Mercor

1.1 Identity, Operating Model, and Leadership

Mercor identifies its legal entity as Mercor.io Corporation and lists San Francisco as its location [CO001]. The company describes a two-sided marketplace in which AI labs and enterprises buy access to global expert contractors, complemented by human-data, evaluation, benchmark, and enterprise-AI offerings [CO002]. Independent profiles date the company's founding to January 2023 and name Brendan Foody, Adarsh Hiremath, and Surya Midha as its three co-founders [CO003]. They also report that former Uber Chief Product Officer Sundeep Jain joined as Mercor's first president in May 2025 [CO004]. The evidence therefore supports a founder-led company with one named experienced operating executive, but it does not establish current functional remits for each founder, a complete leadership roster, board composition, voting control, or succession arrangements. Those omissions matter because rapid scale and concentrated founder knowledge can create execution and governance dependencies that cannot be resolved from public profiles alone.

1.2 Capital Formation and Corporate Milestones

Public profiles describe a steep financing progression: a $30 million Series A at a $250 million post-money valuation in September 2024 [CO005], a $100 million Series B at $2 billion in February 2025 [CO006], and a $350 million Series C at $10 billion in October 2025 [CO007]. A later independent review reported that talks for a $500 million raise at a $20 billion valuation had not produced a confirmed closing by October 1, 2026 [CO008]. Consequently, $10 billion is the latest priced valuation supported by the assigned evidence; the $20 billion figure is a reported negotiation, not an updated valuation fact or cash available to the company. The financing history demonstrates investor willingness to fund Mercor's growth, but the evidence does not disclose secondary-sale mechanics, debt or credit facilities, liquidation preferences, ownership percentages, voting rights, or the current board. The stakeholder map therefore records economic relevance without implying control terms that have not been published.

1.3 Scale Signals, Revenue Quality, and Control Gaps

Mercor said in May 2026 that it had crossed $1 billion in annualized revenue run rate, while Sacra separately estimated $2 billion of annualized gross revenue in June 2026 [CO009] [CO010]. These are not ARR figures and should not be read as net revenue retained by Mercor. Company pages also show a changing payout scale: more than $2 million per day to over 30,000 weekly active contractors in May, versus a current homepage claim of more than $4 million in daily payouts and 30,000-plus experts [CO011]. The homepage separately claims 507,000 roles created [CO012], and a company job page advertised a remote UK commercial-operations expert contract at $150 per hour [CO013]. Mercor's privacy notice expressly excludes the work.mercor.com service from its website-notice scope [CO014]. Its own payments retrospective acknowledges that manual reconciliation and off-platform data had outgrown early controls [CO015]. Third-party estimates that contractors receive 60% to 70% of gross billings imply a materially smaller retained take, but no audited gross-to-net bridge is public [CO016].

1.4 Exhibits

2.1 Market Definition, Scope Boundaries, and Spend Classification

Mercor occupies a newly emergent market layer positioned between traditional professional technical staffing agencies and specialized AI data infrastructure providers. The company's core commercial engine connects elite domain experts—including physicians, corporate attorneys, specialized engineers, and research scientists—with frontier artificial intelligence research laboratories and enterprise model developers requiring human feedback for model training, reinforcement learning, and benchmark evaluation. To accurately evaluate Mercor's commercial opportunity and valuation multiple, total market expenditures must be rigorously disaggregated into distinct budgetary categories rather than conflated into an undifferentiated artificial intelligence market. Included spend spans frontier model reinforcement learning from human feedback (RLHF), expert-authored evaluation benchmarks (such as Mercor's APEX suite), enterprise workflow and operational data licensing, and specialized contractor recruitment. Conversely, commoditized low-skill mechanical crowd data labeling, internal corporate engineering payroll, standard software-as-a-service application subscriptions, and capital expenditures for compute hardware or model inference hosting are explicitly excluded from Mercor's addressable envelope. The status-quo substitutes for Mercor's solution encompass legacy technology staffing firms such as Toptal and Andela, general freelance labor platforms like Upwork, manual internal recruitment pipelines at frontier laboratories, and incumbent synthetic data labeling competitors.[CM001, CM002, CM003, CM004]

2.2 Market Sizing Lenses, Growth Projections, and Economic Realities

Evaluating Mercor's market size requires multiple complementary analytical lenses rather than a single speculative top-down figure. While global professional staffing and human capital management represents an expansive $650 billion addressable spend, Mercor's realistic serviceable market is constrained to specialized AI training labor, reinforcement learning evaluations, and enterprise domain expertise pipelines, estimated between $12 billion and $18 billion. Mercor's serviceable obtainable market, representing expert RLHF and enterprise workflow data monetization, is currently sized between $3.0 billion and $4.5 billion. Mercor reported surpassing $2.0 billion in gross annualized revenue by June 2026, up from $1.0 billion four months earlier, supported by paying out over $1.5 million to $4.0 million daily across an active network of over 30,000 weekly contractors and 300,000 registered professionals. However, this gross annualized volume reflects gross contractor billing rather than high-margin software annual recurring revenue. Market analysts observe that Mercor operates an estimated 30 percent recruiting fee and take-rate structure over pass-through contractor payouts, meaning net revenue represents a fraction of headline gross volumes. Furthermore, peer benchmarks highlight that private markets value data infrastructure peers such as Surge AI ($15 billion valuation on over $1 billion trailing revenue) and Scale AI ($29 billion valuation) at infrastructure multiples, yet Mercor's expansion relies on sustaining gross margins amidst intensifying price discovery.[CM005, CM006, CM007, CM008, CM009]

2.3 Buyer Segmentation, Budget Allocation, and Adoption Pathways

Purchasing authority for Mercor's offerings is distributed across distinct organizational buyers with diverging procurement criteria and budget constraints. Frontier AI model research laboratories—exemplified by OpenAI, Anthropic, and Google DeepMind—represent the primary economic engine of Mercor's rapid scaling. In these accounts, procurement is driven by Chief Scientists, Technical Staff, and Research Directors utilizing core research and development budgets to acquire domain-validated human feedback and proprietary evaluation datasets. The primary adoption trigger in this segment is model capability plateaus, where standard synthetic or public web corpora fail to advance complex reasoning, demanding certified legal, medical, and mathematical judgment. Conversely, enterprise commercial adopters—spanning financial services, healthcare institutions, and logistics corporations—engage Mercor through Chief Information Officers, Chief Technology Officers, and Heads of Enterprise Automation. These enterprise buyers utilize IT innovation, business process transformation, and departmental operational budgets to deploy domain-specific AI agents and monetize internal operational workflows. Unlike frontier labs, enterprise procurement operates with extended 60-to-180-day sales cycles, requiring rigorous data governance, verified SOC 2 and ISO compliance certifications, and demonstrated cost-displacement return on investment before transitioning pilots into production rollouts.[CM002, CM005, CM010, CM011]

2.4 Adoption Drivers, Structural Bottlenecks, and Concentration Headwinds

Mercor's business expansion benefits from powerful secular tailwinds, led by the frontier AI industry's acute requirement for high-tier expert human supervision as public web text for pre-training is exhausted. Neutrality concerns among frontier laboratories following Meta's $14 billion investment in Scale AI created an immediate commercial opening that Mercor aggressively captured. Furthermore, enterprise demand for verified AI agents across accounting, legal, and healthcare workflows has expanded adoption into non-engineering corporate verticals. Nevertheless, severe structural constraints and operational vulnerabilities threaten sustained growth. The March 2026 supply-chain cybersecurity incident—originating through a compromised open-source LiteLLM dependency—resulted in the theft of 4 terabytes of Mercor platform data, contractor personal identifying information, and proprietary lab training methodologies. The breach triggered multiple federal class-action lawsuits and led Meta to indefinitely pause all active work with Mercor, underscoring extreme counterparty concentration risk. In addition, the risk of long-term synthetic data substitution, potential commoditization by incumbent platforms like LinkedIn and Upwork, and the lack of transparent, published buyer-side pricing schedules present material adoption frictions that require ongoing diligence.[CM004, CM012, CM013, CM014, CM015]

2.5 Exhibits

3.1 Post-Training Landscape and Market Structure

The market for AI training data and model evaluation underwent a structural sorting following Meta's June 2025 transaction acquiring a 49% stake in Scale AI for $14.3B [CP001]. That transaction raised immediate neutrality concerns across frontier AI labs such as Google and OpenAI, prompting a rapid reallocation of post-training and reinforcement-learning-from-human-feedback (RLHF) budgets toward independent providers [CP001, CP005]. As frontier models shifted from low-skill bulk data collection—such as image classification and bounding boxes—toward complex reasoning, post-training bottlenecks increasingly require credentialed professionals (including physicians, lawyers, finance professionals, and specialized software engineers) [CP004, CP005]. Within this restructured ecosystem, Mercor competes directly against established data platforms like Scale AI and managed-quality providers like Surge AI, while also facing pressure from specialized staffing challenger networks, academic talent marketplaces such as Handshake AI, and lab internal build efforts [CP001, CP006, CP007]. Mercor differentiates by operating an automated AI-interview and matching marketplace that brokers expert labor hours at speed, in contrast to Surge AI's boutique dataset delivery or Scale AI's bundled enterprise data platform [CP001, CP002, CP008].

3.2 Competitor Profiles and Strategic Positioning

Scale AI, historically the dominant training data provider across autonomous vehicles and early LLM development, has pivoted toward defense, regulated enterprise workloads, and robotics following the departure of major frontier lab post-training volumes [CP001]. Surge AI represents the primary quality benchmark among frontier research labs, operating an elite, highly selective contractor network of approximately 50,000 specialists focused on task-level quality control, rubric authoring, and complex reasoning benchmarks such as the Tuesday Work Index [CP001]. Handshake AI and micro1 represent low-overhead or campus-oriented challenger marketplaces capable of sourcing quantitative and engineering talent at competitive rates [CP001, CP005]. Meanwhile, traditional recruitment networks such as Refery highlight the boundary between employee search and Mercor's primary delivery of expert staffing, AI evaluations, and enterprise AI delivery [CP007]. In response to shifting buyer requirements, Mercor has heavily expanded engineering headcount in San Francisco, with 63 of 107 open roles in engineering, reflecting an aggressive push into proprietary software evaluations and agentic reinforcement-learning environments [CP003, CP006].

3.3 Commercial Models, Pricing Dynamics, and Switching Costs

Commercial terms across the human-data landscape reflect fundamentally divergent units of value: Mercor bills expert hours with a platform take rate of approximately 30% to 35%, Surge AI bills completed deliverable datasets on a per-task or per-project basis, and Scale AI bundles enterprise platform licensing fees with variable labor markups [CP001, CP002, CP005]. Mercor's advertised expert rates have exhibited market softening, moving from $141 per hour in June 2026 to $109 per hour in August 2026, while platform-level daily payouts exceed $4M across a network of over 500k created roles [CP001, CP009, CP010]. While Mercor's model offers unmatched speed and flexibility when spinning up domain-specific expert panels, switching costs for frontier labs remain low [CP001, CP005]. AI research labs routinely multi-home across multiple vendors to benchmark quality, prevent single-vendor dependency, and arbitrage labor rates across platforms [CP001, CP005]. Furthermore, labs retain direct relationships with specialized expert pools or build internal evaluation teams when project confidentiality is paramount [CP001, CP005].[CP001, CP002, CP005, CP009, CP010]

3.4 Moat Durability and Adverse Risks

Mercor's long-term defensibility faces twin structural vulnerabilities: trust and platform security on one side, and customer disintermediation or commoditization on the other [CP001, CP005]. On March 27, 2026, a malicious compromise of the LiteLLM open-source release pipeline resulted in the exfiltration of approximately 4 terabytes of Mercor data, including contractor candidate profiles, personal data, source code, and API keys [CP001]. The breach triggered an immediate pause in work by Meta, federal litigation from contractors, and renewed enterprise scrutiny over data segregation and third-party software supply chain governance [CP001]. Although Mercor deployed comprehensive remediation—engaging Mandiant and Latacora to tighten network controls and rotate credentials—the incident highlighted that marketplace platforms bear systemic enterprise data liabilities [CP001]. Simultaneously, the rise of software-based reinforcement learning environments and synthetic data generation poses an ongoing threat of commoditizing human evaluation labor, forcing Mercor to invest in agent testing benchmarks and software infrastructure to preserve its valuation multiple [CP001, CP005].

3.5 Exhibits

4.1 Revenue model and revenue quality

Mercor presents three commercial surfaces: human data and evaluations for AI developers, data monetization for enterprises, and enterprise AI deployment supported by forward-deployed engineers. Public evidence is strongest for the first surface. AI Weekly, summarizing internal documents reviewed by The Information, reported about $614 million of gross revenue in the first half of 2026 and a June annualized gross-revenue run rate above $2 billion. Those labels are not ARR and should not be valued as recurring software revenue. The same coverage reported that OpenAI and Anthropic represented roughly 91% of the first-half total, making revenue quality dependent on a concentrated set of frontier-lab budgets. Mercor discloses no product-level revenue mix, contract duration, renewal rate, or recognition policy.[CI001, CI002, CI003, CI008]

4.2 Pricing, contractor economics, and GTM

Public pricing is visible on the supply side rather than the customer side. Mercor's website reports an average contracted expert rate of $108 per hour and more than $4 million in daily payouts, while the live opportunity marketplace displays hourly and task-based rates ranging from $20 per task to as much as $150 per hour in the sampled listings. These are contractor compensation points, not customer list prices or realized net revenue. The available evidence therefore supports a project-led, usage-linked service model, but it does not establish sales cycle, CAC, payback, discounting, channel economics, or enterprise contract minimums. Buyer concentration further weakens the usefulness of gross growth as a standalone sales-efficiency proxy.[CI003, CI006, CI007]

4.3 Gross-to-net bridge and cost structure

The best public gross-to-net clue is AI Weekly's report that contractors keep roughly 60% to 70% of quoted gross revenue. Arithmetically, that leaves approximately 30% to 40% before Mercor's own payroll, recruiting incentives, payment operations, fraud controls, insurance, legal costs, product development, and other overhead; it is not a disclosed gross margin. Mercor's privacy notice also describes data-security, compliance, marketing, and service- provider activities, but provides no expense amounts. Sector reporting that 45% of hiring assessments were flagged for integrity issues is not Mercor-specific, yet it illustrates why screening and fraud-control costs cannot be assumed immaterial. Audited gross margin, contribution margin, operating expenses, working capital, capex, and cash conversion remain unavailable.[CI004, CI005, CI013, CI016]

4.4 Capital adequacy and underwriting verdict

Funding trackers do not agree on cumulative capital: RevenueMemo reports $483.6 million across four announced equity rounds, while FundedIQ records $455 million across three rounds. The difference appears to reflect round coverage rather than cash available today, and neither figure establishes unrestricted cash. RevenueMemo reports no disclosed debt, cites a third-party report of $6 million in first-half 2025 profit, and says a roughly $500 million financing at a $20 billion valuation was discussed in 2026 but had not been announced. None of those items supplies a current balance sheet, cash-flow statement, burn rate, runway, or financing trigger. The financial verdict is therefore high growth with plausible positive unit spread, offset by weak metric comparability, concentrated demand, and insufficient evidence to underwrite margin durability or capital adequacy.[CI009, CI010, CI011, CI012, CI014, CI015]

4.5 Exhibits

5.1 Product surfaces and user workflows

Mercor's public product has moved beyond a single recruiting workflow. Its current website separates expert work, human data and evaluations, benchmark construction, and enterprise-agent deployment, while also marketing data monetization from customer workflows. In customer-workflow terms, the common operating model is to recruit or identify domain experts, capture professional work or judgments, convert that input into training data, evaluations, or benchmark environments, and deliver the resulting assets to model developers or enterprise teams. A company-issued job listing also shows that the supply side includes part-time remote expert engagements rather than only permanent placements. The evidence establishes the breadth of the offer, but not the revenue mix, adoption, or production maturity of each surface.[CE001, CE002, CE004, CE015, CE016]

5.2 Architecture and operating model

The most concrete architecture evidence comes from MongoDB's supplier case study. It describes a unified candidate collection on MongoDB Atlas, semantic retrieval through Vector Search and Voyage AI embeddings, Lucene-backed full-text retrieval, hard filters, and Mercor ranking logic. It also describes Kafka-based change-data capture that updates candidate embeddings when profiles or interviews change. Reported results include a reduction in profile latency from roughly thirteen seconds to milliseconds, a move from six-hour search-index delays to near-immediate availability, and database growth from 100 GB to more than 2 TB by May 2026. These are supplier-reported operating outcomes, not independently reproduced benchmarks, so they support architecture maturity more strongly than they support service-level reliability.[CE005, CE006, CE007, CE008, CE009]

5.3 Differentiation, maturity, and roadmap

Mercor's differentiation appears to combine a large expert network, automated screening, workflow-specific data capture, hybrid retrieval, and proprietary evaluation products. A current engineering role describes intended code-task retrieval using dense embeddings, BM25, ranking, reranking, model routing, and quality guardrails; because this is a hiring specification, it should be treated as roadmap and capability intent rather than proof that every component is in production. Independent commentary supports the strategic logic of expert-led post-training but also identifies a durability risk: synthetic data, software-built reinforcement-learning environments, and customer cost pressure could reduce the value of human-intermediation workflows. The only public developer artifact in the assigned evidence is an unaffiliated assessment simulator, which is useful as ecosystem signal but does not establish Mercor code quality, release cadence, or API adoption.[CE009, CE011, CE012, CE014]

5.4 Trust, security, privacy, and quality controls

Mercor markets automated anonymization across more than sixty categories of sensitive identifiers and says customers retain control over shared data. Its website privacy notice also discloses extensive website telemetry, including session replay, while explicitly excluding the separate work.mercor.com service from that notice. The assigned sources therefore do not provide a unified description of contractor-data handling, model-training data boundaries, audit logs, retention, incident response, or independent certifications. This gap matters because an independent tertiary source summarizes reports of a March 2026 supply-chain incident and related customer pauses; that source is not sufficient to quantify scope or remediation, but it raises the diligence threshold. Investment-grade validation requires the platform-specific privacy notice, control reports, penetration-test summary, incident postmortem, and benchmark quality-governance documentation.[CE003, CE010, CE013]

5.5 Exhibits

6.1 A two-sided customer and expert network

Mercor's public surfaces distinguish the paying demand side from the expert supply side. The company says it connects expertise to frontier AI labs and brings the same network to Fortune 2000 companies deploying AI, while its current marketplace lists project-based work across generalist, sales, quality, data-science, and specialist roles. The reported five-million-person domain-expert network is broader than the 30,000-plus experts described for human data and evaluations; these are different populations and should not be combined into one active-user count. Enterprise buyers are routed into human-data, evaluation, workflow-diagnostic, deployment, and data-monetization offerings, but public buyer pricing is not disclosed.[CU001, CU003, CU005, CU006, CU008, CU012, CU016]

6.2 Scale indicators are large but use different denominators

The strongest current adoption signals are company-reported operating measures rather than audited customer metrics. Mercor's newsroom states that more than $4 million is paid to its expert network each day and that the broader network contains more than five million domain experts. Its homepage separately shows an average contracted rate of $108 per hour, 507,000 roles created, and 30,000-plus experts for human-data and evaluation work. These measures indicate substantial marketplace throughput, but none provides active buyer count, active expert count, completed-role count, gross billings, take rate, or repeat usage. Serval's supplier-authored case study adds operational evidence: it says Mercor automated onboarding for more than 4,000 contractors, supporting scale but not buyer retention.[CU001, CU002, CU003, CU004, CU005, CU011, CU016]

6.3 Enterprise proof is quantified but anonymized

UsagePricing's record of Mercor's enterprise page describes six case studies, up from three, and reports quantified outputs across diagnostics, value-creation assessments, production-readiness evaluation, and an SRE deployment. The examples show that buyers have used Mercor for real analytical and evaluation work: 25 candidate workflows across seven functions, 132 expert-graded HR-agent runs, and identified value or mitigation improvements. However, the retained text does not name the external clients, disclose contract values, or establish which engagements became recurring production deployments. The evidence therefore supports engagement breadth and concrete project outputs, not a claim of durable software adoption or realized customer economics.[CU006, CU007, CU008, CU015]

6.4 Retention, concentration, and marketplace durability are open

The retained evidence does not establish NRR, GRR, churn, renewal rates, contract length, repeat-project frequency, or top-customer revenue concentration. Built In's explicitly AI-generated and unreviewed summary nevertheless flags reliance on a small set of frontier labs and large technology buyers, so it is a low-confidence risk signal rather than independent verification of concentration. On the supply side, Breaking Even reports a 138-post spike dominated by empty-queue complaints, indicating that recruiting reach does not guarantee steady utilization for experts. Mercor's website privacy notice also excludes the work.mercor.com service and points readers to a separate Expert Privacy Notice, leaving service-specific controls outside the reviewed source set. Customer-level cohort data and direct buyer references are required to resolve durability.[CU009, CU010, CU013, CU014, CU015]

6.5 Exhibits

7.1 Severity-Ranked Risk Taxonomy and Valuation Implications

Mercor's risk posture underwent a structural deterioration in March 2026 when a software supply-chain backdoor in the open-source LiteLLM dependency enabled unauthorized threat actors to access internal systems. This incident exposed sensitive contractor PII, tax identifiers, and video interview biometrics across its expert marketplace. For a high-velocity talent and human-data marketplace whose private valuation reached $10B on the premise of being the trusted data and human intelligence layer for frontier AI models, security and compliance are existential prerequisites. Customer confidence, enterprise procurement standards, and regulatory exposure under emerging artificial intelligence and data privacy frameworks represent material hurdles. The resulting fallout spans multi-jurisdictional class-action litigation, enterprise procurement freeze risk, and extreme customer concentration among a handful of frontier model labs whose workflow shifts could break the core investment thesis.[CR001, CR002, CR003, CR007, CR008]

7.2 Regulatory, Legal, and Compliance Exposures

The legal fallout from the security incident materialized immediately across federal courts. Between April 1 and April 7, 2026, five distinct class-action lawsuits landed in federal district courts across California and Texas against Mercor.io Corporation, including Deboni v. Mercor.io Corporation, Gill v. Mercor.io Corporation, and Esson v. Mercor.io Corporation in the Northern District of California, alongside White and Beltran v. Mercor.io Corp. in the Northern District of Texas. The complaints allege negligence, inadequate encryption, absence of robust multi-factor authentication, and widespread failure to protect contractor PII, W-9 tax forms, and biometric interview assets. Plaintiffs also allege secondary exposure through contractor monitoring software, while law firms such as Dapeer Law initiated broader formal investigations following Mercor's notification letters and report to the California Attorney General offering 24 months of TransUnion credit monitoring.[CR004, CR005, CR006, CR009]

7.3 Operational Security and Counterparty Dependency Vulnerabilities

Beyond immediate courtroom defense, Mercor's operational architecture exposes systemic vulnerabilities inherent to modern AI supply chains. Attackers compromised PyPI publishing credentials for open-source LiteLLM versions 1.82.7 and 1.82.8, injecting backdoors that harvested cloud credentials and API keys within build pipelines, which threat actors subsequently leveraged against Mercor's data stores. While Mercor announced comprehensive remediation—including auditing dependencies, rotating access credentials, deploying restrictive cloud security policies, engaging Mandiant and Latacora, and implementing 24/7 managed detection—the operational blowback highlighted heavy counterparty dependency. Frontier labs like Meta paused ongoing projects pending security reviews, illustrating how customer concentration among foundational model creators amplifies counterparty risk.[CR010, CR011, CR012, CR013, CR014, CR015]

7.4 Exhibits

8.1 Investment judgment

The evidence supports TRACK / RESEARCH MORE, not a price-insensitive buy. Mercor has an unusually strong scale signal: management said annualized revenue run rate exceeded $2 billion, and reporting described an early financing discussion at twice the last confirmed valuation. Yet the headline ratio is not an investable multiple. Public reporting says gross revenue includes compensation passed through to experts, while public evidence does not provide the gross-to-net bridge, gross margin, customer concentration, preference stack, or a closed 2026 financing. At a reported $20 billion mark, that missing information is central rather than incremental. The decision should remain conditional on verified economic quality and enforceable financing terms.[CV001, CV002, CV003, CV004, CV014, CV016]

8.2 Thesis, anti-thesis, and financing context

The thesis is that Mercor is extending a large expert network into evaluations, benchmarks, and enterprise agent workflows, which could make the platform more valuable than a labor marketplace alone. The anti-thesis is that reported revenue quality may be materially weaker than the headline: contractor payouts are embedded in gross revenue, the public work marketplace displays project-based compensation, and third-party funding totals disagree. Tracxn reports $492 million raised, RevenueMemo reports $483.6 million, and Exa's directory still reports $136 million through only three rounds. These discrepancies do not change the confirmed $350 million Series C, but they show why primary financing documents and a fully diluted cap table are required before underwriting entry price or dilution.[CV002, CV004, CV006, CV007, CV008, CV009, CV010, CV011]

8.3 Scenarios and return logic

The scenario bands below are decision anchors bounded by the last confirmed $10 billion mark and the reported $20 billion discussion; they are not estimates of fair value. A bull case requires the reported financing to close and management's run-rate claim to translate into durable net revenue, margin, retention, and customer diversification. The base case treats the public marks as a reference interval while withholding a target return. The bear case is a repricing below the last round if gross-to-net economics disappoint or legal and customer fallout persists. No public evidence supports a defensible holding period, exit proceeds, or return distribution because fully diluted ownership and preference terms are not available.[CV001, CV002, CV003, CV004, CV012, CV013, CV014, CV015, CV016]

8.4 Exit readiness and final diligence

Exit readiness is not established by the available evidence. RevenueMemo reports that Mercor remains private and has not filed for an IPO, while the reported 2026 round was still unannounced. A security incident followed by at least seven reported class actions and a reported Meta pause raises the required proof threshold for customer retention, cyber controls, insurance, and contingent liabilities. Before a valuation recommendation can move to invest, diligence should obtain audited gross and net revenue, gross margin, cohort retention, top-customer concentration, current cash and debt, the fully diluted cap table, preference and secondary terms, lawsuit status, cyber-remediation evidence, and a board-approved exit or liquidity plan. Failure to reconcile those items is a thesis-break condition at a premium valuation.[CV001, CV012, CV013, CV015, CV016]

8.5 Exhibits