Yosemite Valley, black and white
2026 · CONFIDENTIAL INVESTOR MEMO · FOR ACCREDITED INVESTORS ONLY

CollectedAI

Automating garment manufacturing — the Foxconn of fashion

We're bringing garment manufacturing back to American soil with robots, then using that same platform to sell directly to brands and to our own label — capturing the full margin of a $2 trillion industry, not just the sourcing fee.

Yosemite Valley, California
01The Opportunity

Clothing is still made almost entirely by hand. We're building the robots that replace that labor — and running our own factories with them.

A $2 trillion global industry running on labor that is increasingly expensive, unreliable, and far from home. CollectedAI is building the robots that replace that labor and using those robots to run its own clothing operations. Think of it like Tesla: instead of selling the technology to other car companies, Tesla built its own cars. We're doing the same with garment factories — proving the technology in Vietnam, then bringing automated factories home to the United States, where American-made, robotically produced clothing can undercut offshore competitors while CollectedAI keeps the entire margin.

The tech we're developing for fine-grained robot manipulation is among the most valuable in the world right now. No one else has a real-world data flywheel for it.

New Jersey

R&D + Brand Pilot

Our robot lab and small pilot factory. Engineers train robots on sewing tasks, iterate on designs, and produce initial brand samples. Closed loop: every failure teaches the system something. Brand lead Jan Nusbaum drives U.S. brand development and launch.

Da Nang, Vietnam

Scale + Data Collection

Our Da Nang facility — 35 staff, growing to 120 — runs live production while we deploy robots alongside human workers. High-volume, real-world data collection at scale. Revenue-generating from day one. Proof point for the larger raise.

Where we stand today
Next milestone: a robot performing a sewing task fully autonomously. This validates the core thesis for the larger institutional raise.
02Progress

The lab, the factory floor, and the robots — as they stand today.

Unedited footage and photos from New Jersey and Da Nang. Videos are sped up so the long clips read as quick loops — mute is on by default.

NJ Robot Lab team
NJ Robot LabThe engineering team
NJ Robot Lab workstation
NJ Robot LabRobot arm workstation
NJ Robot Lab setup
NJ Robot LabTraining setup
Da Nang factory floor
Da Nang, VietnamFactory floor
Da Nang factory team
Da Nang, VietnamThe floor team
Closing the Da Nang lease
Da Nang, VietnamClosing the factory lease
Robot PolicyHanging a shirt — real-world arm
Embodiment TransferIn-painting research — multi-camera view
SimulationPolicy checkpoint comparison (R2 vs R3)
03Team & Investors

Frontier robotics, run by people who know fashion and factories from the inside.

Maxwell Nusbaum
CEO & Co-Founder
Business development, brand strategy, factory operations, and investment lead.
Alexander "Sasha" Khazatsky
CTO & Co-Founder
Stanford PhD; lead author of DROID, the largest and most diverse robot manipulation dataset of its time. Leads all robot and technology development.
Jan Nusbaum
Brand Lead
Former SVP, Tommy Hilfiger / President, Nanette Lepore. Leads U.S. brand strategy and go-to-market.
Hung Xuan Nguyen
Factory Lead / Operator
20-year veteran across Adidas, Hanesbrands, and Coats; built and scaled factories from the ground up. Future head of our Vietnam scrubs factory.
Maya Alva
Data Collector
Runs hands-on data collection with our robots — capturing the fine-grained manipulation demonstrations that train our production policies.
Leonard John Lee
Lead Automation Engineer
15 years building intelligent factory automation systems from the ground up.
Varun Giridhar
Robot Learning Engineer
MS, Georgia Tech; trained in Animesh Garg's robot learning lab.
Ayush Tarachandani
Robot Learning Intern
MS student, Carnegie Mellon University. Works in Katerina Fragkiadaki's robot learning lab.
Sepehr Nasiriany
Robot Learning Intern
MS student, Columbia University. Works with Professor Yunzhu Li on robot manipulation.
Backed by
Jeff Dean
Chief Scientist, Google
$750K anchor investor · priced this round
Pieter Abbeel
Head of Robotics, Amazon
Professor of Robotics, UC Berkeley
Ashvin Nair
Engineer, OpenAI / Cursor
Google Cloud
$350K compute grant
non-dilutive, extending our GPU runway
04The Round

The next angel extension — already priced by Jeff Dean.

$800K raised from angels plus $180K from founders to date. This raise extends that same angel round.

Target raise
$2.0M
SAFE · this extension round
Post-money cap
$30.56M
Priced by Jeff Dean, Chief Scientist, Google
Raised to date
$980K
$800K angels + $180K founders
Use of funds
Engineer Salaries
~35%
Recruiting and retaining world-class robot learning engineers — our core competitive advantage.
Factory Operations
~25%
Maintaining NJ and Vietnam facilities until each becomes self-sustaining through brand and B2B revenue.
Robotic Equipment
~15%
Arms, sensors, cameras, and custom hardware for the R&D lab and production floor.
Compute
~15%
GPU clusters for training neural networks and diffusion models on our proprietary sewing data.
Brand Development
~10%
Design, production, and U.S. market launch of our scrubs and clothing line.
05Roadmap & Vision

From garment-production data to U.S. robotic manufacturing to full vertical ownership.

Production data improves the robots; the robots improve manufacturing economics; B2B clients validate quality, cost, throughput, and reliability; the owned brand then turns the same manufacturing advantage into consumer upside.

$79.3B
U.S. apparel imports — the largest single-country apparel import market in the world, mostly sourced from Asia
35.1%
Average U.S. apparel tariff rate as of December 2025, up from apparel being just 2.5% of import value but 15.6% of tariff duties collected
$1.5B+
Tariff exposure reported by Nike alone; Gap reported a 100–110 bps margin hit worth $150–175M
2016
Ever since 2016, nearshoring has been a stated priority for U.S. & European apparel executives — actual nearshore share has stayed flat, held back by cost and capacity
01

Data lab in the U.S. and a Vietnam operating base

Primary proofRobot lab live in NJ, two Vietnam factories, first production-linked dataset, NC factory-owner conversations underway.
02

Scale the Vietnam data engine

Primary proofMore operators, more task variation, a real revenue base, U.S. factory financing preparation.
03

Robotic demos at the North Carolina factory

Primary proofReal garment demos, a client pipeline, material research.
04

First U.S. robotic factory

Primary proofMedium-scale modular factory/campus with buyer-facing systems.
05

Established-brand production

Primary proofCommercial B2B proof with cost, lead-time, quality, and tariff-avoidance evidence.
06

Scale the platform + prepare the brand

Primary proofMore cells, material partners, first owned-brand collection, DTC infrastructure.
07

Dual-track production + online brand launch

Primary proofB2B utilization running alongside owned-brand online drops.
08

Larger U.S. platform + upstream optionality

Primary proofFactory network/campus planning, fabric-mill build-vs-buy analysis.
09

Global manufacturing + full vertical expansion

Primary proofRepeatable U.S. playbook, first European factory case, broader brand portfolio.
Strategic conclusion

CollectedAI's roadmap follows a staged industrial strategy: data, then robotic garment production, then U.S. manufacturing for established brands, then the owned brand, then full vertical expansion through additional factories, upstream assets, and a global rollout. We bring garment manufacturing closer to the American market at a cost structure that historically required Asia — then use that same platform to sell better-designed, better-priced products directly to consumers. This turns the upside of a U.S.-based CollectedAI into more than a patriotic story.

06Paths Not Taken, for Good Reasons

Eight directional decisions, run through the same evaluation framework, before we committed capital.

Before committing capital, CollectedAI and its advisors ran eight distinct directional decisions through the same evaluation framework: pain intensity, technical feasibility, deployment friction, buyer potential, ROI visibility, and expansion potential. This is that record — the full evaluation summary for each decision, unedited. Click any row to collapse it; the full 20+ page analysis is linked at the bottom.

01

Software vs. Production

Operate production directly+
Evaluation summary

Software licensing is the easier-looking path, but it is the weaker business model. The target market for robotic sewing software is weak due to limited R&D budgets, and incumbents face little urgency to adopt robotics when manual labor remains cost-effective. Internal deployment of the same software gives CollectedAI full control over hardware, workcell design, data feedback, and deployment quality, while opening a much larger commercial opportunity.

Owning production is more demanding, but it is where the real value sits. Leveraging robotic arms with our proprietary software is our competitive edge. It is capable of outworking humans due to its 24/7 robotic nature. It is more scalable, factories can be built more efficiently. Taking out the human hiring cost makes traditionally expensive areas suddenly viable again for manufacturing. These factories do not yet exist, CollectedAI will be the first of its kind.

02

Production Geography

United States, primary commercial market+
Evaluation summary

Asia preserves the supply chain problems that make offshore sourcing painful, and Europe lacks a clear cost advantage while carrying restrictive regulations. The United States offers the strongest thesis: a massive consumption market, a documented appetite for nearshoring, and a tariff structure that makes domestic production financially compelling when robotics reduces the labor share. North Carolina emerged as the most practical starting location due to its existing textile workforce, mill infrastructure, and university research support. The trade-off is higher upfront capital against a superior landed-cost and working-capital case that offshore manufacturing cannot match under the current trade regime.

03

Factory Scale Model

Modular medium-to-large campus model+
Evaluation summary

One single gigafactory would maximize standardization but scares customers as it concentrates operational, geographic, and capacity risk. Many scattered microfactories would improve flexibility but may struggle with unit economics, maintenance consistency, and real production scale. The stronger model is a controlled, modular manufacturing platform: start with one flagship U.S. campus with multiple medium to large factory units, then replicate standardized cells or medium-sized facilities as demand proves itself.

04

Vertical Integration

Yes — as a later-stage layer+
Evaluation summary

Retail margin is where the largest profit pool sits, and the cross-industry examples demonstrate that owning both production and the consumer relationship is viable when the economic and operational fundamentals are right. Robotic manufacturing is the wedge that makes this possible in apparel in a way that previous attempts could not sustain. Legacy companies like American Apparel failed because human labor at U.S. wages could not compete with Asian production costs. AI-driven robotics alters this equation by making U.S. manufacturing cost-competitive without tariffs and with faster turnaround. This shift validates the full vertical model as viable in a way that has never been possible before.

05

Customer Sequencing

Established brands first+
Evaluation summary

Manufacturing for established brands addresses a clear and urgent market need: brands face rising tariff costs, supply chain pressure, and a restructuring import landscape that creates immediate demand for a cost-competitive domestic alternative. Launching an owned brand from the start would add design, marketing, and consumer acquisition burdens before the manufacturing system is proven, while also requiring consumer education on the value proposition.

Serving established brands first builds production credibility, operating cash flow, and manufacturing capability, creating a foundation from which an owned brand can later be launched with lower risk. The economics of operating both a factory and a consumer brand are fundamentally different, and the added complexity of design, merchandising, inventory risk, and retail execution should not be taken on before the manufacturing system is proven. The careful path is to treat the owned brand as a later-stage opportunity, not the first step.

06

Dual-Track Production

Hybrid — one platform, two demand channels+
Evaluation summary

The optimal decision is to manufacture for other brands while developing and growing the owned brand. The market is not winner-take-all, and the strongest precedent supports a dual model: B2B production for utilization and credibility, owned brand for margin and strategic expansion.

07

Brand Differentiation

"Thrift-store diamonds" — 60s/70s designs, modern sizing+
Evaluation summary

The concept our Nusbaum dual team will pursue is mass-producing beautiful "thrift-store diamond" inspired designs from the 60s and 70s. Garments with the emotional appeal of rare vintage finds, but made available in modern sizing, consistent quality, and at substantially lower prices than boutique vintage-inspired brands. The concept has precedent in vintage-inspired brands and archive reissues, but the larger opportunity is to combine that aesthetic with a manufacturing system that can make rare-looking designs scalable.

08

Brand Roll-out

DTC-first, not DTC-only+
Evaluation summary

CollectedAI should not launch the future brand as a traditional retail-first fashion company. The strongest rollout is DTC-first, but not DTC-only. Online should be the commercial core because it allows demand testing, waitlists, controlled drops, customer data collection, and fast feedback into the manufacturing system. Physical retail should come later, first through temporary pop-ups or fit showrooms, then through selective wholesale or flagship stores only once demand is proven. This creates a staged rollout that matches CollectedAI's advantage: test demand quickly, produce efficiently, replenish winners, and avoid locking the brand into expensive fixed retail before product-market fit is clear.

07ROI & Financials

The path to $1B+ in annual revenue.

A staged, dual-channel model for robotic U.S. garment manufacturing and an owned consumer brand. Illustrative 5–10 year benchmark case, using public-company scale analogues rather than a top-down market-share assumption — and excluding internal manufacturing revenue for the owned brand.

$443M
External manufacturing
Gokaldas-scale B2B revenue
+
$631M
Owned brand
FIGS-scale consumer revenue
=
$1.074B
Combined annual revenue
external B2B + retail, no double-count

Gokaldas and FIGS represent the two distinct revenue engines in CollectedAI's long-term model. Gokaldas is a modern, diversified apparel manufacturer whose product mix and operating complexity align with our medium-term capability roadmap. FIGS is a leading pure-play scrubs brand — a relevant proxy for the scale a focused, owned apparel brand can achieve while remaining modest relative to the broader fashion market. This is a conservative frame: manufacturing revenue is trending upward, inflation isn't priced in, and CollectedAI's robotics cost advantage is not reflected in either analogue's own economics.

Path to scale
01

Data engine + early revenue. Operate controlled Vietnam production to collect labeled workflow data while serving existing factory clients.

02

Prove robotics + unit economics. Validate cycle time, quality, uptime, labor reduction, and total landed cost through the NJ lab and NC factory environment.

03

Scale U.S. B2B manufacturing. Sell to established brands — scrubs first, then dresses, loungewear, and adjacent products as robotic cells expand.

04

Add the owned-brand layer. Use the same platform to capture retail revenue once manufacturing reliability, capacity, and demand are proven.

Earnings sensitivity at $1.074B revenue
Mature-scale caseNet marginAnnual net profit
Peer-like baseline (rounded)~5.0%~$53.7M
Internal automation upside*27.5% blended~$295.5M

*Based on a 40% reduction in manufacturing COGS, with non-COGS expenses held constant — producing 36.0% manufacturing and 21.5% integrated-brand net margins.

Anticipated subsequent rounds
YearRoundRangeTrigger
2Series A$20M–$30M~20 factory tasks automated
3TBD$45M–$50MEstablished-brand production scaling
08Sources & Research

Every footnote behind the Paths document, the Roadmap, and our financial one-pagers.

Organized by topic and numbered to match how they're cited across our research. Where a source has a public link, it's linked directly.

Additional named sources referenced in our research

Wanjie Textile (garment margin analysis) · Global Growth Insights (garment manufacturing market sizing) · Garment Manufacturing Market 2026–2035 report · McKinsey EU regulatory analysis · McKinsey (procurement & CPO priorities) · Compliance in Global Value Chains (working paper) · Swatch Group (company disclosures) · L'Oréal (company disclosures) · Urban Outfitters (recirculated garment figures) · Mandala (scrubs production cost analysis) · NurseMoneyTalk (scrubs brand sentiment) · Statista (retail & e-commerce data) · State of Apparel 2026 · Ecommerce Fashion in 2026 Trends · Insider Trends (flagship retail analysis) · JLL (U.S. retail market report) · Warby Parker (retail expansion case) · Inditex FY2025 annual results

09The Mission

We're building the company that controls manufacturing for American brands, and captures that same uplift again with our own labels — a disciplined path through a $2 trillion industry.

Software licensing was the easy path and the wrong one. Production is harder, and it's where the value is. We start by manufacturing for established brands, where the tariff pain is real and the ROI is provable today. From that foundation, CollectedAI adds its own brand to capture the margin that currently belongs to retailers — one manufacturing platform, two demand channels, and a staged path to full vertical ownership.

Maxwell Nusbaum
CEO & Co-Founder, CollectedAI
Confidential. For accredited investors only.
collected-ai.com