Theseus Robotics, a seven-person startup from Y Combinator's Fall 2026 batch, has set its sights on a problem that has vexed grocery chains for years: getting products onto shelves efficiently and picking online orders without breaking the bank. The Switzerland-based company joins a growing roster of ventures betting that humanoid robots can solve retail's persistent labor shortage.
The need is real enough. Grocery store productivity dropped 3.1 percent in 2025, according to the Bureau of Labor Statistics, even as overall retail trade productivity climbed 2.9 percent over the same period. Progressive Grocer reported in July that labor continues to rank as the top concern for grocery operators. A June 2026 McKinsey report, "State of Grocery in North America," found AI beginning to deliver tangible value across the grocery supply chain, particularly in labor productivity and fulfillment efficiency.
Theseus describes its mission on its website as "automating the last 30 metres of retail fulfilment" through shelf restocking, stock rotation, and order picking. CEO Samarth Malhotra wrote on LinkedIn that the company is creating "the brain for the Retail AI Robot: precise manipulation, a careful touch, and a mind that learns every store it meets." CTO Kristóf Floch posted that "a robot that only works when nothing goes wrong won't last a single night shift."
Whether Theseus can deliver on that promise is another matter entirely.
Where the Robots Actually Are
Shelf-scanning robots have achieved commercial scale. Simbe Robotics reported in a September 2026 press release that it had surpassed 3,000 autonomous units under contract, with deployments across Wakefern (the ShopRite cooperative), Schnuck Markets at 62 locations, and B&R Stores in Nebraska. Brain Corp, according to its August 2026 press release, logged 68 percent growth in global deployments and 5.3 million autonomous hours in the first half of 2026 alone.
These wheeled machines—Simbe's Tally and Badger Technologies' Marty among them—scan aisles for pricing errors and out-of-stock items. They generate intelligence for stores. What they don't do is restock.
Humanoid robots capable of manipulation remain largely experimental when it comes to retail applications. HIVE Robots is testing Unitree G1 humanoids at a Føtex supermarket near Copenhagen with hopes of achieving autonomous restocking capability by early 2027. The robots currently operate under human guidance while collecting training data. RockingRobots reported in October that Føtex employees spend 40 to 50 percent of their working time restocking shelves.
Sanctuary AI completed a week-long pilot at a Mark's retail store in Langley, British Columbia in March 2023, performing 110 retail tasks under human supervision. The company has since pivoted toward industrial deployments. Agility Robotics, which unveiled its Digit 5 humanoid in September 2026 and reported $300 million in orders as of May, has focused commercial efforts on warehouse logistics with Amazon and other partners rather than in-store retail.
"It's brutally hard to build robotics that operate reliably across thousands of stores for the world's largest retailers," said Aidan Madigan-Curtis, a partner at Eclipse Ventures, in Simbe's September 2026 press release. That perhaps understates the challenge.
The Push Behind the Pitch
Market projections tell a seductive story. A 2026 Fortune Business Insights report pegged the retail automation market at $31.21 billion in 2026, forecasting growth to $77.36 billion by 2034. The smart retail market shows even steeper trajectory: from $74.75 billion in 2026 to $637.01 billion by 2034, a 30.71 percent compound annual growth rate.
Goldman Sachs Research projects the broader humanoid robotics market could reach $38 billion by 2035, with some Goldman commentary citing potential shipments of 6.5 million units by that year. Stanford's Emerging Technology Review 2026, published in February, forecasts that as safety and dexterity improve, humanoids will expand into hospitality and retail.
Labor scarcity persists as a chronic constraint rather than a cyclical pressure, giving grocery chains reason to explore alternatives. Fulfillment costs for online grocery orders remain structurally uneconomical for most retailers without automation. Kroger operates five Ocado automated customer fulfillment centers as of 2026, after closing three others in early 2026. Walmart acquired Alert Innovation's Alphabot micro-fulfillment technology in October 2022 and continues building next-generation fulfillment centers.
The technology itself shows signs of maturation through vision-language-action models and teleoperation-to-autonomy pipelines. HIVE's trial with Unitree robots uses data collection under human supervision to bootstrap autonomous behavior, the same approach Figure AI and Apptronik are testing in automotive manufacturing with BMW and Mercedes-Benz.
Amazon CEO Andy Jassy wrote in his April 2026 shareholder letter that "robotics provides a step-level change for how we can deliver faster, reduce the cost of carrying more selection, and automate movements that cause strains and injuries to our teammates."
Still, intent and execution rarely follow the same timeline in robotics.
What Theseus Hasn't Said
The startup has not disclosed pilots, customer names, or unit deployments. Its LinkedIn page shows 536 followers and lists company size as two to ten employees. Y Combinator's batch partner is Kulveer Taggar. The founding team includes CEO Samarth Malhotra, CTO Kristóf Floch, and CPO Ádám Weinhardt-Kovács.
Theseus emphasizes dual-use capability: overnight restocking when stores close and daytime order picking for e-grocery fulfillment. Malhotra wrote that "because we touch every item, we give stores visibility into what's really on their shelves." That claim positions the company against both intelligence robots like Simbe and logistics humanoids like Agility's Digit.
Simbe, by contrast, has disclosed multi-year contracts across regional and national banners. Wakefern scaled up Store Intelligence deployments in 2024. Schnuck Markets expanded from an initial 15 stores chainwide. Harmons in Utah deployed Tally across all locations after a pilot completed in October 2025. The robots run multiple scans per day but rely on human associates to act on the data.
Agility Robotics has accumulated 65,000 hours of real production experience, according to its website, and filed an S-4 for a SPAC combination with Churchill Capital XI in October 2026. An exhibit in that filing shows illustrative revenue per robot over five years ranging from $400,000 to $500,000 under both robotics-as-a-service and ownership models, though the company cautioned those figures are management estimates rather than contracted terms.
Figure AI is piloting its humanoid at BMW manufacturing facilities in Spartanburg, South Carolina, handling tasks like inserting sheet metal into fixtures. Apptronik entered a commercial agreement with Mercedes-Benz in March 2024 to test its Apollo humanoid in production logistics.
These automotive deployments offer something grocery aisles don't: controlled environments with no customers, no product variety changes, and defined workflows.
IEEE Spectrum cautioned in a 2025 analysis that translating AI progress to reliable multipurpose robots remains the gating challenge. Agility's Oregon factory capacity exists, but unit economics and safety at scale are unproven outside narrow logistics tasks. IEEE Robotics and Automation Society wrote in September 2026 that Digit 5 "may be the first humanoid robot worker that's truly safe," underscoring that even leading vendors treat full human-robot collaboration as a future milestone rather than a current capability.
The Reality Check
Retail humanoid deployments face technical, economic, and regulatory hurdles that shelf-intelligence robots have already cleared. UL 3300 for service robots in commercial spaces reached a 2024 revision and appears on OSHA's list of appropriate test standards. The EU Machinery Regulation 2023/1230 becomes applicable January 20, 2027, governing service robots in European stores.
Biometric privacy rules add complexity. Illinois BIPA litigation continues across retail use cases. Chicago delivery robot companies faced class-action filings in October 2026 for alleged biometric capture.
Safety certification routes exist but require validation at fleet scale. Fixed costs for compliance, insurance, and fleet management favor large chains over independents. Grocery operates on thin margins—net profit often under three percent—which makes capital expenditure thresholds unforgiving.
The automation pathway forward likely diverges into distinct tracks. Shelf-intelligence robots will continue scaling through multimodal sensor fusion. Simbe launched Tally Spot in January 2025, combining mobile robots with fixed cameras. Back-of-house micro-fulfillment centers will absorb more automation as Walmart and Kroger refine their systems. Alert Innovation's Alphabot and Ocado's grid-based CFCs already handle tens of thousands of orders per week at installed sites.
Humanoid restocking will remain pilot-heavy into 2027 before any potential inflection. HIVE's target of autonomous restocking in early 2027 at a single Føtex location would represent a milestone, not a deployment wave.
Theseus and competitors must prove not just that a robot can place a can on a shelf, but that a fleet can do so overnight across stores with different layouts, product mixes, and staffing models while generating positive ROI versus hiring third-shift workers. That's a considerably taller order.
McKinsey's March 2026 consumer survey of 4,989 respondents found that willingness to use AI decreases as automation levels rise. Customer acceptance lags technical capability. Labor unions watch closely. The Los Angeles Times reported in October that grocery workers in Southern California express concern as Tally robots roll through Ralphs and other chains.
The Long Game
Founders betting on humanoid retail face the same adoption curve that self-driving cars encountered: the last ten percent of reliability takes longer than the first ninety.
Investors should watch for named pilots with disclosed unit counts, verifiable uptime, and transparent economics. Until then, "automating the last 30 metres" remains a compelling pitch and an unproven business. Theseus may eventually deliver on its vision of robots that learn every store they meet. For now, the company joins a crowded field where ambition runs well ahead of deployed reality.
