Volumes

Cover Block

PUBLIC

Name	Volumes
Tagline	Capturing ground-truth reality in motion and turning it into accurate, licensable spatiotemporal data for AI labs. [Volumes, retrieved 2024]
Headquarters	New York, NY
Stage	Seed
Business Model	B2B
Industry	Deeptech
Technology	AI / Machine Learning
Growth Profile	Venture Scale
Founding Team	Co-Founders (2)
Funding Label	Seed

Links

PUBLIC

• Website: https://www.volumes.cloud/
• LinkedIn: https://fr.linkedin.com/company/volumesmedia

Executive Summary

PUBLIC

Volumes is building a pipeline for ground-truth spatiotemporal data, a foundational input for the emerging physical AI stack where models must learn to perceive and interact with the real world [Volumes, retrieved 2024]. The company captures real-world events using multi-sensor volumetric rigs, processing the raw feeds into licensable 3D and 4D Gaussian Splatting reconstructions and synthetic data [Volumes, retrieved 2024]. This focus on measurable, physics-accurate data differentiates it from purely synthetic or lower-fidelity video datasets, positioning the company as a potential infrastructure provider for AI labs developing robotics, autonomous systems, and embodied agents.

The venture is led by Chet Ellis, a co-founder with a background in multimodal AI data and large-scale media production [Volumes, retrieved 2024]. His experience suggests a practitioner's understanding of the data quality and scale requirements for training advanced models. The company is headquartered in New York and has secured Seed-stage capital, though the specific amount and investor syndicate are not yet public [Crunchbase, retrieved 2026].

The immediate watch points are commercial traction with AI research labs, the scalability of its capture and processing operations, and the evolution of the competitive landscape which includes established volumetric video specialists like Evercoast [evercoast.com, retrieved 2026]. Over the next 12-18 months, the key indicator will be whether Volumes can convert its technical capability into recurring, high-value data licensing agreements.

Data Accuracy: YELLOW -- Core product claims are sourced from the company's website. Seed funding is confirmed but details are limited. Founder background is partially corroborated.

Taxonomy Snapshot

Axis	Value
Stage	Seed
Business Model	B2B
Industry / Vertical	Deeptech
Technology Type	AI / Machine Learning
Headquarters	New York, NY
Growth Profile	Venture Scale
Founding Team	Co-Founders (2)
Funding	Seed

Company Overview

PUBLIC

Volumes is a New York-based startup that emerged to address a specific gap in AI development: the scarcity of high-fidelity, real-world data that captures the physics of motion. The company's founding narrative centers on the premise that next-generation physical AI models require training on data that preserves the full spatiotemporal dynamics of reality, not just static images or synthetic approximations. While the exact founding date is not public, the company's product focus and Seed-stage status place its emergence in the broader context of the 2023-2024 surge in investment for physical AI and robotics infrastructure [Crunchbase, retrieved 2026].

The company's headquarters in New York, NY, situates it within a significant hub for media production and AI research, aligning with its focus on large-scale capture and data processing. The sole confirmed founder, Chet Ellis, serves as CEO and brings a background in multimodal AI data and large-scale media production, a profile that matches the technical and operational demands of volumetric capture [Volumes, retrieved 2024]. Ellis's educational background includes Harvard, though his specific degree and prior founding experience are not detailed in public sources [Volumes, retrieved 2024].

Public milestones are currently limited to the establishment of its core product offering and the securing of Seed funding. The company has defined its product suite, which includes raw ground-truth capture, 3D/4D Gaussian Splatting reconstructions, and synthetic data generation, all detailed on its public website [Volumes, retrieved 2024]. A Seed round has been logged on Crunchbase, though the amount, lead investor, and precise timing remain undisclosed [Crunchbase, retrieved 2026]. This positions Volumes in the early build phase, with capital likely directed toward refining its capture rigs, data pipelines, and initial customer engagements.

Data Accuracy: YELLOW -- Core company details (HQ, founder role, product) confirmed by primary source; funding stage corroborated by one database; founding date and round specifics not public.

Product and Technology

MIXED

The company's core offering is a data-as-a-service platform for physical AI, built on a volumetric capture pipeline. The product suite is segmented into three distinct data outputs, each targeting a specific need in the AI training and simulation workflow [Volumes, retrieved 2024].

• Raw Ground-Truth Data. This is the foundational product: unprocessed volumetric captures from multi-sensor rigs deployed in the real world. The company positions this as a source of pure, unadulterated spatiotemporal data, capturing spatial relationships, geometry, temporal dynamics, environmental context, and physical interactions [Volumes, retrieved 2024].
• 3D & 4D Gaussian Splatting. The platform processes raw captures into two forms of reconstructions. Static 3D Gaussian Splatting produces high-fidelity scene models, while 4D Gaussian Splatting adds the temporal dimension, creating a measurable, licensable copy of space, time, and physics [Volumes, retrieved 2024].
• Synthetic Volumetric Data. To expand available training sets, the company also generates AI-synthesized scenes. This suggests a pipeline that can augment real-world captures with simulated variations, though the specific generative techniques are not detailed publicly [Volumes, retrieved 2024].

The technology stack is not explicitly documented, but the product description implies a heavy reliance on computer vision, sensor fusion, and novel-view synthesis techniques centered on Gaussian Splatting. The mention of "multi-sensor rigs" indicates a hardware component for data acquisition, while the generation of synthetic data points to internal AI model development. The entire system is architected to produce "accurate, licensable spatiotemporal data," framing the output as a commercial asset for AI labs rather than a one-off capture service [Volumes, retrieved 2024].

PUBLIC The demand for high-fidelity, real-world data is becoming a primary constraint for the next generation of AI models, shifting the competitive edge from model architecture to training data quality. For companies building physical AI, such as embodied agents or robotics, the scarcity of accurate spatiotemporal data that captures real-world physics and dynamics is a recognized bottleneck. Volumes positions itself directly at this intersection, offering a data capture and generation service for this nascent but critical market.

Defining the total addressable market for volumetric capture and synthetic spatiotemporal data is challenging due to its early stage, but analogous markets provide a useful frame. The broader market for AI training data was valued at approximately $2.5 billion in 2023, with projections for high growth as models become more complex and multimodal [Grand View Research, 2023]. More specifically, the market for computer vision data, which includes 3D and video data, represents a significant segment of this total. While a specific TAM for 4D Gaussian splatting or licensable volumetric data is not yet established in public reports, the rapid investment in physical AI and robotics suggests a SAM that could scale into the hundreds of millions as the technology matures.

Key demand drivers are well-documented in adjacent research. The push towards more capable embodied AI and robotics requires training on data that reflects real-world object permanence, physics, and temporal sequences, a need highlighted in recent papers from leading AI labs. Furthermore, the proliferation of consumer and industrial applications for augmented and virtual reality continues to fuel demand for high-quality 3D content. A third driver is the growing regulatory and ethical scrutiny around the use of scraped internet data for training, which may incentivize AI developers to seek licensed, rights-cleared datasets like those Volumes aims to produce.

Adjacent and substitute markets include traditional 3D asset creation studios, photogrammetry services, and synthetic data generation platforms. These markets are larger and more established but often lack the specific spatiotemporal and ground-truth fidelity Volumes emphasizes. The primary competitive risk is not substitution from these adjacent markets, but from large AI labs deciding to build in-house volumetric capture capabilities, given their substantial resources.

Given the absence of a specific, cited market sizing report for Volumes' exact niche, the following table presents sizing claims from analogous, publicly reported markets to contextualize the opportunity.

Market Segment	2023 Size	Projected CAGR	Source
AI Training Data	$2.5B	21.5% (2024-2030)	[Grand View Research, 2023]
Computer Vision (Global)	$16.3B	7.8% (2024-2030)	[Grand View Research, 2023]
Synthetic Data Generation	$0.2B	35.4% (2023-2030)	[MarketsandMarkets, 2023]

The analyst takeaway is that while the core market for volumetric data-as-a-service is not yet quantified, it sits at the convergence of several high-growth, billion-dollar adjacent markets. The company's potential SAM is likely a fraction of these larger pools today, but the tailwinds from physical AI development could accelerate its expansion rapidly.

Data Accuracy: YELLOW -- Market sizing is drawn from analogous, third-party industry reports; specific TAM for the company's niche is not publicly available.

Competitive Landscape

MIXED

Volumes operates in a specialized niche where the competitive map is defined by technical fidelity, data capture methodology, and the intended end-use for the resulting assets.

Evercoast | 6 | companies
4DViews | 5 | companies
Volumes | 4 | companies

Company	Positioning	Stage / Funding	Notable Differentiator	Source
Volumes	B2B provider of ground-truth spatiotemporal data for physical AI training.	Seed stage; amount undisclosed.	Focus on raw, unprocessed volumetric captures and licensable data for AI labs, rather than end-user applications.	[Volumes, retrieved 2024]
Evercoast	Enterprise volumetric video platform for immersive experiences and Gaussian Splatting.	Venture-backed; latest round undisclosed.	Offers a full-stack platform from capture to playback, targeting enterprise clients in media and entertainment.	[evercoast.com, retrieved 2026]
4DViews	Provider of professional volumetric capture systems and post-production software.	Established company; funding history not public.	Long-standing focus on high-end film, VFX, and broadcast production with proprietary hardware and software.	[4DViews]

The analyst takeaway is that the competitive set is small but distinct, with each player emphasizing a different layer of the value chain, from raw data to finished applications.

Segmenting the competitive landscape reveals three primary clusters. The first includes incumbent capture specialists like 4DViews, which have built businesses around high-fidelity systems for media and entertainment. Their advantage lies in established hardware and software pipelines, but their focus is on content creation, not data licensing for AI. The second cluster comprises platform-oriented challengers such as Evercoast, which aim to democratize volumetric video capture and rendering, including Gaussian Splatting outputs, for broader enterprise use. They compete more directly on the application layer. Volumes sits in a third, emerging cluster: data-as-a-service providers for AI. Its closest substitutes are not other capture companies but providers of synthetic data for simulation or 2D video datasets for computer vision, which lack the spatiotemporal ground truth Volumes promises.

Volumes' current defensible edge appears to be its stated focus on raw, licensable ground-truth data for AI labs [Volumes, retrieved 2024]. This is a perishable edge, however. It depends on maintaining perceived superiority in data accuracy and a first-mover advantage in building relationships with leading physical AI research teams. If competitors like Evercoast begin offering similar raw data feeds or if AI labs develop in-house capture capabilities, this differentiation could erode. The edge is also tied to the company's proprietary multi-sensor rigs and fusion techniques, which are not publicly detailed but constitute its core technical moat.

The company's most significant exposure is to platforms that can move up or down the stack. A well-funded competitor like Evercoast, with its enterprise sales motion and full-stack platform, could decide to package and sell raw data bundles, directly challenging Volumes' core offering. Furthermore, Volumes is exposed to potential competition from the very customers it serves. Large AI labs with sufficient capital could invest in building their own volumetric capture studios, internalizing the data supply chain. The company's lack of publicly disclosed commercial deployments or named lighthouse customers makes this risk more acute, as it has not yet demonstrated an unassailable distribution channel.

The most plausible 18-month scenario hinges on adoption by a major AI research institution. If Volumes successfully lands a flagship partnership with a lab like OpenAI, Google DeepMind, or Tesla, validating its data's utility for robotics or autonomous system training, it becomes the winner if early AI adoption proves decisive. It would gain a reference customer, a revenue stream to fund R&D, and a powerful narrative. Conversely, Volumes becomes the loser if platform players capture the market. If Evercoast expands its data offerings and leverages its existing enterprise relationships to become the default volumetric data source, or if 4DViews pivots to serve the AI market, Volumes could be relegated to a niche supplier. Its seed-stage funding, against more established competitors, means capital could become a constraint in a land-grab scenario.

Data Accuracy: YELLOW -- Competitor positioning and stage data are drawn from public sources and company websites; Volumes' own details are from its primary source. Funding specifics for competitors are not fully corroborated.

Opportunity

PUBLIC If Volumes can establish its ground-truth spatiotemporal data as the foundational layer for training and validating physical AI models, the company could become the de facto standard for real-world AI perception, a role with significant enterprise and platform value.

The headline opportunity for Volumes is to become the primary supplier of trusted, licensable reality data for the emerging physical AI stack. The company's focus on raw, unprocessed volumetric captures positions its output as a form of high-fidelity measurement, not just another synthetic dataset. This is critical for AI models that must interact safely and predictably with the physical world, such as those powering autonomous systems, embodied robotics, and advanced simulation. The company's early articulation of its product suite, from raw data to 4D Gaussian Splatting, suggests a deliberate approach to building a comprehensive data pipeline [Volumes, retrieved 2024]. The outcome is reachable because the demand signal is clear: AI labs require vast, accurate, and legally licensable data of real-world physics and motion, a need that existing 2D image and video datasets cannot fully satisfy.

Growth would likely follow one of several concrete paths, each with a distinct catalyst.

Scenario	What happens	Catalyst	Why it's plausible
The Foundational Data Partner	Volumes becomes the go-to source for training data among leading AI research labs and large model developers.	A formal, public data partnership with a major AI lab (e.g., OpenAI, Anthropic, Google DeepMind) is announced.	The company's stated mission directly targets "AI labs" as its primary customer [Volumes, retrieved 2024]. The technical specificity of its offering (4D Gaussian Splatting, multi-sensor fusion) aligns with the advanced needs of these organizations.
The Simulation Standard	The company's synthetic volumetric data generation becomes the preferred method for creating expansive, physically accurate training environments for robotics and autonomous vehicle developers.	A top-tier autonomous vehicle or robotics company (e.g., Waymo, Boston Dynamics) publicly cites using Volumes' synthetic data to train a new model.	The product line explicitly includes "AI-generated scenes for expanded training sets" [Volumes, retrieved 2024], addressing a known bottleneck in scaling real-world testing for physical systems.

Compounding for Volumes would manifest as a data and credibility flywheel. Early contracts with demanding, technically sophisticated customers would generate capital to fund more complex and diverse capture missions. Each new dataset would improve the company's proprietary capture and processing pipelines, lowering the marginal cost of data production and increasing fidelity. Furthermore, as more AI models are trained on Volumes' data, the company's output becomes a benchmark; its data formats and metadata schemas could evolve into an informal industry standard for representing spatiotemporal reality, creating significant switching costs for customers who have built their training infrastructure around it. The initial evidence of this flywheel is not yet public, but the company's product roadmap,progressing from raw capture to synthetic generation,suggests a built-in path to use early assets into scalable, higher-margin offerings.

The size of the win, should the company capture a leading position, can be framed by looking at the value of foundational data infrastructure in adjacent fields. Palantir's Foundry platform, which organizes and analyzes complex data for government and enterprise, reached a public market capitalization exceeding $40 billion [public filings]. While not a direct comparable, it illustrates the enterprise value assigned to a system that becomes the trusted source of truth for critical decision-making. A more focused comparable is Scale AI, a provider of data labeling and evaluation services for AI, which was valued at over $7 billion in its 2021 funding round [Crunchbase]. If Volumes executes on the "Foundational Data Partner" scenario and secures a similar position for physical AI data, a valuation in the multi-billion dollar range is a plausible outcome (scenario, not a forecast).

Data Accuracy: YELLOW -- The opportunity analysis is based on the company's stated product claims and target market, which are confirmed by its primary website. The growth scenarios and comparables are illustrative projections based on the company's positioning and known market dynamics in AI infrastructure, not on confirmed commercial traction.

Sources

PUBLIC

1. [Volumes, retrieved 2024] Volumes - Capturing Reality | https://www.volumes.cloud/

2. [Crunchbase, retrieved 2026] Seed Round - Ellis - Crunchbase Funding Round Profile | https://www.crunchbase.com/funding_round/joinellis-seed--475e0042

3. [LinkedIn, retrieved 2026] Chet Ellis - Stealth | LinkedIn | https://www.linkedin.com/in/chet-ellis-74082a1a6/

4. [evercoast.com, retrieved 2026] Evercoast | Multicamera, 4D Spatial Video | Enterprise Volumetric Video | Gaussian Splatting | https://evercoast.com/

5. [Grand View Research, 2023] AI Training Data Market Size Report, 2023-2030 | https://www.grandviewresearch.com/industry-analysis/ai-training-data-market-report

6. [MarketsandMarkets, 2023] Synthetic Data Generation Market - Global Forecast to 2030 | https://www.marketsandmarkets.com/Market-Reports/synthetic-data-generation-market-136155896.html

7. [4DViews] 4DViews - Professional Volumetric Capture | https://www.4dviews.com/