DePIN Connected | Ep. 26 | CEO Nils Pihl with Auki Network

DePINConnection avatar
DePIN Connection
@DePINConnectionSep 25, 2025

The Spaces convened a focused discussion on robotics, geo-spatial DePIN, and Alki’s push to build a Real-World Web for robots and AR. Host Brad interviewed Nils (Alki) on their breakthrough: a humanoid robot navigating a large indoor venue it had never visited, using a collaboratively built 3D map processed on community GPUs. Nils detailed Alki’s polycentric, privacy-first architecture that lets venues self-host visual positioning and digital twins, contrasting it with centralized data-collection models. The talk covered near-term deployments in retail and hospitality, the ‘Cactus’ app and staff co-pilot, China’s dominance in robotics, and Alki’s funding plan ($75M now; scale to $0.5–$1B by 2027–28). Nils argued robots need hyperlocal, low-latency compute rather than decentralized GPU training, and predicted semi-humanoids in grocery stores by 2026, with ‘robot skill app stores’ emerging. The room highlighted geospatial DePIN winners like GeoNet, Helium/XNET, Hivemapper, Wingbits, and closed with Q4 picks (Alki, GeoNet, 375 AI), and macro notes on AVAX/Solana.

Robotics + DePIN Twitter Space: Alki’s Real‑World Web, Hyper‑Local Compute, and Geospatial DePIN

Speakers and roles

  • Brad (Host; content creator running livestreams/YouTube; DePIN investor)
  • Nils (Alki Robotics; founder/lead; robotics and spatial computing)
  • Tyler (Contributor; DePIN enthusiast)
  • William (Contributor; DePIN analyst)
  • Suburban (Contributor; DePIN practitioner)

Market backdrop and momentum

  • Robotics and decentralized infrastructure are entering an acceleration phase. A brief market correction was framed as healthy.
  • Notable signals:
    • Nvidia’s Jensen Huang has highlighted robotics as a next major market.
    • Figure Robotics reportedly raised $1B at a ~$40B valuation.
  • Expect near-term mainstream adoption via autonomous vehicles, with broader robotics following.

Alki: vision, architecture, and privacy stance

  • Mission: build collaborative spatial computing—a “Real‑World Web” (decentralized machine perception network) that makes physical venues browsable, searchable, and navigable for AI, robots, and AR devices.
  • Core idea: robots/glasses/phones collaboratively map and understand physical spaces so devices can “browse” the physical world like the internet enables browsing digital content.
  • Visual positioning: sensors alone (cameras/LiDAR) are insufficient; robots also need maps (analogous to self-driving cars requiring HD maps). Maps answer where the bathroom is, where aisles/products are, and provide consistent spatial context.
  • Polycentrically decentralized design:
    • Not a Filecoin/torrent model. Different components reside uniquely under local control.
    • Venues can fully self-host their data; nothing must reside on Alki’s infrastructure.
    • Blockchain supports discovery/authentication of the correct hyper‑local resource.
    • Strong privacy posture: avoids centralized collection of camera streams and digital twins of private spaces (addresses enterprise/consumer reluctance). Nils cited Elon Musk’s claim Optimus would be the “greatest source of data”; Alki aims to make surveillance-centric approaches unnecessary.
  • Differentiation from data‑collecting DePINs:
    • Projects like Hivemapper, Natix, and OVR pay contributors and centralize data. Alki neither pays for nor collects venue data.
    • Instead, Alki enables venues/cities to run their own bounties and own their data (e.g., City of Hong Kong funding its map via Alki’s protocol).
    • Market inefficiency avoided: centralized networks struggle to price data per-location; Alki leaves valuation and procurement to the free market (venue-level demand).

Breakthrough: first large-scale indoor navigation of a never‑before‑visited venue by a humanoid robot

  • Demo (two weeks prior):
    • Phones captured the conference venue; a few markers placed.
    • Upload to Alki network; processed on community nodes using consumer GPUs.
    • Produced a 3D map usable by a humanoid robot, which navigated on arrival despite never physically visiting beforehand.
  • Claim: first in robotics (not just Web3 robotics)—a significant lead in collaborative machine perception.

Retail use cases and measured ROI

  • AR sticky notes/co-pilot for staff:
    • Pilot in a 50,000 sq ft supermarket with 30,000 SKUs (Northern Europe) saved at least 15 min per employee per day on handovers.
    • Demonstrates immediate operational value once devices understand spatial context.
  • Cactus (web app):
    • Customers/staff scan a QR code to access a browser-based navigator.
    • Dotted path guides to specific products; improves in‑store experience and staff efficiency.
  • Hospitality: Alki expects upscale hotel deployments in the West relatively soon.

Market size, geography, and partners

  • China dominance:

    • 50% of global robot sales and installed base already in China; gap widening rapidly.

    • Drivers: demographic crisis (aging population post one‑child policy), strong government mobilization/funding, manufacturing scale, and strategic focus (robots/drones in warfare).
  • Alki partners: SlamTech, Unitary, Engine AI, PadBot, and others.
  • Addressable market:
    • ~500,000 supermarkets/grocery stores globally.
    • Goal: deploy to ~100,000 locations by end of 2028.
    • Economics:
      • Robot ROI: ~13–16 months; then ~$40,000/year per robot.
      • Handheld co‑pilot: targeting ~$300/month/location; at 100k locations implies ~$360M annual recurring revenue.

Funding roadmap and scale-up plan

  • Near-term raise: $75M (in SF) to prove scaled distribution (~500 robots) and validate unit economics by 2026–2027.
  • Follow-on raise: $0.5–$1B to deploy 10,000–20,000 robots once unit economics and distribution readiness are proven.

Robot form factors, capabilities, and timelines

  • Semi‑humanoid (wheeled base, no legs) for better battery/stability in store environments: targeted deployments by end of 2026.
  • Walking/talking humanoids: feasible within months for tour‑guide style roles.
  • Manipulation tasks: VLAs (Vision‑Language‑Action models) are a major breakthrough accelerating progress, but general‑purpose physical intelligence remains years away; tasks require specific training/policies.
  • Anticipated “robot app store”: purchase discrete capabilities (fold laundry, do dishes, etc.) before broadly general physical intelligence.

Compute, latency, and DePIN hot takes

  • Hyper‑local compute is essential:
    • Robotics/drones require ultra‑low latency spatial computing (<8 ms, ideally <4 ms); battery life and responsiveness benefit from nearby offloading.
    • Aggregated idle cloud compute lacks the locality/latency needed for robotics.
  • Skeptical on decentralized GPU training for LLMs:
    • The “biggest data center wins” arms race isn’t relevant to physical AI.
    • Nils believes decentralized training is a red herring for robots/drones and that such web3 AI narratives are often disingenuous to retail investors.
    • Notes potential misallocation in massive centralized training capex if inference becomes dominant and cheaper (e.g., DeepSeek-like efficiencies).
  • Bullish DePIN segments:
    • Geospatial components: GeoNet (very high praise; Mike’s leadership), Hivemapper, Natix.
    • Hyper‑local compute networks.
    • Connectivity with geospatial distribution: Helium, XNET.
  • Bearish DePIN segments:
    • Non‑geospatial decentralized GPU aggregators (e.g., Render, Aethir) face race‑to‑bottom and eventual obsolescence.

Digital twin upkeep

  • Near-term: venue staff maintain the twins.
  • Medium-term: visitors wearing camera glasses (e.g., upcoming consumer devices) sell RGB streams back to venues for twin updates, with venue-set budgets via Alki’s protocol.

Additional technical notes

  • Floor cleaning robots typically build 2D maps (walls/shelves) and often lack world-facing cameras; useful but insufficient for product-level mapping.
  • Expert commentary: Steve (ROS Nav2 core contributor) noted that even $75M a year ago wouldn’t crack basic manipulation in grocery environments; VLAs shift timelines forward, enabling simpler tasks sooner.

Macro predictions and socioeconomics

  • Bold 10-year prediction: average Chinese household spends more on robots than cars; Western households follow within ~20 years.
  • Reasons China leads: demographic labor shortfall, government-funded campuses/facilities, unmatched production scale, and strategic military emphasis on robots/drones.
  • Labor and inflation:
    • Robots will take many jobs; capitalism may need a “software upgrade” (DePIN enables individuals to fill roles traditionally held by corporations); UBI may be considered.
    • Nils questions traditional inflation framing given rising life-quality bundles; William counters that technology is inherently deflationary; Brad notes fiat expansion and widening wealth gaps complicate the picture.
  • Gig economy disruption: autonomous vehicles likely displace ride-hailing incomes in coming years; Tesla “taxi mode” could let owners monetize idle time, but affordability divides persist.

Ecosystem mentions and collaborations

  • Tashi team (JMR, Ken): robot coordination; consensus in real-time mapping.
  • Hivemapper (Ariel): SF-based; mapping via camera streams.
  • GeoNet: correction data/survey-grade positioning; “single source of truth” for autonomy; rovers with LiDAR mapping cities; sustained network growth and token burns.
  • Wingbits: drone-related DePIN.
  • Daba: hotspot onboarding; teasing aviation expansion.
  • Conduit: potential disruptive play (suggested invite for deeper dive).
  • 375 AI (a.k.a. 3,75 / 3,70,5 AI): Q4 token launch; edge data collection via Upfront Media’s 50,000 billboards (six cameras each; Nvidia edge compute).
  • Avalanche (AVAX) and Solana: large-cap ecosystems with favorable catalysts (spot ETFs, digital asset treasury movement, market structure bill).

Investment picks and perspectives (Q4 focus)

  • Brad: Alki (rebalancing heavily), GeoNet (largest DePIN holding), 375 AI (edge data; strong founder track record). Large-cap conviction: AVAX, Solana.
  • Tyler: Alki, GeoNet, 375 AI, Deep tokens.
  • William: Alki, Deep, GeoNet, 375 AI, Toshi Network; plus AVAX; Helium and XNET for 5G and offload; overall DePIN bullishness.
  • Suburban: Deep (executing, revenue), Wingbits (drones), Daba (hotspots; aviation tease), Conduit (sleeper). Notes personal ramp in local GPUs; energy constraints.

Key takeaways

  • Alki achieved a first-in-robotics demonstration: pre-mapped, large-scale indoor navigation for a humanoid robot, processed on community GPUs.
  • Alki’s Real‑World Web is privacy-first and polycentrically decentralized; venues self-host and control their spatial data.
  • Retail is immediate and measurable: AR notes/coprots save time; Cactus app improves navigation; hotels next in line.
  • China is the epicenter of robotics adoption; Alki is deeply partnered there.
  • Funding path is clear: $75M to validate scale/unit economics, then $0.5–$1B to deploy 10–20k robots.
  • Robotics compute will be hyper‑local and geospatial; DePIN winners emphasize locality and positioning; generalized decentralized GPU aggregation is structurally mismatched to robotics’ latency needs.
  • Societal impact: expect gig economy displacement via autonomy; broader reconsideration of economic structures as physical AI scales.

Actionable notes

  • Follow Alki for demos and updates; expect more proof points on semi‑humanoid deployments and hotel pilots.
  • Watch GeoNet’s growth (base stations/rovers/token economics) and geospatial DePIN plays (Hivemapper, Natix).
  • Track 375 AI’s Q4 token launch and billboard-edge data rollout.
  • For investors: overweight geospatial/locality-centric DePIN; be cautious on non‑local GPU aggregation narratives.
  • Expect continued policy and capital markets catalysts (spot ETFs, treasury adoption, market structure bill) benefiting AVAX/Solana ecosystems.

Closing

  • Nils departed to continue the $75M raise; strong community support and admiration for execution.
  • Brad will host an upcoming interview with Nils; recording of this Space is available and encouraged to be shared for broader exposure to Alki and the robotics + DePIN narrative.