OpenAI Deep Value Investment Research

Hunting ten-year five-baggers among great growth stocks — pressing the upside question: "Can it get much bigger?"

• QHow large is its market ceiling? Is it expanding an existing pie, or creating an entirely new market?8/10

  A

  The ceiling is extremely high. OpenAI is both expanding existing markets that are being rewritten by intelligence, such as software, search, and customer service, and trying to use general intelligence to open a new market that does not yet exist. The further the thesis moves toward a brand-new market, the more it rests on long-term belief and the harder it is to validate with cash flow.

  Start with the absolute size of the pie. Citing Gartner, the report says global generative AI spending will be about 644 billion dollars in 2025, up 76.4% year over year, and total global AI spending will reach about 2.52 trillion dollars in 2026, up 44% year over year. A McKinsey survey shows that enterprise use of generative AI has risen from 33% in 2023 to 71% in 2024. These figures can still be externally corroborated: Gartner's global GenAI 2025 spending forecast is exactly 644 billion dollars, with growth of 76.4%. In other words, OpenAI's track itself is expanding much faster than traditional software, so the ceiling is not the constraint.

  More important is its practical grip on expanding the pie. OpenAI is no longer a lab selling a single model. It now controls three entry points at once: consumer access, work access, and developer access. The report states that as of February 2026, ChatGPT had more than 900 million weekly active users, more than 50 million paid consumer subscriptions, and more than 1 million paying enterprise customers. These numbers can be externally checked: OpenAI and multiple media outlets confirmed that ChatGPT had reached 900 million weekly active users and 50 million paid subscriptions. This means the company is not trying to erode one vertical application, but a whole row of existing markets being repriced by intelligence: search, office collaboration, customer service, programming, education, and more. This is the expansion of the existing pie. Its scale is measured in trillions of dollars, and the certainty is relatively high.

  The more attractive story, creating a brand-new market, still has more imagination than proof today. The report notes that management has put advertising, commercial shopping guidance, licensing, and outcome-based revenue sharing on the roadmap, and that the business model follows the idea of expanding with the value intelligence creates for users. OpenAI's own article, “A business that scales with the value of intelligence”, does indeed include subscriptions, usage-based billing, advertising, commerce referrals, and outcome sharing in its commercial logic. The issue is that most of these new fee layers based on taking a share of AI-created value have not yet been monetized at scale. Whether they work, what take rates they can command, and who sets the price all remain unanswered.

  The honest conclusion is that the ceiling is high enough: the existing pie is already at the trillion-dollar level and still growing quickly, and OpenAI is indeed betting on both lines at once. But the part that could truly support a fivefold return in ten years, turning general intelligence into a new market priced by OpenAI itself, is still closer to a vision than a validated business. The report's 3/5 score for business understandability and 3/5 score for moat reflect exactly this compromise: the direction is right and the size is large, but the long-term portion has not yet been proven.

  Jun 11, 2026
• QCan its revenue at least double over the next five years? Will growth be driven mainly by volume, price, or new businesses?8/10

  A

  Revenue can almost certainly double, and by far more than double. That is exactly the issue: the growth is real, and it is driven mainly by volume, meaning explosive growth in users and usage, plus the new-business optionality of new fee layers such as advertising, referrals, and enterprise. It is not driven by price, and the high growth is being bought with equally astonishing cash burn.

  First, anchor the reality of the growth. The report says OpenAI had about 3.7 billion dollars of revenue in 2024 and about 13 billion dollars of actual revenue in 2025. Annualized revenue was 10 billion dollars in June 2025, more than 20 billion dollars at year-end, and more than 25 billion dollars by the end of February 2026. This curve can be externally checked: multiple media outlets confirmed that OpenAI's 2025 revenue was about 13.1 billion dollars, and its annualized run rate had exceeded 25 billion dollars in early 2026, or about 2 billion dollars per month. Based on that slope alone, at least doubling revenue over the next five years is barely in doubt. The company's own message to investors is that 2030 revenue will exceed 280 billion dollars, implying a target of about 20 times the 13.1 billion dollars of 2025 revenue in five years. Even after a large haircut, doubling is more than achievable.

  The growth structure is mainly volume. One side is the step-change expansion in users: weekly active users rose from 800 million in October 2025 to 900 million in February 2026. The other side is penetration of enterprise and developer usage. The report says the enterprise side is expanding quickly, and external sources show that enterprise revenue already accounts for more than 40% of OpenAI revenue and is expected to match the consumer side by the end of 2026. Price, by contrast, is a headwind. The AI model layer is in intense price competition, single-token prices are trending down, and OpenAI's pricing power is constrained by the three-way contest among performance, price, and deployment. Relying on price increases to drive growth is not realistic today.

  New business is the second engine, but it is still an option. The report makes clear that advertising, commerce referrals, and outcome sharing are on the roadmap. These are incremental fee layers beyond subscriptions and API revenue. If they work, they could meaningfully amplify revenue. But their contribution to revenue is still small today, and the timing is unknown. They can enlarge the upside, but they should not be the basis for assuming a doubling.

  The key honesty lies in the quality and cost of that growth. The report notes that adjusted gross margin fell from 40% to 33% in 2025, while inference costs rose fourfold. External sources further confirm that 2025 inference costs increased about 4 times to 8.4 billion dollars, gross margin fell from 40% to 33%, and that was far below the company's own 46% forecast. In other words, while revenue doubles, costs and cash consumption are scaling at the same pace or even faster. The report says OpenAI burned 2.5 billion dollars of cash in the first half of 2025, and the company expects it will not turn cash-flow positive before 2029. More concerning, reports in April 2026 said OpenAI had missed monthly revenue targets several times, and the CFO warned that if growth does not accelerate, it will be difficult to support future compute-contract payments.

  Conclusion: doubling revenue over five years is highly probable and even a low bar. The main drivers are volume, meaning users and usage, plus optionality from new fee layers, not price. But being able to double is not by itself an investment reason. The real unresolved question is whether this high-speed growth can, at some scale point, turn gross margin and cash flow positive together, instead of remaining a business that burns more as it gets bigger.

  Jun 11, 2026
• QFive years from now, what will take over as the next growth engine? Does this second curve exist today?5/10

  A

  The second curve does already exist in the roadmap and in early forms today, but none has yet been proven as a mature engine that can carry the business on its own. They look more like several branches growing from the same intelligence flywheel than a new tree that is already bearing fruit.

  First define OpenAI's current first curve: consumer subscriptions such as ChatGPT Plus/Pro, enterprise/team subscriptions, and usage-based API billing. These are the core support for about 13 billion dollars of 2025 revenue and about 25 billion dollars of annualized revenue in early 2026. See the external figure that OpenAI's annualized run rate exceeded 25 billion dollars in early 2026. The question is what takes over when this main curve slows.

  The report identifies three candidate second curves, none of which is pure PPT anymore. First, advertising and commercial shopping guidance/outcome sharing. The report states clearly that management has included advertising, commerce referrals, licensing, and outcome sharing in the business logic, and OpenAI's own business-model article also confirms this path. Given the entry-point scale of 900 million weekly active users, advertising has the largest imaginative space, but its revenue contribution is still small and its format remains unsettled. Second, enterprise deployment and agent workflows. The report mentions deeper business-process embedding through DeployCo/Frontier Alliance and similar forms. The direction is to move from selling model access to selling agents embedded into workflows. This is the branch closest to current revenue and has the strongest extensibility, but it is essentially depth within the first curve rather than a brand-new pole. Third, expansion into deployment and consulting. The report notes that OpenAI is pushing its boundary outward from making models into services.

  But the implicit premise in this chain question needs to be added: the real test of a second curve is not whether candidates exist, but whether it is independent of the first curve and can self-fund when the main business loses speed. By that stricter standard, all three lines today still depend heavily on the same underlying model capability and the same pool of compute. None has been proven able to create high-quality cash flow independently even if the core model is caught up with. Compared with platforms that have grown truly independent second curves, such as cloud, advertising networks, or hardware ecosystems that each can profit on their own, OpenAI's branches still share one trunk. If that trunk, model leadership, is eroded, the branches will come under pressure at the same time.

  More important, all these second curves consume more compute and burn more money. The report says the company expects compute spending alone to be about 50 billion dollars in 2026 and cumulative compute targets before 2030 to be about 600 billion dollars. External sources confirm that OpenAI has told investors it is targeting about 600 billion dollars of compute spending before 2030. In other words, the cost of growing a second curve is to push capital intensity even higher, not to open new businesses lightly on top of an existing platform the way mature software companies can.

  The honest conclusion: the second curve exists today, and there is more than one. They have all moved beyond the concept stage. But none has been proven able to take over independently if the main business loses speed. For now, they are upside-amplifying options, not a safety net that has already paid off. This is consistent with the report's judgment that both moat and return on capital remain uncertain.

  Jun 11, 2026
• QWhat is its core competitive advantage? Will this moat widen or narrow over the next three to five years?5/10

  A

  The core advantage is a composite of brand default entry point, massive-scale distribution, data feedback loops, and compute-organization capability. But this moat is more like a lead in fast-moving water than a hardened barrier. Over the next three to five years, it can broadly hold on the consumer side, while it is already narrowing on the enterprise side.

  Start with the real sources of the moat. The report summarizes them in five points: brand and default entry point, with ChatGPT among the strongest AI mindshares among the public; scale and distribution, with 900 million weekly active users and tens of millions of paying users; data and feedback loops, where high-frequency interaction helps product optimization; developer and enterprise embedding through API, Business, and Enterprise; and compute-organization capability, no longer relying on a single cloud vendor. Distribution scale is a hard asset: ChatGPT has reached 900 million weekly active users and 50 million paid subscriptions. This gives product iteration, monetization experiments, and ecosystem expansion an extremely high starting point, and the probability that consumer mindshare is overturned wholesale within three to five years is not high.

  The key issue is that this is not the type of strong consumer-goods moat where price can be raised without losing volume. The report states that enterprise-customer loyalty is constrained by three variables: performance, price, and deployment/governance. If any one of these is overtaken by competitors, stickiness loosens. The overtaking has already happened on the enterprise side. The report cites Ramp as saying Anthropic has exceeded OpenAI in paid enterprise AI adoption in the United States for the first time. This inflection point can be externally verified and is still worsening: Ramp's May ranking shows that Anthropic's enterprise adoption rose to 34.4%, OpenAI fell to 32.3%, Anthropic quadrupled its enterprise adoption over the past year, while OpenAI grew only 0.3%. By June 2026, Anthropic had risen to about 41% and became the most adopted model in enterprises. The driver is the breakout of Claude Code, a coding agent. This directly hits the enterprise-embedding pillar inside OpenAI's moat.

  Competitors' structural advantages make narrowing a trend rather than noise. The report notes that in 2026 Google unified enterprise AI under Gemini Enterprise and controls cloud, data, chips (TPU), governance, and native enterprise relationships. Microsoft, although adjusting the partnership, still holds Azure channels and deep enterprise distribution. In other words, enterprise competitors combine channels, cloud, and data in one stack, while OpenAI's grip outside the model layer is relatively thinner.

  Even more important, the restructuring of OpenAI's relationship with Microsoft objectively weakens one of its original distribution moats while also freeing it. The report says Microsoft's IP license to OpenAI became non-exclusive, continuing to 2032, OpenAI can serve customers across clouds, and Microsoft no longer pays OpenAI revenue share. External sources confirm that in April 2026 the two sides revised the agreement: Microsoft's IP license changed from exclusive to non-exclusive through 2032, OpenAI can serve customers on any cloud, and a total cap was placed on OpenAI's revenue share to Microsoft. This is double-edged. It frees OpenAI from single-cloud binding and reduces the single-point risk of not getting chips, but it also means OpenAI no longer has exclusive access to Microsoft's enterprise distribution channel and must compete for enterprise customers directly on product strength.

  The honest conclusion: the consumer moat is broadly solid for three to five years. The enterprise moat is being materially thinned by Anthropic in coding/workflows and by Google in cloud plus data plus chips plus governance. The direction is narrowing, not widening. This is fully consistent with the report's 3/5 moat score and its explicit judgment that there is narrowing risk on the enterprise side.

  Jun 11, 2026
• QIf its core business is disrupted, does it have the genes for self-reinvention? How does it treat mistakes and bad news?5/10

  A

  The genes for self-reinvention are very strong. OpenAI has repeatedly iterated and even disrupted its own products and business model. But on how it treats mistakes and bad news, blemishes in its governance history make it hard to feel comfortable about internal conflict and information transparency. The two need to be judged separately.

  First, look at the implicit premise in the chain question: the self-reinvention gene when the core business is disrupted. OpenAI's record is quite notable here. The flywheel described by the report is more compute → better models → more usage and revenue → more reinvestment. The company has kept rewriting itself from a model-selling lab into an intelligence platform with tiered monetization, and it has actively put advertising, commerce referrals, outcome sharing, and other new fee layers into its commercial logic, as seen in OpenAI's own business-model article. What shows self-reinvention even more clearly is the restructuring of its dependence on infrastructure. The report says the company no longer relies on a single cloud vendor and has diversified compute across partners including Stargate, Oracle, SoftBank, and Amazon. External sources say the latest round involved about 30 billion dollars each from SoftBank and Nvidia, and about 50 billion dollars of Amazon compute credits. The company has proactively dismantled the historical structure of being tied to Microsoft's single cloud. A company willing to keep overturning its product form, business model, and supply structure usually has strong adaptability when disruption arrives.

  But the record on how it treats mistakes and bad news is a deduction. The report describes in detail that in November 2023 the board publicly removed Sam Altman because he had not been consistently candid in his communications with the board. In the March 2024 review, WilmerHale found that the core issue was a breakdown of trust and confirmed that the prior board's original statement accurately reflected the decision and reasoning at the time, while ultimately concluding that the conduct did not necessarily require removal. This means the company has experienced a serious failure of trust and communication when facing internal disagreements.

  In 2025–2026, governance complexity did not disappear; it became institutionalized, and new tests of bad-news handling have continued to surface. The report mentions that court filings in May 2026 showed Altman held more than 2 billion dollars of equity across nine companies that had business dealings with OpenAI, prompting multiple state attorneys general to ask the SEC for a review. This can be externally verified: court filings show Altman held more than 2 billion dollars in companies transacting with OpenAI, including about 1.7 billion dollars in the fusion company Helion, and ten Republican state attorneys general wrote to the SEC requesting strict scrutiny before an IPO. At the same time, reports in April 2026 said OpenAI had missed monthly revenue and user targets several times, and the CFO warned that if growth did not accelerate, it would be difficult to support compute contracts. This type of bad news appears to have been forced into view more by external reporting and legal processes than by proactive, transparent company disclosure. That is far from the value investor's expectation that bad news should be disclosed immediately and voluntarily.

  The honest conclusion: in product, business-model, and supply-structure self-reinvention, OpenAI has strong genes, and adaptability to disruption is one of its few positive marks. But in honesty and transparency when facing internal mistakes and bad news, the record is blemished, from the 2023 breakdown of trust to the 2026 related-equity controversy. This is why the report judges management trustworthiness as uncertain leaning negative and scores management and capital allocation at 2/5. It is exactly the dual assessment: able to reinvent, but not necessarily candid.

  Jun 11, 2026
• QDoes management, especially the founder, have a long-term vision and interests deeply tied to the company? Is it willing to sacrifice current profits for five to ten years out?4/10

  A

  The long-term vision is extremely strong, and the willingness to sacrifice current profits for five to ten years out is unequivocal. OpenAI gets full marks on those two points. But founder interests deeply tied to the company is precisely the awkward part: Altman does not get rich through OpenAI equity. That weakens the incentive to push the stock price up in the short term, but it also weakens the classic founder alignment where one's net worth is tied to the same stock. It is further complicated by conflict-of-interest controversy around related holdings.

  Start with long-term vision and sacrificing the present for the future. This point is almost beyond reproach. The report says the company does not expect to turn cash-flow positive before 2029, compute spending alone in 2026 is expected to be about 50 billion dollars, and cumulative compute targets before 2030 are about 600 billion dollars. External sources confirm that OpenAI has anchored compute spending before 2030 at about 600 billion dollars and expects cumulative cash burn by 2030 to be about 665 billion dollars, while setting a 2030 revenue target above 280 billion dollars. A company that knowingly will not turn positive for five years and still dares to put hundreds of billions of dollars behind long-term compute and models has written the willingness to sacrifice current profits for five to ten years out into its financial structure. There is no need to doubt that.

  But founder interests deeply tied to the company needs to be separated, and the conclusion is negative. The report points out a counterintuitive fact: in October 2025, reports said OpenAI publicly stated that Altman would not directly receive company equity in the new structure. This can be externally checked. Multiple reports confirm that Altman himself does not hold OpenAI equity and does not get rich through OpenAI equity. This makes alignment double-edged. The good side is that he has no direct incentive to push valuation up for personal shareholding gains in the short term. The bad side is that he also lacks the strong alignment typical of founders whose personal net worth lives and dies with the stock price. For a ten-year growth bet, founder non-ownership actually weakens the signal that he is in the same boat as minority shareholders.

  The more complex issue is the conflict of interest created by related holdings. The report says court filings in May 2026 showed Altman held more than 2 billion dollars in total across nine companies that had business relationships with OpenAI, prompting multiple state attorneys general to request an SEC review. External sources confirm that Helion (fusion) accounted for about 1.7 billion dollars, Stripe about 633 million dollars, and Retro Biosciences about 258 million dollars, and ten Republican state attorneys general wrote to the SEC requesting a review of self-dealing concerns before an IPO. In other words, when OpenAI signs contracts with or spends money on suppliers in which Altman personally has large stakes, there is controversy over blurred boundaries between his personal portfolio and company interests. This is the opposite of founder interests aligned with the company; it is a potential reverse binding.

  Add the governance-history discount: the report records that in November 2023 the board removed Altman for not being consistently truthful and candid in communications, and the 2024 review confirmed that the core issue was a breakdown of trust. Although governance is now institutionalized, with the nonprofit foundation controlling OpenAI Group directors through special voting rights, the true boundary of minority shareholder rights remains in motion.

  The honest conclusion: long-term vision and sacrificing the present for the future are full-mark items. But founder interests deeply tied to the company is a deduction: Altman does not hold OpenAI equity, making alignment weaker, and his more than 2 billion dollars of personal holdings in related companies have triggered self-dealing scrutiny, creating potential reverse alignment. This is the core basis for the report's 2/5 score on management and capital allocation and its judgment that management trustworthiness is uncertain leaning negative.

  Jun 11, 2026
• QIf it disappeared tomorrow, how much would customers miss it? Is its growth model sustainable and not dependent on social harm or regulatory friction?5/10

  A

  If it disappeared tomorrow, consumers would miss it greatly, while enterprise customers would miss it but have ready substitutes. Indispensability is high on the consumer side and being diluted on the enterprise side. Its growth model is not clean on social and regulatory sustainability: copyright, privacy, and antitrust lines are all real, ongoing legal and regulatory pressures.

  First, how much would customers miss it? This needs to be split by customer layer. Consumers would miss it a lot. ChatGPT is the mass default entry point with 900 million weekly active users (ChatGPT has reached 900 million weekly active users and 50 million paid subscriptions), and for large numbers of individual users it is already embedded into daily workflows. There is no single experience-equivalent substitute in the short term, so disappearance would create real withdrawal. But enterprise attachment is being diluted. The report cites Ramp as saying Anthropic has overtaken OpenAI in enterprise adoption, and external sources confirm that Anthropic's enterprise adoption has risen to about 34.4%, OpenAI has fallen to 32.3%, and by June 2026 Anthropic rose further to about 41% to become the most adopted model in enterprises. Enterprise customers face a multi-vendor market where performance, price, and deployment are comparable. If OpenAI disappeared, a meaningful portion of enterprise workloads could move to Anthropic or Gemini. Migration friction exists, but it is not fatal. That is exactly the discount to indispensability on the enterprise side.

  Now add one implicit premise in the chain question: whether the growth model depends on social harm and regulation. OpenAI is not clean here, and all three lines have landed in specific proceedings:

  First, copyright. The report says multiple author and media copyright lawsuits have been consolidated in New York. This can be externally checked: multiple copyright-infringement lawsuits against OpenAI have been consolidated by the multidistrict litigation panel into the Southern District of New York (In Re: OpenAI, Inc. Copyright Infringement Litigation, 1:25-md-03143), with Judge Sidney Stein overseeing pretrial proceedings, and in October 2025 the court denied parts of OpenAI's motion to dismiss. The legality of model-training data is a structural issue hanging over its growth model.

  Second, privacy. The report mentions that Italy's privacy regulator fined OpenAI 15 million euros. External sources confirm that Italy's data protection authority, Garante, fined OpenAI 15 million euros in December 2024 for processing personal data without adequate legal basis and violating transparency obligations. With EU AI rules and data transparency requirements still evolving, compliance cost is moving up rather than down.

  Third, antitrust/governance scrutiny. The report records that Altman held more than 2 billion dollars in nine related companies, and ten state attorneys general requested SEC scrutiny (court filings and a letter from ten Republican state attorneys general to the SEC). Although this leans toward governance, when combined with the regulatory spotlight before an IPO, it further pressures social and regulatory sustainability.

  The honest conclusion: consumer indispensability is high, while enterprise indispensability is being diluted. The social and regulatory sustainability of the growth model is a clear deduction. Copyright through consolidated litigation, privacy through a 15 million euro fine, and governance/antitrust through an SEC review request are all real and ongoing, not abstract risks. The report accordingly lists this as one of the core risks and gives no social-sustainability bonus to the moat. That matches this question's conclusion.

  Jun 11, 2026
• QWhat are the unit economics of this business, in gross margin and incremental returns? Do they improve or worsen as scale grows? Where does the money it earns go?3/10

  A

  The unit economics are currently poor and worsening. Adjusted gross margin is falling rather than rising, incremental returns are being swallowed by inference costs, and greater scale has not yet produced the lighter economics of software. The money it earns and raises is almost entirely going into R&D, compute, channels, and deployment, with no dividends and no buybacks. This is the biggest distinction between OpenAI and traditional high-gross-margin software companies.

  Start with gross margin, the most direct thermometer for unit economics, and the reading is moving down. The report says adjusted gross margin fell from 40% to 33% in 2025 because inference costs rose fourfold. External sources not only confirm this but make it look worse: in 2025 OpenAI's adjusted gross margin fell from 40% to 33%, far below its own 46% forecast, while inference costs increased about 4 times to 8.4 billion dollars, above the 6.6 billion dollars forecast last summer, and 2026 inference costs are expected to rise to 14.1 billion dollars. In other words, scale is growing, yet unit gross profit is being eroded in reverse by compute costs. This is the opposite of the mature SaaS pattern where larger scale drives marginal cost toward zero and gross margin thickens.

  Now look at incremental returns and cash flow. The report gives a crucial set of facts: first-half 2025 revenue of 4.3 billion dollars, cash consumption of 2.5 billion dollars, R&D expense of 6.7 billion dollars, and ending cash and securities of 17.5 billion dollars. The company expects it will not become cash-flow positive before 2029. Put the report's conservative estimate of owner earnings plainly: it explicitly does not apply the standard formula of net income plus depreciation and amortization minus maintenance capital expenditure because public data are lacking, and instead conservatively treats 2025 owner earnings as a negative value of about -5 billion to -8.5 billion dollars. This means each additional dollar of revenue currently does not settle into positive free cash flow. Incremental returns are negative.

  Do they improve or worsen as scale grows? In the short term, they worsen; over the long term, it is unresolved. The report's core inference is that OpenAI today looks more like a company exchanging future scale and network position for today's cash consumption and infrastructure commitments. The most important and fragile parts of its cost structure are not sales expense but compute, inference, R&D, and elite talent. This is a hybrid of software plus compute-infrastructure commitments, not light-capital SaaS. In theory, if inference costs decline with hardware and model-efficiency improvements, and revenue scale absorbs fixed R&D, unit economics can improve. But today's direction is deterioration: gross margin 40%→33%, inference costs up fourfold. Improvement remains an assumption, not a fact.

  Finally, where does the money it earns go? The answer is almost entirely reinvestment, and at astonishing scale. The report says compute spending alone is about 50 billion dollars in 2026, cumulative compute targets before 2030 are about 600 billion dollars, and there are no dividends and no buybacks. External sources confirm that OpenAI has anchored compute spending before 2030 at about 600 billion dollars and cumulative cash burn at about 665 billion dollars. For a high-growth immature company, full reinvestment is not inherently wrong. But from a unit-economics perspective, it means you are not buying a machine that is already putting cash in your pocket. You are buying a machine that keeps consuming capital and whose unit economics remain unproven.

  The honest conclusion: current unit economics are poor and worsening, with declining gross margin, negative incremental returns, and no cash-flow positivity before 2029. Scaling has not yet delivered software economics. All money is going into compute, R&D, and channels, with no shareholder returns. The report therefore judges stable free cash flow as failed and excellent return on capital as uncertain. This question is fully consistent with that.

  Jun 11, 2026
• QWhat conditions must all be true for it to rise fivefold in ten years? Are those conditions realistic? What expectations are embedded in today's share price?3/10

  A

  For a private valuation of about 852 billion dollars to rise fivefold over ten years, to about 4.3 trillion dollars, a string of difficult conditions must all come true at the same time. Today's price already embeds expectations of near-perfect execution. It sits roughly in the low-to-middle part of the report's optimistic valuation range, with a negative margin of safety. Honestly, a fivefold return in ten years is not impossible, but it requires not merely doing well; it requires almost everything to go right while competitors fail to catch up.

  Anchor the starting point first. The strongest public pricing anchor is that at the end of March 2026 OpenAI completed about 122 billion dollars of financing at a post-money valuation of about 852 billion dollars, making it the most expensive private company in history. A fivefold return in ten years would mean a market value of about 4.26 trillion dollars, already larger than any listed company in the world today.

  Add one implicit premise in the chain question: what conditions must all be true for a fivefold return in ten years. Working backward from the report's valuation logic, at least the following must all be met:

  1. Revenue must materialize at the hundreds-of-billions-of-dollars level. Citing external reports, the report says the company's 2030 revenue target is above 280 billion dollars (OpenAI's own 2030 revenue target given to investors exceeds 280 billion dollars), and high growth would then need to continue toward around 2035. This is the premise of the report's optimistic scenario.
  2. Margins must move from negative to high. The report's optimistic scenario assumes an owner-earnings margin of 22%–25% in 2035. The reality is that in 2025 adjusted gross margin fell from 40% to 33%, and inference costs rose 4 times to 8.4 billion dollars and are expected to reach 14.1 billion dollars in 2026. To move from today's negative owner earnings to 20%+, inference costs must structurally fall faster than price competition.
  3. The enterprise side must stop bleeding and regain leadership. The report says the enterprise side is being overtaken by Anthropic (Ramp's measure shows Anthropic's enterprise adoption has exceeded OpenAI's and reached about 41% in June 2026). The fivefold narrative requires OpenAI at least not to keep losing share in the most profitable enterprise layer.
  4. Compute commitments must not bite back against cash flow. The report says cumulative compute targets before 2030 are about 600 billion dollars. A fivefold return requires these hundreds of billions of dollars of investment to become high returns, not fixed obligations that crush cash flow when growth slows. Yet in April 2026 the CFO warned that if growth does not accelerate, it will be difficult to support compute contracts.
  5. Governance and regulation must stop disturbing shareholder rights. Related-holding SEC scrutiny, consolidated copyright litigation, privacy fines, and similar issues all need to fade, and IPO valuation must not detach from fundamentals.

  For all five to be true at the same time is not very realistic. If any one fails obviously, the fivefold logic breaks.

  Now answer what expectations today's share price embeds. This is the part that should be stated most plainly. The report's reverse valuation is clear: the current valuation of about 852 billion dollars roughly sits in the low-to-middle part of the optimistic scenario range of 900 billion–1.3 trillion dollars and is already above the report's reasonable intrinsic-value range of 350 billion–600 billion dollars. As an intuitive sanity check, 852 billion dollars against about 13 billion dollars of 2025 revenue is about 66 times sales, and even against more than 25 billion dollars of early-2026 annualized revenue is about 34 times annualized revenue. By contrast, as mature profitable comparisons, Microsoft's current PE is about 24 times, and Alphabet's is about 29 times; both are large-scale profitable public companies with mature cash flow and strong liquidity. In other words, today's price has already bought forward most of the expectation that OpenAI becomes a consumer + enterprise + developer super-platform and delivers high owner earnings around 2030. The report further calculates that to support today's valuation and still receive a mature-platform terminal multiple of 20–25 times 2030 owner earnings, OpenAI would need about 50 billion–62 billion dollars of sustainable owner earnings in 2030, implying an 18%–22% margin on 280 billion dollars of revenue. That is itself an extremely high embedded hurdle.

  The honest conclusion: a fivefold return in ten years requires a series of low-probability conditions to all occur at once: hundreds of billions in revenue, margins moving from negative to 20%+, enterprise share turning back up, compute commitments not biting back, and governance/regulatory issues calming down. Today's price already embeds near-perfect execution and has a negative margin of safety. This is the core arithmetic behind the report's failing judgments on valuation below intrinsic value and sufficient margin of safety, and its final Watch rating.

  Jun 11, 2026
• QWhy has the market not realized all this yet? Is it because it does not understand, looks down on it, or cannot look far enough? What will become the narrative inflection point?3/10

  A

  For OpenAI, this question needs to be asked in reverse. The market has not failed to realize its greatness; it has already priced that greatness fully, and even excessively. It is not being looked down on, nor is the market unable to look far enough. It is being looked at too fully: the roughly 852 billion dollar valuation has already bought forward the most optimistic growth story. What the market has not fully absorbed is the downside: the difficulty of turning profitable, enterprise-side loss of ground, and governance/regulatory tail risks. So the narrative inflection point is more likely downward disenchantment than upward discovery.

  First correct the premise of the question. The normal use of this Baillie Gifford question is to find a neglected gem that the market does not understand. OpenAI is the opposite: it is one of the most watched and richly priced private assets in the world. At the end of March 2026, it raised about 122 billion dollars at a post-money valuation of about 852 billion dollars, making it the most expensive private company in history. On about 13 billion dollars of 2025 revenue, that is about 66 times sales, far above mature profitable companies such as Microsoft at about 24 times PE and Alphabet at about 29 times PE. The market does not lack willingness to look far ahead for OpenAI. It lacks enough discount for whether the story can land.

  Now break down understand/look down/not look far enough:

  • Does not understand? Partly true, but in the opposite direction. The report repeatedly emphasizes that OpenAI's financial transparency is below public-company standards: no complete audited statements, no share-dilution table, no standard cash-flow statement, and revenue figures mix ARR, annualized revenue, run rate, and other measures. So there is indeed a lack of visibility. But in private-market enthusiasm, the narrative has overwhelmed that lack of visibility. The result is not undervaluation; it is a high valuation despite information gaps.
  • Looks down on it? No. No one is dismissing OpenAI. It is the synonym for the AI wave.
  • Cannot look far enough? Also no. If anything, the opposite: the market is looking too far. The report states that the current price already sits in the low-to-middle part of the optimistic valuation range of 900 billion–1.3 trillion dollars and has already discounted the long-term imagination of more than 280 billion dollars of 2030 revenue and 20%+ margins. That is looking too far, not failing to look far enough.

  So what has the market not fully realized? It is that downside risks have begun to materialize, but the valuation has not reflected them. Citing external reports, the report notes that in April 2026 OpenAI had missed monthly revenue and user targets several times, and CFO Sarah Friar warned that if growth does not accelerate, it will be difficult to support future compute contracts. On the enterprise side, Anthropic has overtaken it in adoption and reached about 41% in June 2026. Gross margin has slipped from 40% to 33%, and inference costs have risen 4 times to 8.4 billion dollars. These cracks have not been fully incorporated into the high valuation.

  Add the implicit premise in the chain question: what will become the narrative inflection point. For an asset priced too richly, the inflection point is more likely to be a downward disenchantment signal. The report lists several categories to watch:

  1. Continued enterprise loss of ground: if Anthropic (Claude Code/coding workflows) and Google (Gemini Enterprise+TPU+cloud) further widen the adoption gap in enterprises, the mindshare story that OpenAI equals the AI leader will be rewritten.
  2. Further delay in the profit path: if 2029–2030 cash-flow positivity keeps being pushed back and gross margin continues to fall, the market will switch from valuing it as a growth platform to valuing it as a high-growth but high-cash-burn business.
  3. Governance/regulatory hammer drops: if Altman's more than 2 billion dollars of related holdings, which prompted ten state attorneys general to request SEC scrutiny, consolidated copyright litigation, or privacy fines produce major adverse outcomes, they will become sentiment inflection points.
  4. IPO pricing: a listing will, for the first time, bring the private-market narrative under public-market cash-flow discipline. If the IPO valuation detaches from fundamentals or the stock breaks issue price after listing, the narrative will reverse quickly.

  The honest conclusion: OpenAI is not a greatness the market has not yet realized. It is greatness that the market has already over-recognized and priced in advance. The direction of the perception gap is negative. What has not been fully priced in is downside: the difficulty of profitability, enterprise-side loss, and governance/regulatory tail risk. Therefore the narrative inflection point is more likely downward disenchantment, from continued enterprise losses, further profit delays, regulatory hammer drops, or IPO validation, rather than upward discovery. This is internally consistent with the report's Watch rating and its explicit judgment that the current price is not cheap enough and the margin of safety is insufficient.

  Jun 11, 2026