CE010, CE011, CE012, CE013]

5.4 Product diligence conclusion

Product evidence supports a useful wedge: fast 3D asset ideation and pipeline acceleration for creators, developers and smaller studios. The evidence is less conclusive on enterprise-grade reliability and on whether the platform can consistently produce final assets without cleanup. Vast has expanded from generation into APIs, plugins and post-processing, which can increase switching costs if integrations become embedded in customer workflows. The central technical risk is quality variance at the boundary between a compelling demo and a repeatable production asset: topology, hollow geometry, UV seams, organic forms, licensing and operational controls all remain diligence items. Treat current adoption, model performance and unreleased world-model initiatives as separate evidence categories rather than assuming one proves the others.[CE008, CE009, CE010, CE012, CE013]

5.5 Exhibits

6.1 Who uses and pays for Tripo

Vast’s customer surface spans individual creators, developers, studios and companies, and potentially industrial users. The company describes Tripo AI as serving creators and developers, while external coverage reports nearly 10 million individual users and approximately 90,000 studios and companies; another current account reports 20 million users but also describes the figure as coming from the company. These counts are not directly comparable because their populations and dates differ, so they should be treated as reported reach rather than a verified active-customer base. The commercial path is visible in the public pricing structure: a free, non-commercial tier is separated from paid plans beginning at $20 per user per month, with higher limits and commercial features. This supports a plausible self-serve acquisition funnel, but no public source here discloses paid conversion, account-level revenue, contract value, or the share of users who become production customers.[CU001, CU002, CU003, CU004]

6.2 Adoption signals and durability

The clearest trajectory signal is the increase in reported platform scale across contemporaneous accounts: Tripo’s March release said more than 6.5 million creators and 90,000 developers had generated nearly 100 million models, while May coverage cited nearly 10 million individual users and roughly 90,000 studios and companies. A later funding account reported tens of millions of users globally and integrations into NetEase, Tencent, ByteDance, and Microsoft, but it did not provide active-user, paid-account, or revenue denominators. The named production evidence is stronger than logo-only evidence where a use case is specified: the company’s GDC material describes game-engine workflows and a 250-person developer event, and current coverage identifies NetEase use in Where Winds Meet. Still, the public record does not establish repeat usage, renewal, churn, gross retention, net retention, or whether named relationships are pilots, production deployments, or ecosystem integrations.[CU005, CU006, CU007, CU008, CU009]

6.3 Named proof, expansion, and concentration

Named-company evidence indicates that Tripo is moving beyond an individual creator tool toward developer and studio workflows. Implicator reports NetEase use in the game Where Winds Meet and identifies the company’s stated developer and enterprise/API reach; Dealroom reports integrations into NetEase, Tencent, ByteDance, and Microsoft. These are useful validation signals, but they remain principally company-originated or secondary summaries rather than customer-authored outcome studies. The public evidence does not quantify production seats, spend, deployment locations, time-to-value, asset acceptance rates, expansion from a first team to a wider organization, or revenue concentration by customer. Pricing research also indicates that commercial use, private models, higher concurrency, batch generation, and API/integration access sit behind paid tiers, making those upgrade triggers plausible but not proof of conversion. Investors should therefore distinguish reach and integration from durable, revenue-producing adoption and request customer-level cohort evidence before underwriting expansion.[CU010, CU011, CU012, CU013]

6.4 Evidence limits and diligence priorities

Public customer evidence is directionally positive but incomplete. The sources support segmentation, reported reach, named ecosystem relationships, and a visible usage-based subscription model; they do not support a precise active-customer count, paid conversion rate, cohort retention, expansion rate, average contract value, or concentration percentage. Conflicting scale figures should not be reconciled by averaging them: the 6.5 million creator figure, nearly 10 million individual-user figure, and 20 million user figure have different sources, dates, and denominators. The next diligence step is a customer-level export that ties account IDs to plan, geography, use case, production status, gross and net retention, expansion, churn reason, and revenue. A second step is reference verification with at least three named customers, including one that can attest to production use rather than experimentation.[CU014, CU015, CU016]

6.5 Exhibits

7.1 Ranked risk profile

Vast's highest-consequence risks are not disproven incidents; they are exposed dependencies that public evidence does not yet close. The company operates a generative-AI service whose output and training-data obligations can create legal exposure, while its rapid financing trajectory raises expectations that may outpace disclosed revenue. Commercial proof is described through user, studio, and enterprise-integration counts, but those measures do not establish paid conversion, retention, gross margin, or customer concentration. The operating thesis therefore depends on converting adoption into durable revenue while maintaining documented data governance, output review, and resilient service operations. Kill criteria should be tied to observable events rather than inferred from the absence of public incidents.[CR001, CR002, CR003, CR004]

7.2 Regulatory and legal exposure

The most concrete external risk is the unsettled boundary between generative-AI training, copyrighted inputs, and infringing outputs. The retained legal analyses describe divergent U.S. district-court outcomes and emphasize that fair-use treatment is fact-specific, that unauthorized acquisition can undermine a defense, and that outputs can still create liability. China-focused guidance describes controls for training-data provenance, harmful-content filtering, personal-information consent, monitoring, complaints, and continuity. These are risk indicators and compliance expectations, not proof that Vast violated a rule. The diligence priority is a documented inventory of training sources, licenses and consents, complaint handling, model monitoring, and jurisdiction-by-jurisdiction obligations.[CR005, CR006, CR007, CR008]

7.3 Operational, financial, and partner dependencies

Vast's operating model has several linked execution risks. The sources describe a fast sequence of funding rounds, a large claimed user and studio base, enterprise integrations, and a plan to build both foundation models and a distribution platform. Those signals can support scale, but they also create exposure to inference and storage costs, model-quality regressions, platform uptime, enterprise procurement, and the need to turn broad usage into paid recurring or usage-based revenue. The investor mix includes strategic and state-linked participants, which may provide capital and distribution but can also complicate governance, concentration, and future financing expectations. Public evidence does not establish runway, cash burn, customer concentration, or contractual protections.[CR001, CR002, CR003, CR009, CR010, CR011]

7.4 Mitigations, monitoring, and thesis-break triggers

The appropriate mitigation is evidence-led operating control rather than a blanket assumption that Vast is either safe or unsafe. Management should be asked for training-data lineage and rights records, model and output incident logs, privacy and consent procedures, service-continuity tests, unit economics by plan and workload, customer renewal and concentration data, and financing documents. Monitoring should distinguish a disclosed fact from an unresolved gap: a new complaint, takedown, outage, failed enterprise deployment, material model-quality regression, down-round, or inability to reconcile funding and cash availability would be more probative than a marketing metric alone. Until those records are produced, the residual risk remains material and the investment case should use conservative assumptions.[CR006, CR008, CR009, CR010, CR011, CR012]

7.5 Exhibits

8.1 Investment case and anti-thesis

Vast has a credible category and financing signal: independent coverage describes Tripo AI as a text- and image-to-3D platform, while public reporting attributes nearly $200 million of 2026 financing to the company and says it crossed the billion-dollar valuation threshold. Reported distribution is also notable: AI Weekly describes 10 million individual users and 90,000 studios, and names Sony and NetEase as customers, although those figures are company-reported or reported without audited denominators. The anti-thesis is that this evidence establishes attention and capital access, not economic value. No retained source verifies ARR, recognized revenue, gross margin, retention, customer concentration, preference terms, or a precise post-money valuation. A second set of 2026 data pages reports a much larger Series B and substantially higher cumulative funding, but those figures conflict with the nearly $200 million account and require cap-table and announcement-level verification before underwriting.[CV001, CV002, CV003, CV004, CV005]

8.2 Price discipline and comparable evidence

The available public comparable evidence supports a framework, not a defensible mark for Vast. Multiples' October 2026 public software sample reports a 3.5x EV/revenue median for artificial-intelligence software and a 2.1x median across the listed horizontal SaaS categories; the source says its public-market data is powered by FactSet consensus estimates and Morningstar historical data. Acquiry's private-market discussion places high-growth AI-native SaaS at 10x–20x ARR, but explicitly says private transactions are not directly comparable to public companies and stresses NRR, CAC payback, gross margin, and profitability. Vast's retained evidence does not provide those inputs, and the company's model combines subscription and usage-based revenue according to a market-data profile. Applying either benchmark to an unverified revenue base would create false precision, so the missing metric pack is itself the central valuation finding. TechCrunch's report of a $20 billion financing valuation for Chinese AI company Moonshot is a useful signal of private-market appetite, not a direct comp: Moonshot builds language models, and this chapter has no evidence that its revenue quality, economics, or financing rights match Vast's.[CV006, CV007, CV008, CV009, CV015]

8.3 Scenarios, exit readiness, and diligence triggers

A bull case requires converting reported user and studio reach into paid usage, repeat production workflows, durable gross margins, and a clean financing record. A base case assumes the product remains strategically relevant but public evidence cannot yet demonstrate monetization or common-equity value; that supports tracking rather than a buy call. A bear case combines financing-data inconsistency, AI-3D competition from Tencent and other model providers, customer or geopolitical concentration, and a valuation reset if growth does not become measurable revenue. Exit readiness is therefore conditional: a future financing, secondary, acquisition, or public-market process should disclose audited or investor-grade revenue, retention, unit economics, cap-table preferences, dilution, and customer concentration. The immediate thesis-break triggers are a failure to reconcile the rounds, materially weaker paid conversion than reported reach implies, or evidence that inference costs prevent attractive gross margins. A low-reputation but directionally adverse category survey adds a caution rather than a benchmark: it characterizes generative AI and generic AI platforms as crowded or overvalued and consumer-focused startups as riskier. That warning does not prove Vast is overvalued, but it reinforces the need to price verified economics rather than category enthusiasm.[CV010, CV011, CV012, CV013, CV014, CV016]

8.4 Exhibits