Restricted Access
Limited Partner Materials
Enter the access code provided to you by Desert Forge Ventures.
Incorrect access code. Try again.
DFV Fund I · Confidential
Enter the access code provided to you by Desert Forge Ventures.
A Las Vegas–based company building centralized robot data farms to supply the training data the robotics industry cannot collect at scale on its own. DFV Fund I executed a $150,000 SAFE on May 1, 2026. This SPV gives Limited Partners the opportunity to invest on identical terms, alongside the fund.
Every major robotics company faces the same constraint, and simulation alone cannot solve it for tasks involving physical contact, force, or deformable materials. Amplibotics is the dedicated supplier of foundational physical AI datasets for the broad market that vertically integrated players like Tesla and Boston Dynamics will not serve — because their data is collected exclusively for internal use.
Over $15B has flowed into humanoid and robotics companies in the last two years. Every one of them needs training data. Most cannot collect it at scale internally. Scale AI has already indicated a preference to buy outsourced physical data rather than build the infrastructure in-house, which is the strongest available validation of the thesis.
Note on valuation. The original February investment memo modeled a priced seed round at a $25M–$50M post-money valuation. The round closed instead as a $16M post-money cap SAFE on the same terms as the existing pre-seed financing structure. Co-investing LPs receive identical economic terms to DFV Fund I and benefit from the lower entry point.
The founders ran Haddington Dynamics — acquired by Ocado for $25M in 2020. Every team member came from Haddington, minimum three to four years working together. Among the most sophisticated founders DFV has met in the Nevada ecosystem.
Hardware-agnostic data infrastructure. The data holds its value regardless of which robot form factor or which foundation model wins. The vertically integrated players (Tesla, Boston Dynamics) keep their data internal — that validates and expands Amplibotics' addressable market.
AgiBot, the largest Chinese player, licenses datasets non-commercial only — eliminating it as a commercial substitute. Western enterprise buyers in defense, automotive, healthcare, and government cannot verify provenance of Chinese-sourced data. US-sourced, auditable data is structurally preferred.
Fully loaded cost of approximately $122 per hour of sellable data. Operating cost approximately $21 per hour. Non-exclusive data sells for $50/hr to ten or more customers; exclusive data sells for $500/hr. Every subsequent sale of the same dataset is near pure margin.
12-robot data farm, functional software platform, pilot network, and approximately $9M of pipeline — all on $1.25M of pre-seed capital. First machine operational January 2026.
Praxis Ventures (pre-seed lead, 12.5%) is continuing into this round with $500K. Praxis previously backed Proto Labs ($1.8B exit) and meets weekly with the Amplibotics team.
Across five named data customers and one real-world deployment partnership, spanning two revenue streams (data sales and integrated deployment).
| Customer | Type | Status |
|---|---|---|
| Scale AI | Data buyer | Committed to a ~500-hour initial purchase. Likely first revenue. |
| RichTech | Buyer + Reseller | Publicly traded ~$1B market cap. Active engagement. Interested in resale rights. |
| Bright Machines | Data buyer | Multi-million dollar deal potential. Raised $400M. |
| OpenAI | Data buyer | Re-engaging now that 200 Hz capture is achieved. 1,000 hours needed before re-engagement. |
| Big Robot | Data buyer | In active discussions. Will pay for pilot time and resulting data. |
| Costco (via EIS) | Deployment | Phase 1 warehouse automation pilot in progress. |
The path is more concrete than the original memo suggested. The founders have laid out three time horizons:
Keep the existing 12-robot fleet and add graveyard and weekend shifts. Las Vegas talent pool supports it — 132 pilot applications in two weeks, many requesting overnight shifts. Current facility supports 6–10 simultaneous on-site pilots.
Add five more robot racks to bring capacity to 36 stations. Requires a larger facility (current rent $2K/month; target ~$25K/month) with cooling and bandwidth for 24/7 operation. Pilots transition to remote work from US homes or lower-cost markets.
Replicate the data farm across geographies to optimize energy costs, real estate, and proximity to pilots and customers. Funded by a Series A, targeted for early 2027.
Synthetic data, egocentric capture, UMI, toy robots. Customers have not switched to any of these. The thesis: physical AI follows self-driving — real-world data will always be required regardless of how much synthetic data exists.
Standard Bots, Franka, AgiBot. Each produces hardware-locked datasets designed to sell their robots. Amplibotics runs a mixed fleet across manufacturers, producing data that generalizes across hardware — a long-term structural advantage.
The largest competitor. Only the best-funded companies can do this well. Amplibotics' wedge: serve the long tail of well-funded but not-Tesla-scale robotics companies, and undercut in-house economics for those who can.
First machine operational January 2026. Scale AI initial purchase pending. Approximately $300K cash and $50K monthly burn at original diligence; this SPV and the broader round extend runway materially.
12 robots may not establish a durable position quickly enough. Mitigated by the staged scaling plan and by the fact that 100–200 stations would make Amplibotics the largest physical AI data operator in the world.
No dedicated business development hire. Mitigated by founder pedigree, inbound interest from named customers, and the integrator model (EIS) for real-world deployments.
Training data is not patentable. Defensibility rests on data volume, quality, and capture speed. The 200 Hz capture threshold is a real near-term technical moat.
The round closed at a $16M post-money SAFE rather than the $25M–$50M priced seed originally modeled. Co-investing LPs benefit from the lower entry point. The business has progressed materially since February (200 Hz milestone, OpenAI re-engagement, pipeline growth, first machine operational).
LPs invest via SAFE, which converts at the next priced round or liquidity event. No board seat, no voting rights. Standard non-participating preferred behavior on liquidity. DFV holds pro rata rights via side letter, which the SPV will exercise on behalf of LPs in the next round.
Amplibotics becomes infrastructure for a larger robotics or AI company. The data assets plug directly into the acquirer's model training pipeline. Probable acquirers include Figure, 1X, Physical Intelligence, or foundation model companies expanding into physical AI.
Accumulate the largest robotics dataset, train models once architecture stabilizes, and sell inference tokens — the OpenAI model applied to physical AI. Requires winning the data accumulation race and timing model development correctly.
Both paths scale with total hours of data collected. The more data accumulated, the stronger both paths become.
Submit your indication of interest below. We will follow up with subscription documents within one business day. Capital is called within five days of commitment.
Your submission has been received. We will follow up with subscription documents within one business day.