Eka Robotics

1.1 Identity, product, and stage

Eka Robotics is described as a Cambridge, Massachusetts robotics company developing manipulation systems that combine visual perception, force-sensitive hardware, and learned physical actions. The product thesis is a full-stack approach: robotic hands supply contact information to a Vision-Force-Action model so a robot can adjust how it handles an object. Reporting places the company at the Series A stage and says unnamed industrial pilots are under way for pick-and-place and wire-handling tasks. That evidence supports an early deployment-stage identity, not a proven commercial business model. The assigned sources do not disclose pricing, contract structure, named customers, revenue, or a complete product catalogue, so these remain diligence gaps rather than inferred scale.[CO001, CO002, CO003, CO006, CO013]

1.2 Founders, leadership, and governance

Current reports identify Pulkit Agrawal and Tuomas Haarnoja as the founders. Agrawal is an MIT professor, while Haarnoja previously worked at Google DeepMind and is associated with reinforcement-learning research for robot control. This pairing is directly relevant to a force-aware manipulation architecture and provides strong technical founder-market fit. However, the public pool does not identify a broader executive team, board composition, investor board rights, succession coverage, or operating ownership outside the founders. A current company-issued job repost shows active hiring for senior hardware test leadership, including reliability testing and certification work, but a vacancy does not establish that those capabilities are already staffed or institutionalized.[CO004, CO005, CO013, CO014]

1.3 Funding history and stakeholders

SiliconANGLE reports an $11 million seed round in May 2025 and says investors have put about $161 million into the company after the October 2026 Series A. Multiple sources report that the latest round raised $150 million at a $1.45 billion post-money valuation, co-led by Andreessen Horowitz and F-Prime. TechStartups also names Obvious Ventures, E14 Fund, MFV Partners, Bison Ventures, Glasswing Ventures, Samsung Next, and unspecified strategic investors as participants. These reports establish financing history and investor interest, but not cash available, ownership percentages, liquidation preferences, debt exposure, or board control. The valuation is a negotiated financing outcome and should not be treated as independently verified enterprise value or operating performance.[CO006, CO007, CO008, CO009, CO015, CO016]

1.4 Scale, readiness, and milestone evidence

Public reporting attributes roughly 1,500 object-picking operations per hour and greater than 95% pick success to company-reported tests, with the throughput compared against a reported human benchmark of about 600 operations per hour. Those figures preserve the stated task and attribution; they are not independent production benchmarks. The gripper is also reported as rated for five million duty cycles against a company goal of 20 million. Servicta and SiliconANGLE highlight the unresolved transition from demonstrations to sustained industrial use, including uptime, task breadth, recovery, integration cost, and maintenance. Named pilot results and revenue are absent. The evidence therefore supports a well-financed technical program with promising reported demonstrations, while commercial scale and long-run reliability remain material open questions. No recorded incident in this assigned pool should be interpreted as proof that no incident exists.[CO003, CO010, CO011, CO012, CO013, CO014]

1.5 Exhibits

2.1 Market boundary and buyer problem

The relevant market is not all artificial intelligence or all industrial automation. Eka's reported product sits in robotic manipulation: systems that combine visual sensing, force feedback and physical action to pick, place, insert or otherwise handle objects in industrial workflows. The included spend is the hardware, gripper and sensing package, control/model software, integration, validation, maintenance and support needed to automate a defined manipulation task. Excluded from this boundary are general warehouse software, autonomous driving, pure vision inspection, and robotics projects whose main value is navigation rather than contact-rich handling. SiliconANGLE reports that Eka's model uses force data as well as camera images and that pilots are underway for pick-and-place and wire handling, while its cited expert cautions that generalization beyond pick-and-place remains open. This narrows the immediate wedge but also makes the market testable: a buyer can compare an exact task's completed cycles, interventions, safety stops and maintenance against its current process. Eka's own site frames the ambition as a single model that generalizes across objects, tasks, and environments, plus robots that are fast and reliable, but that positioning is not evidence of market share or deployment scale.[CM001, CM002, CM003, CM004]

2.2 Sizing lenses and evidence limits

Public evidence supports multiple lenses but not a defensible company-specific TAM, SAM or SOM. The first lens is the installed industrial-robot base: the WorkNation source cites International Federation of Robotics data of 603,000 industrial robots installed in 2025, up 11%, with global operational stock above five million. That is an industry-scale denominator, not Eka's addressable spend, because it includes robots that do not require generalized manipulation and does not disclose replacement, software or integration budgets. The second lens is physical-AI financing: WorkNation cites Crunchbase's $47.4 billion across 521 physical-AI venture deals in the first half of 2026, while warning that the category includes autonomous vehicles, aerospace and drones. It is a capital-market signal, not customer spend. The third lens is task-level throughput: SiliconANGLE reports a company comparison of roughly 1,500 picks per hour versus a human baseline of about 600, with more than 95% pick success and objects up to one kilogram. Those figures are company-reported and lack public independent test conditions. A valuation and funding round can indicate investor expectations, but cannot be converted into revenue, ARR or serviceable market share. Until Eka discloses customer count, contract value, deployment duration and task-level economics, SAM/SOM remain unresolved.[CM005, CM006, CM007, CM008, CM009]

2.3 Buyer segments, drivers, and adoption constraints

The likely buyer is an industrial operator with a repetitive manipulation workflow where object variation or contact uncertainty limits conventional automation. Operations and manufacturing own the throughput and quality outcome; automation or engineering teams own integration; safety and compliance own release risk; finance owns the labor, uptime and total-cost case. The adoption trigger is a bounded task with measurable baseline performance, not a generic promise of general-purpose autonomy. The strongest drivers are labor substitution or augmentation, higher throughput, and flexibility when the object mix changes. The constraints are equally material: the public record does not name industrial pilot customers or provide sustained line results; an expert cited by SiliconANGLE questions whether the system can run on an assembly line for a year without failure; and the gripper is reported at five million duty cycles today against Eka's stated goal of twenty million, leaving lifecycle evidence incomplete. A current Eka hardware test-engineer posting also describes functional, environmental, stress and reliability testing plus certification work, including ISO 10218 and ingress protection, which indicates that validation and compliance are part of the deployment burden rather than solved assumptions. Buyers should therefore require a pilot scorecard covering completed-task rate, cycle time, exception recovery, damage, safety stops, uptime, integration cost and maintenance.[CM010, CM011, CM012, CM013, CM014]

2.4 Exhibits

3.1 Competitive landscape and substitutes

The competitive set is broader than manipulation-model startups. Eka combines visual understanding, force sensing, and action control for variable physical tasks, placing it against AI-native robotics companies such as Physical Intelligence, Generalist, Skild AI, THEKER, and Dream Labs. Those peers are visible primarily through financing signals in the assigned evidence, not matched product or deployment data. Buyers can also choose an incumbent robot-and-integrator cell, commission specialist vision or gripper automation, build internally, or preserve a human-operated workflow. These alternatives compete on the whole job rather than on model architecture alone: implementation risk, service access, safety validation, exception handling, and the cost of changing an operating process. The named startup financing figures indicate a capital-intensive race, but they do not measure revenue, installed base, market share, or production reliability.[CP001, CP003, CP004, CP005, CP006, CP007, CP008]

3.2 Capability, trust, and production evidence

Eka's reported differentiation is the coupling of vision, force feedback, learned control, and custom robotic hardware. Humanoids Daily says the model is trained with reinforcement learning in simulation, while a syndicated report describes force-sensitive grippers paired with a vision-force-action model. The company's reposted hardware-test role also covers validation from PCBAs and actuators through grippers and complete robotic arms, including safety and ingress-protection certification. These facts make the technical thesis more specific than a generic robotics-AI claim. They do not demonstrate lower total cost, smoother integration, or sustained reliability. The public video is expressly not a controlled competitor comparison, and the reviewed evidence provides no common test across Eka, an incumbent automation cell, a specialist system, internal build, and the current human process.[CP001, CP002, CP010, CP011, CP012, CP015]

3.3 Switching costs, distribution, and moat durability

A moat could emerge from force-rich task data, simulation-to-real learning, integrated hardware, and validation knowledge accumulated in deployments. The same integration can become a buyer liability if interfaces, maintenance, model updates, or support are proprietary. Incumbents and integrators may counter with installed equipment, service relationships, and proven safety processes, while buyers may multi-home across arms, grippers, controls, and software when interfaces permit. The assigned evidence does not disclose Eka customer contracts, revenue, or a detailed deployment timeline, and it does not disclose realized pricing or matched total cost. Servicta therefore recommends measuring interventions, recovery time, damage, safety stops, uptime, and integration cost in a pilot. Without those data, technical differentiation is credible but commercial lock-in and pricing power remain hypotheses.[CP009, CP011, CP013, CP014, CP016]

3.4 Competitive diligence verdict

Eka merits a buyer-controlled pilot, not a conclusion that it has won the manipulation market. The evidence is strongest for a force-aware architecture and a company-reported throughput claim of about 1,500 operations per hour versus a human baseline of about 600. It is weakest for named production deployments, shift-length reliability, realized pricing, customer outcomes, and comparable economics. An external expert quoted by SiliconANGLE questioned whether the system could run on an assembly line for a year without failure and whether it generalizes beyond pick-and-place. A decisive test should use the same objects, cycle definition, exception rules, safety boundary, and duration for Eka and at least one credible alternative. Until that evidence exists, the durable advantage is unproven and capital-rich peers, incumbent distribution, internal build, and the status quo remain viable displacement paths.[CP008, CP009, CP012, CP013, CP014, CP015, CP016]

3.5 Exhibits

4.1 Financing provides capacity, not operating proof

The assigned sources consistently describe a $150 million Series A led by Andreessen Horowitz and F-Prime at a reported $1.45 billion post-money valuation. WorkNation identifies Obvious Ventures and E14 Fund as continuing investors, while KuCoin attributes the announcement to third-party reporting and includes an explicit disclaimer that the information may be obtained from third parties and carries no representation or warranty. These are financing-event facts, not evidence of cash on hand, revenue, ARR, gross margin, or cash burn. The available reporting says the funding is intended to expand the Vision-Force-Action model, broaden tasks, and accelerate customer deployments, but no source in this pool provides a budget, runway, debt obligation, or financing covenant. The valuation is a negotiated financing signal and should not be used as a substitute for operating performance or an intrinsic-value conclusion.[CI001, CI003, CI004, CI005]

4.2 Monetization and unit economics remain unverified

The public material describes a robotics system aimed at industrial and manufacturing workflows, including pick-and-place, bin-picking, wire manipulation, and other contact-rich tasks. It does not publish a list price, lease or usage model, software fee, realized revenue, customer contract, deployment count, CAC, sales cycle, gross margin, service cost, or payback period. Renascence explicitly says that funding terms, customers, and independent performance tests are not covered in its available material, while Servicta distinguishes a financing valuation from deployed robots, customer savings, production uptime, and commercial revenue. A financial model should therefore separate a possible equipment or deployment transaction from recurring software economics until contracts and invoicing evidence are obtained. The company-reported throughput and gripper-cost figures are useful diligence inputs, but they do not establish contribution margin because labor, integration, maintenance, downtime, consumables, safety systems, and support costs are not disclosed.[CI006, CI007, CI008, CI009, CI010]

4.3 Capital intensity and the next financing trigger

Eka's disclosed plan to expand model capability and accelerate deployments implies continued research, hardware, deployment, and support spending, but the pool does not quantify any of those categories. The a16z announcement describes simulation-based reinforcement learning and integrated model, sensing, controller, and hardware development; WorkNation says the round will support research and development, capability expansion, and faster customer deployment. Those activities can create a capital-intensive path even if the model ultimately improves automation economics. SiliconAngle and Servicta both emphasize that laboratory or demonstration throughput still needs validation in sustained production, with integration, downtime, maintenance, safety, and repeatability affecting customer economics. The financial verdict is consequently evidence-constrained: financing capacity appears stronger after the round, while revenue quality, margin path, runway, and the timing of a next financing remain unknown. A buyer or investor should require production-pilot economics and a cash plan before underwriting scale.[CI011, CI012, CI013, CI014]

4.4 Exhibits

5.1 Product and customer workflow

Eka's product is described as a robotic manipulation system rather than a standalone software model. The reported workflow runs from sensing and perception through force-aware action, physical manipulation, deployment, and operational support. Servicta describes the core model as using visual and force information together, while the company-described recruiting material places the system at customer sites for configuration, calibration, demonstrations, and issue triage. This implies a buyer workflow in which Eka must integrate with existing arms, grippers, cameras, edge compute, safety controls, and site procedures. The public record does not identify production customers, contract terms, or measured deployment outcomes, so the product should be evaluated as an integrated workflow with material implementation work rather than as a plug-in model.[CE001, CE002, CE003, CE004]

5.2 Technical architecture and maturity

The available evidence points to a layered architecture: visual and force sensing feed a learned policy; simulation and reinforcement learning support training; real robots provide transfer and feedback; and production software provides deployment, telemetry, diagnostics, and safety functions. Recruiting requirements mention ROS2 or similar middleware, Linux, edge deployment, TensorRT or ONNX, production C++ and Python, automated testing, simulation, and system-identification experiments. Hardware validation roles cover PCBAs, actuators, grippers, robotic arms, environmental and stress testing, failure analysis, and certification documentation. These are concrete engineering activities, but job requirements establish intended capabilities and process needs, not proof that every component is mature or deployed at scale.[CE005, CE006, CE007, CE008, CE009]

5.3 Reliability, differentiation, and technical diligence

Eka's differentiation claim is the combination of a vision-force-action model, a force-sensing gripper, simulation-based reinforcement learning, and an operating loop that feeds telemetry and field failures back to engineering. Independent coverage reports company claims of approximately 1,500 object-picking operations per hour and more than 95% success, but does not provide a matched benchmark, customer identity, or enough test detail to establish production performance. The same coverage says a specialist believes the system still needs to prove itself with integrators and sustained industrial use. The decisive diligence questions are therefore task-specific transfer, intervention and recovery rates, uptime, safety stops, maintenance, data handling, update controls, certification scope, and performance after object or layout changes. Public evidence does not establish those measures.[CE010, CE011, CE012, CE013, CE014]

5.4 Exhibits

6.1 Buyer and use-case segments

The available evidence places Eka in a B2B industrial-automation motion rather than a self-serve software market. SiliconANGLE reports that the company is running undisclosed industrial-customer pilots on pick-and-place and wire-handling work, while its reported product positioning centers on a force-sensing gripper and a vision-force-action model. That combination suggests a buyer set involving manufacturing or logistics operators, automation integrators, and operational teams responsible for repetitive manipulation. It does not establish which party signs the contract, owns the robot, or pays for integration. The practical initial segment is likely a bounded workflow where object variation creates labor or automation friction and failures can be safely recovered. This is a diligence framing, not evidence of an installed customer base. The assigned public material does not name a customer, facility, geography of deployment, contract value, or production status, so logos or investor interest should not be treated as proof of customer adoption.[CU001, CU002, CU003]

6.2 Adoption trajectory and durability

Adoption evidence is at the pilot-and-demonstration stage in the retained material. SiliconANGLE says pilots are underway but does not name the industrial customers or publish their operating results; the same report attributes a 1,500-operations-per-hour figure and a greater-than-95% pick-success figure to Eka. Those metrics describe reported system performance, not active accounts, paid deployments, renewal, or customer return on investment. A separate Servicta analysis makes the same commercial distinction, stating that pilot conditions, uptime, task breadth, and long-run reliability still need independent validation. Emergent makes a materially stronger claim that systems are commercially available and shipping to early adopters, but it supplies no customer identity or operating result and is not reconciled with the pilot-stage accounts. A company-issued test-engineer posting also shows Eka building functional, environmental, stress, reliability, and certification processes. The trajectory is therefore best described as commercialization in progress, not verified expansion. The most decision-useful next evidence is a customer-attributed production deployment with a defined baseline, operating duration, intervention rate, uptime, quality result, and denominator for any success percentage.[CU004, CU005, CU006, CU013, CU014, CU015]

6.3 Named proof, expansion, and concentration risk

No retained source names a paying customer or provides a customer-controlled outcome. WorkNation and KuCoin repeat the financing and technical thesis, while BuiltIn's page explicitly warns that its material was generated from popular LLM responses and was not reviewed or approved by Eka; none supplies customer references or retention evidence. That absence is not proof that customers do not exist, but it prevents verification of production deployment, repeat purchase, concentration, channel dependence, or land-and-expand behavior. The reported use cases—pick-and-place and wire handling—are narrower than a general-purpose automation claim and should be evaluated separately. Emergent's assertion that systems are shipping to early adopters raises, rather than resolves, the need to distinguish a shipment from a paid production rollout. Expansion should be treated as an open commercial hypothesis until Eka discloses how a pilot becomes a paid rollout, who performs integration, what support obligations remain with the vendor, and whether performance survives new objects, layouts, shifts, and sites. A customer reference list, redacted contracts, or procurement records would materially improve confidence.[CU007, CU008, CU009, CU013, CU014]

6.4 Exhibits

7.1 Risk ranking and thesis-break triggers

The highest-consequence risk is commercialization transfer. Public accounts describe pilots, demonstrations, and reported laboratory throughput, but do not identify customers, disclose pilot revenue, or provide sustained production-line results. That leaves a material gap between technical promise and deployable operating performance. A thesis break would be repeated failure to meet a buyer's agreed cycle-time, completed-task, quality, safety, and uptime thresholds on representative production work, or an integration burden that erases the expected labor and throughput benefit. The next risk is model generalization: force feedback and simulation may improve recovery from contact variation, but simulation-to-reality transfer remains a recognized challenge and the public sources do not establish performance across shifts, layouts, materials, or unfamiliar task families. Financial risk follows from opacity: the reported $150 million Series A and $1.45 billion post-money valuation provide capital and market validation, not evidence of revenue, cash runway, margins, or customer concentration. Investors should require a task-specific pilot protocol, failure and intervention logs, lifecycle-cost assumptions, and a financing plan before treating the round as proof of reduced execution risk.[CR001, CR002, CR003, CR005, CR005]

7.2 Regulatory, legal, and safety exposure

No public source in the assigned pool establishes that Eka holds a particular industrial safety certification, has faced enforcement, or has been involved in litigation. That is an evidence gap, not evidence that no incident or legal exposure exists. The relevant exposure is operationally coupled to deployment: a flexible gripper applying force around people, equipment, or delicate parts must fit the customer's guarding, emergency-stop, validation, data, and change-control procedures. The reported custom gripper, force sensing, and manipulation claims should therefore be assessed through documentation and site testing rather than inferred from marketing or investor language. Before deployment, diligence should obtain the applicable machine-safety conformity basis, hazard analysis, validation records, software-update controls, incident ownership, insurance allocation, and the customer's responsibility for integration. A kill criterion is any unresolved safety hazard, missing validation artifact, or change to model behavior that cannot be tested and rolled back before production use.[CR007, CR007, CR008, CR008]

7.3 Operational, partner, and dependency risks

Eka's product depends on a stack rather than a model alone: sensing, a force-aware gripper, robot hardware, simulation and reinforcement-learning infrastructure, controls, site integration, maintenance, and human exception handling. The investor description explicitly frames sensing, model, controller, and hardware as a coupled system, which creates differentiation but also expands the failure surface. Public reporting places the work in industrial pick-and-place and wire-manipulation pilots while withholding customer names and operating results. That makes integrator compatibility, spare parts, calibration, support response, and the boundary between Eka and the customer especially important unknowns. Supplier or partner concentration is also not disclosed; a single gripper, robot-arm, compute, or integration dependency could delay deployment or increase cost. Diligence should map every critical counterparty and substitute, test degraded operation, and require service-level commitments. A thesis break is a dependency that cannot be replaced within the buyer's required recovery window or a pilot that needs persistent manual intervention to maintain nominal output.[CR010, CR011, CR010, CR013]

7.4 Financial and execution model risks

The financing headline should not be confused with cash available for operations or with commercial traction. Public coverage reports a $150 million Series A and a reported $1.45 billion post-money valuation, but it does not disclose revenue, gross margin, burn, runway, contract value, or production deployment economics. A robotics company can consume capital through hardware iteration, testing, integration, support, inventory, and field reliability work even when its model is technically promising. The reported 1,500 picks per hour and comparison with a roughly 600-per-hour human baseline are not a complete unit-economics test: sources do not provide matched task definitions, intervention rates, downtime, damage, maintenance, or total cost of ownership. The immediate diligence ask is a bridge from pilot metrics to paid deployment economics, including hardware and software pricing, gross margin by deployment, installation labor, warranty exposure, support costs, working capital, and customer payback. A thesis break is a sustained inability to reach a buyer-approved payback period without undisclosed subsidies or continuing engineering services.[CR014, CR015, CR016, CR016]

7.5 Monitoring plan and unresolved evidence

The public record supports a monitored, evidence-first posture rather than a conclusion that Eka has failed. Investors should track named production customers, months of continuous operation, completed-task rate, intervention and recovery time, unplanned downtime, safety stops, defect or damage rate, changeover performance, gross margin, paid conversion, and concentration by customer and integrator. Each metric needs a denominator, evaluation period, task definition, and attribution. Public sources repeatedly distinguish reported demonstrations from the still-unproven test of sustained production-shift performance. Until those records are available, residual exposure is high for commercialization, integration, safety validation, and financial model risk. The decision can improve if an independent or customer-authored evaluation shows repeatable results on representative tasks with transparent exceptions and a credible path to positive deployment economics; it should worsen if customer results remain undisclosed while valuation expectations rise.[CR018, CR019, CR020, CR021]

7.6 Exhibits

8.1 Investment thesis and anti-thesis

Eka's positive case is a differentiated robotics thesis backed by a reported financing, a force-sensitive gripper, and a Vision-Force-Action model trained through simulation and reinforcement learning. Multiple reports describe a claimed 1,500 operations per hour versus roughly 600 for a human and applications including pick-and-place, wire handling, light-bulb insertion, and other contact-rich tasks. The anti-thesis is equally material: the performance numbers are company-reported, no independent test is cited in the assigned evidence, pilots are described without named customers, and revenue and contracts are not disclosed. The financing therefore validates investor interest more than business-model maturity. A buy call would require evidence that the laboratory or pilot claims translate into paid, durable production deployments.[CV001, CV002, CV003, CV004, CV005, CV016]

8.2 Financing context and entry discipline

The reported financing context is unusually clear on headline terms but weak on the operating denominator needed for valuation. Pomegra, DutchStartup, TechFundingNews, WOWTALE, Robothot, Jawlah, and Multiples each report a $150M Series A around October 20, 2026, at approximately $1.4B to $1.45B post-money, with a16z and F-Prime leading. These are largely syndicated or overlapping accounts, not independent marks. Multiples presents the company as privately held with licensing and product sales, while Pomegra explicitly says revenue and customer contracts were not disclosed. Tracxn's October 15 record still labels Eka unfunded and reports no funding, a direct data conflict that should be resolved against transaction documents rather than averaged away. Entry discipline should therefore be milestone- and price-sensitive, with no supported revenue multiple or cash-runway calculation.[CV006, CV007, CV008, CV009, CV010, CV016]

8.3 Scenarios, comparables, and downside

The bull case is that force sensing and simulation-trained control produce repeatable deployment economics, pilots convert to named paid customers, and the reported performance advantage supports a higher value as the company expands tasks. The base case is that the round funds technical and commercial validation while public evidence remains insufficient to price operating performance. The bear case is that sim-to-real transfer, uptime, durability, safety, or customer integration costs prevent conversion, leaving a premium private mark unsupported. Multiples lists several comparables and reports very different company-level valuations, including Physical Intelligence at $11B and Skild at $14B in the surrounding coverage, but those figures are reported private marks rather than a comparable revenue or EBITDA set. No probability-weighted range is supportable without revenue, margin, retention, deployment, dilution, liquidation-preference, and cap-table inputs.[CV011, CV012, CV013, CV014, CV016]

8.4 Recommendation and decision gates

The recommendation is research-more rather than buy, track, or avoid. The evidence supports a credible technical and financing narrative, but not a sufficiently observable commercial cash-flow base or a reliable denominator for entry valuation. The next decision should be tied to named production customers, paid contract value, deployment uptime, gross margin or contribution economics, repeatability across tasks, and financing terms. A thesis break would be a failed or delayed production deployment, materially lower throughput or success under independent testing, evidence that force-sensing hardware cannot meet durability or safety requirements, or a financing structure that creates substantial preference or dilution overhang. If those gates clear, a new price-sensitive assessment can use a business-model- matched method; until then, the current reported valuation should be treated as an investor mark with substantial execution uncertainty.[CV002, CV005, CV008, CV010, CV014]

8.5 Exhibits