OpenAI

1.1 Identity, mission, and structure

OpenAI describes itself as an AI research and deployment company whose mission is to ensure that artificial general intelligence benefits all of humanity. The present structure is unusual for a venture-backed startup: the nonprofit OpenAI Foundation controls OpenAI Group PBC, and the for-profit arm is intended to advance the mission while still supporting scale and capital formation. That dual structure is the anchor for every later diligence question in this report. That is unusual enough to change how a buyer, partner, or investor should think about control, information rights, and mission drift.[CO001, CO002, CO003, CO004, CO005]

1.2 Leadership, governance, and scale

The board and senior leadership remain highly concentrated. Sam Altman remains the central public executive; the Foundation-appointed board includes a small group of experienced operators, investors, and safety/governance specialists. Publicly visible scale is now extraordinary: OpenAI says it reaches more than 1 billion weekly active users, more than 1 million business customers, and more than 7 million ChatGPT workplace seats. That scale is real enough to support a category-leader thesis, but it also means even small execution mistakes are now enterprise- and consumer-grade events. The disclosed investor base is also broad, spanning Microsoft, SoftBank, Amazon, and NVIDIA, which makes the company both unusually well-capitalized and unusually dependent on a small set of strategic relationships. The disclosed investor base is also broad, spanning Microsoft, SoftBank, Amazon, and NVIDIA, which makes the company both unusually well-capitalized and unusually dependent on a small set of strategic relationships. It also means external diligence needs to focus on concentration risk, because a few people still carry outsized responsibility for strategy, financing, and trust.[CO006, CO008, CO009, CO010, CO013]

1.3 Capital base and milestone chronology

OpenAI’s latest disclosed capital event is the March 2026 $122 billion round at an $852 billion valuation. The company also says it is now generating $2 billion in monthly revenue. The public milestone record is therefore no longer just a founding story; it is a chronology of recurring fundraising, product launches, governance resets, and service incidents. The open question for diligence is not whether the company has scale, but how much of that scale is already durable versus still dependent on a narrow capital and infrastructure stack. The current finance picture therefore combines public valuation, strategic partner exposure, and unresolved economics in the cap stack. That combination is exactly why the investor map matters even though the full ownership details are not public. The current finance picture therefore combines public valuation, strategic partner exposure, and unresolved economics in the cap stack. That combination is exactly why the investor map matters even though the full ownership details are not public. The relevant chronology now includes not only product launches but also recapitalization, financing, and service incidents, all of which affect the public narrative. That is the right lens for understanding dilution, control, and the practical meaning of the valuation.[CO007, CO011, CO012, CO014]

1.4 Exhibits

2.1 Market boundary and substitutes

The relevant market is not a single TAM headline. For OpenAI, the practical boundary is the spend on consumer assistants, enterprise workflow automation, and developer/API infrastructure that can replace search, software tooling, analysis, and some service labor. The company’s own materials show that this is a market for repeat workflows, not one-off queries, and the substitutes include in-house build, other model providers, and existing enterprise software stacks. This boundary also captures adjacent tools, because buyers compare OpenAI against legacy software, in-house builds, and other AI platforms at the same decision table. That is why the category keeps stretching beyond the old chatbot label.[CM001, CM007, CM012, CM013]

2.2 Sizing lens and evidence-constrained adoption

OpenAI’s current monetized base is already large enough to justify a top-down sizing lens from inside the company. The enterprise report and funding announcement show more than 1 million business customers, 7 million workplace seats, 900 million-plus weekly users, and $2 billion in monthly revenue. Rather than a single TAM figure, the more useful diligence view is a multi-lens picture: consumer scale, enterprise seats, and developer/API usage all point to a category that is still expanding its monetizable surface area. The right interpretation is therefore a layered adoption stack, not a single-market number. That is a better fit for a company whose usage can deepen after a sale. It also explains why revenue, not just usage, is the cleanest current proxy for market size.[CM002, CM003, CM004, CM005, CM006, CM008]

2.3 Buyers, drivers, and constraints

The earliest buyers are not random consumers; they are workers and organizations with clear budget owners, workflow pain, and enough volume to justify integration. OpenAI’s own enterprise report emphasizes productivity gains, more frequent use, and higher adoption intensity in technology, healthcare, manufacturing, finance, and professional services. The key constraints are governance, integration, and trust. As those constraints ease, the market expands from experimentation toward embedded workflow infrastructure. The buyers care about time saved, governance, and whether AI can be embedded in repeatable internal workflows rather than treated as a novelty. In other words, the buyer is paying for a workflow outcome, not a model demo.[CM008, CM009, CM010, CM011, CM012]

2.4 Exhibits

3.1 Competitive set and substitutes

OpenAI no longer competes only with other model vendors. The company’s current product stack crosses into repository workflows, coding agents, and internal workflow automation, which puts it up against GitHub-like surfaces, in-house builds, and modular toolchains that can be stitched together by enterprise teams. The main threat is not one rival; it is the possibility that customers multi-home across model vendors while keeping control of their workflow surfaces. The result is a market where customers can multi-home across vendors while keeping the workflow surface in a different place. That makes distribution and retention the more important questions than any single benchmark release.[CP001, CP003, CP007, CP010]

3.2 Bundle power and distribution

Codex changes the story because it is not a generic chatbot. It is a workflow agent that runs tasks in parallel, fits into CLI and ChatGPT flows, and can sit close to code review, refactoring, and pull-request behavior. That makes distribution a first-order moat variable: OpenAI needs to be present in the places where engineers already work, especially GitHub, terminal, and enterprise automation surfaces. In practice, that means a developer can choose the same model but different operating surfaces, which is the key competitive twist. The company’s advantage is therefore partly a placement problem, not only an intelligence problem.[CP002, CP004, CP005, CP006, CP008]

3.3 Moat durability and displacement risk

The moat is therefore moving from model quality alone to workflow integration, governance, and product breadth. The more OpenAI bundles chat, agents, API usage, and repository-aware tooling, the more it can reduce the likelihood that customers swap only one layer at a time. The counter-risk is commoditization: if competitors match the model layer and keep the workflow layer open, OpenAI may face price pressure rather than winner-take-all economics. The moat will hold only if the workflow layer stays sticky and customers do not treat OpenAI as a swappable model API. If usage migrates to adjacent surfaces, OpenAI will still win attention but lose some pricing power.[CP005, CP007, CP009, CP010]

3.4 Exhibits

4.1 Revenue scale and margin profile

The current public picture is striking: OpenAI says it generates $2 billion per month, while Sacra estimates annualized revenue near $40 billion and Q2 2026 revenue at $6.7 billion. The same third-party analysis says enterprise is now more than half of revenue and that gross margin is around 33%, which makes the company look more like a very large, still-scaling infrastructure platform than a mature software business. The mix matters because the company is monetizing both consumer behavior and enterprise workflow depth rather than a single channel. It is a strong scale story, but it still lacks the cost structure of a mature software franchise.[CI002, CI003, CI004, CI008]

4.2 Burn, liquidity, and financing stack

The problem is burn. Public analysis places 2026 cash burn around $27 billion and 2027 burn around $63 billion, which means the business still needs enormous ongoing capital to keep pace with its own demand curve. OpenAI’s $4.7 billion revolver helps, but it does not change the core fact that the company remains financing-intensive. The March 2026 round also broadened the cap table to strategic investors with compute relevance, which reinforces the view that finance and infrastructure are now inseparable. This is why liquidity discipline and strategic investors matter as much as the headline valuation. The financing stack therefore matters as much as the operating plan.[CI005, CI006, CI007, CI010]

4.3 Valuation read-through

At $852 billion, the valuation asks the market to believe that revenue growth, product breadth, and platform lock-in will outrun the economics of compute. The public data supports a premium growth narrative, but not yet a clean margin story. That is why the valuation stance remains stretched even with the latest revenue expansion: the company may be growing fast enough to justify scale, but it still has to prove that scale can translate into durable free cash flow. Public comparables are scarce, so the best read-through comes from burn discipline, margin trajectory, and the credibility of the IPO story. The current multiple still assumes the market will pay for optionality long before the profit line is obvious.[CI001, CI009, CI011, CI010]

4.4 Exhibits

5.1 Current product stack

OpenAI now sells across chat, coding, enterprise workflows, vertical use cases, and security/access controls. That breadth matters because the company is no longer a pure model vendor; it is a product platform that wraps the model layer with workflow, governance, and deployment surfaces. The stack is now broad enough that a buyer can evaluate OpenAI as a platform rather than a single model endpoint. It also now includes controls, vertical packaging, and distribution mechanics that matter to buyers. That platform framing is now central to how customers evaluate the company. Buyers now compare the bundle against broader enterprise platforms.[CE001, CE007, CE008]

5.2 Codex and workflow integration

Codex is the clearest example of the platform shift. OpenAI describes it as a cloud software-engineering agent that can work on many tasks in parallel, and the surrounding ecosystem increasingly treats it as a workflow product rather than a narrow coding assistant. That makes GitHub-like surfaces and repository-aware tooling part of the competitive context. That matters because engineering teams buy tools that fit their repository and review habits, not just tools with strong benchmarks. The external ecosystem is already teaching teams how to use it, which is a strong adoption signal. It also helps explain why workflow proximity can matter more than benchmark bragging rights. That is a meaningful shift in how the product gets evaluated.[CE002, CE003, CE009]

5.3 Safety, access, and deployment controls

The company’s newest frontier capabilities come with new guardrails. Astra is explicitly framed as a critical cybersecurity-capability model, and Daybreak now lets admins scope access by project or user. The product direction is therefore toward more controllable enterprise deployment, even if that means more friction for legitimate tasks. The frontier-product story is increasingly inseparable from the governance layer that controls who can use the most capable models. At the same time, more controls create more support and governance overhead for customers. Governance is becoming a selling point rather than only a compliance burden. It also pushes the company closer to regulated software norms.[CE004, CE005, CE006, CE010, CE011]

5.4 Exhibits

6.1 Customer segments and buyer motion

The current customer motion spans consumers, individual workers, business seats, and large enterprise deployments. Buyers often differ from users: IT or function leaders approve the spend, workers use the product, and admins govern access. The strongest public evidence is not just seat count but repeated use, internal GPTs, and department-level workflow integration. The best way to read this is as a land-and-expand motion that starts with familiarity and ends with departmental deployment. That pattern makes the customer base look broader than a simple consumer app install base. That also creates a broader multi-seat growth story. That also creates a broader multi-seat growth story.[CU001, CU002, CU006, CU007]

6.2 ROI and vertical adoption

OpenAI’s own enterprise report says workers save 40–60 minutes per day on average and that 75% of surveyed users report faster or higher-quality output. The same report and the company’s vertical case studies point to finance, healthcare, manufacturing, technology, and professional services as the clearest adoption leaders. This is the classic pattern for a platform moving from experimentation into repeatable operating usage. Partner-led case studies show that the company is also moving into complex enterprise buying processes where procurement and change management matter. The practical question is whether those gains persist once the novelty wears off. The company still needs longer-run proof of that pattern. The company still needs longer-run proof of that pattern.[CU003, CU004, CU005, CU009]

6.3 Durability, concentration, and gaps

The customer story is strong, but not complete. OpenAI still does not publish NRR, churn, contract length, or top-account concentration in a reconcilable way. That leaves an important diligence gap: the business appears durable because users keep coming back, but the public record does not yet quantify how concentrated or sticky that durability really is. Those missing metrics are exactly what would separate a fast-growing product from a truly durable enterprise platform. Without those numbers, the market still has to infer quality from usage and seat growth alone. That is the remaining diligence gap for the customer thesis. That is the remaining diligence gap for the customer thesis.[CU008, CU009, CU010]

6.4 Exhibits

7.1 Reliability and customer experience risk

OpenAI’s own status history shows repeated incidents across ChatGPT, the API, login, and model-specific services, and external review surfaces still show meaningful dissatisfaction. For a company that is now infrastructure for work, even modest outage rates or support failures can become enterprise adoption blockers. The broader complaint surface is visible in review sites and user forums, even if exact incident counts remain unpublished. The important diligence issue is not simply whether incidents happen, but whether those incidents now affect customers who are embedding OpenAI into daily work and cannot tolerate long outages. It is now a revenue-retention question as much as a product-ops question.[CR003, CR004, CR009]

7.2 Cyber risk and access controls

The Path to Astra post makes cyber risk explicit by saying Astra meets a critical cybersecurity threshold, while Daybreak adds scoped access controls. That is a sign of maturity, but it also means OpenAI is now managing a product that can be both powerful and brittle: the controls reduce misuse risk, yet they can slow legitimate work or trigger false positives in legitimate workflows. In practice, that means OpenAI is balancing safety against usability every time a sensitive task or cyber-adjacent workflow is attempted. The challenge is to reduce misuse without turning the product into a constant source of interruptions for legitimate users.[CR001, CR002, CR005, CR007, CR008]

7.3 Governance and legal risk

OpenAI’s governance story remains unusual, and that unusual structure invites recurring public scrutiny. The company’s nonprofit-controlled PBC form, the long-running legal and mission disputes around its structure, and the IPO-related legal commentary all mean that governance risk is not a side issue. It is part of the core investment case. For a future public company, that kind of structure can create disclosure friction, stakeholder tension, and extra scrutiny from investors and regulators alike. That makes governance more than a headline risk; it becomes an operating constraint on financing and strategy. That makes the issue material at scale.[CR006]

7.4 Exhibits

8.1 Valuation lens

The latest public valuation is $852 billion, and the company itself says it is generating $2 billion in revenue per month. That combination puts the question squarely on whether rapid scale can become durable operating leverage. The market is currently willing to finance the company as a category-defining platform, but the valuation still has to absorb burn, capital commitments, and the eventual public-listing discount that comes with more disclosure. The key diligence issue is that the market is pricing a future platform, not today's cash generation. That keeps upside open, but it also means the valuation can re-rate quickly if growth or margin signals slip.[CV001, CV004, CV008]

8.2 Growth versus burn

Sacra’s estimates point to $40 billion in annualized revenue, roughly 33% gross margin, and burn that could still be $27 billion in 2026 and $63 billion in 2027. That is enough to support a very large growth narrative, but not yet enough to support a clean mature-software multiple. In other words, the valuation is buying both future margin expansion and execution on a still-capital-intensive business model. The right question is not whether scale exists, but whether scale arrives fast enough to outrun inference spend, support burden, and financing drag. If those curves diverge, the multiple compresses even if revenue keeps rising.[CV002, CV005, CV009]

8.3 IPO path and diligence read-through

The public narrative has already shifted to an IPO frame. The FT and Tech Insider both treat a public listing as live, and the financing stack now includes more strategic participants and broader liquidity options. That lowers funding risk, but it does not remove the core diligence questions: what exactly is the preference stack, how much of the burn is structurally committed, and how fast can the company translate scale into free cash flow? The financing story may still work even before the operating story is clean, which is why governance, side letters, and preference rights matter so much here. Those terms determine how much of the upside actually reaches common holders.[CV003, CV006, CV007]

8.4 Exhibits