OpenAI IPO Economics in 2026: $40 Billion Revenue Run Rate, $750 Billion Compute Plan and the Race to Become Profitable

OpenAI's annualized revenue run rate has reached roughly $40 billion, nearly double its pace at the end of 2025, while the company prepares for a possible public-market transition. But the number that may matter just as much is $750 billion: the approximate cloud-computing spending OpenAI is now reported to be planning through 2030. Between those figures sits the central question any future OpenAI investor will have to answer: can rapidly improving economics per unit of intelligence become durable company-level profitability before the cost of building the next generation of intelligence outruns even OpenAI's extraordinary revenue growth?

Artificial Intelligence and Markets · August 15, 2026 · KG-OPENAI-IPO-001 · Last Verified August 15, 2026

Research basis: This report was developed from POPR's final sealed, post-hostile-audit OpenAI IPO knowledge graph and reconciled against OpenAI's own first-party disclosures alongside financial reporting from Bloomberg, the Financial Times, Reuters, The Information, The Wall Street Journal, Axios and CNBC. Annualized revenue run rate is kept separate from annual revenue. Cash burn, operating loss and net loss are kept separate from one another. OpenAI's own statements about IPO timing are distinguished from externally reported possibilities, and improving gross margins are presented alongside, rather than obscured by, the company's extraordinary R&D and infrastructure spending. KG-OPENAI-IPO-001.

OpenAI's financial story can be made to sound spectacular or alarming depending on which number is selected.

The bullish version begins with revenue. OpenAI's annualized revenue run rate has risen from more than $20 billion at the end of 2025 to approximately $24 billion by the end of March 2026 and roughly $40 billion by August. Bloomberg reported the latest acceleration on August 13, with the Financial Times independently corroborating the approximately $40 billion level.

The bearish version begins with spending. OpenAI burned approximately $3.7 billion of cash during the first quarter of 2026, reported an operating loss of roughly $9.3 billion, spent approximately $8.6 billion on research and development during the same quarter, and is now reported to be planning approximately $750 billion of computing expenditure through 2030.

Neither version is sufficient. The more important story is that OpenAI is trying to make artificial intelligence dramatically cheaper to produce and consume while simultaneously spending at a scale rarely associated with a company that is still private. Those two forces are moving in opposite directions: the economics of delivering each unit of intelligence appear to be improving, while the amount of intelligence OpenAI wants to build and sell is expanding so rapidly that the total bill continues to grow. That tension may ultimately define the economics of an OpenAI IPO.

Is OpenAI Really Making $40 Billion a Year?

Not in the accounting sense that phrase normally implies. The approximately $40 billion figure is an annualized revenue run rate, and that distinction should survive every discussion of OpenAI's financial performance.

A run rate takes a recent revenue pace and annualizes it. It does not mean OpenAI has already recorded $40 billion of audited revenue over the preceding twelve months. Bloomberg reported that OpenAI's annualized revenue run rate had exceeded $40 billion by August 2026, roughly twice the level reported at the end of 2025.

The growth path itself is nevertheless extraordinary. OpenAI CFO Sarah Friar previously disclosed that the company ended 2025 at more than a $20 billion annualized revenue pace. By March 31, 2026, OpenAI was generating roughly $2 billion per month, equivalent to approximately $24 billion annualized. By August, that pace had reached roughly $40 billion. That does not establish what OpenAI's final 2026 revenue will be. It establishes that the company's current commercial velocity has increased dramatically. And for investors considering a future IPO, velocity matters. The question is what happens to the economics underneath it.

OpenAI's First-Quarter Numbers Tell Two Different Stories

OpenAI generated approximately $5.7 billion of revenue in the first quarter of 2026, according to The Information, citing shareholder documents. The company burned approximately $3.7 billion of cash during that same quarter. Its gross margin reached approximately 39%, up from 33% a year earlier.

Those three figures need to be read together. The $5.7 billion shows substantial revenue. The $3.7 billion cash burn shows that the company is still consuming enormous amounts of capital. The six-percentage-point gross-margin improvement shows that the economics of delivering its products are improving despite that burn.

This is where simplistic descriptions such as "OpenAI loses money on every dollar it makes" become inadequate. Gross margin measures something different from total company profitability. A rising gross margin suggests that the direct economics of providing OpenAI's products and services improved year over year. That does not mean OpenAI is profitable. It means the company is showing evidence that greater revenue does not necessarily require proportionately greater direct delivery cost. The remaining problem is everything being spent beyond that layer.

Why Did OpenAI Report a $9.3 Billion Operating Loss if Cash Burn Was $3.7 Billion?

Because accounting loss and cash consumption are not the same measurement. OpenAI's reported first-quarter financial hierarchy needs to be kept intact and the figures should never be collapsed into one another.

Revenue was approximately $5.7 billion. Cash burn was approximately $3.7 billion. Operating loss was approximately $9.3 billion. Net loss was more than $21.3 billion. The approximately $9.3 billion operating loss included non-cash items, including more than $2.3 billion in stock-based compensation. The more than $21.3 billion net loss included an approximately $12.4 billion non-cash accounting charge associated with revaluing convertible interests and warrant liabilities. The $3.7 billion cash-burn figure answers the most direct cash question: how much net cash left the company during the quarter. The operating-loss number answers a broader accounting question about the cost of running the business. The net-loss figure incorporates still more accounting effects. All are real financial measures that describe different things.

The $8.6 Billion R&D Number May Be More Important Than the Loss Headline

OpenAI spent approximately $8.6 billion on research and development in the first quarter of 2026, according to the shareholder-document reporting. That was more than the company's entire quarterly revenue. The number helps explain why improving gross margins have not translated into overall profitability.

OpenAI is not simply paying to serve existing ChatGPT users and API customers. It is simultaneously funding the development of the products and models intended to generate future revenue. That creates an unusual economic structure: the company can become more efficient at providing today's intelligence while spending enormous amounts trying to create tomorrow's intelligence. For an ordinary mature software company, investors often expect scale to produce increasing operating leverage. For OpenAI, the cost of remaining at the frontier may itself expand with scale. That is the unresolved economic problem.

OpenAI's Gross Margin Is Improving

This may be the most important balancing fact in the entire profitability discussion. OpenAI's first-quarter gross margin improved from approximately 33% a year earlier to 39% in Q1 2026. That means the evidence does not support a blanket claim that OpenAI has failed to generate efficiency as it grows. At the gross-margin level, efficiency improved.

The more accurate conclusion is narrower. OpenAI is showing better delivery-level economics while company-wide spending remains far above the level required for profitability. That difference matters enormously to the IPO thesis. If gross margin continues improving while revenue compounds rapidly, OpenAI could eventually create substantial operating leverage. If frontier-model R&D and infrastructure requirements increase faster than those gains can accumulate, the company could remain structurally capital-intensive even at enormous revenue scale. Both paths remain possible from the evidence currently available.

OpenAI Is Also Making Intelligence Cheaper

The pricing story adds another layer. On July 30, OpenAI reduced pricing on two GPT-5.6 models. GPT-5.6 Luna's price fell approximately 80%, to $0.20 per million input tokens and $1.20 per million output tokens. GPT-5.6 Terra fell approximately 20%, to $2 per million input tokens and $12 per million output tokens. GPT-5.6 Sol pricing remained unchanged. OpenAI attributed the reductions to efficiency improvements in how the models are built and served, while Axios reported the company's position that the changes allowed it to deliver more intelligence per dollar.

That is economically significant. If OpenAI can reduce the price customers pay while simultaneously improving its own underlying cost structure, lower unit prices do not necessarily imply worse economics. Lower prices can increase consumption, make more use cases financially practical, allow businesses to automate tasks that were previously too expensive, and drive usage high enough to compensate for declining price per unit. The sealed graph does not assume that outcome indefinitely. It establishes that OpenAI is actively pursuing lower-cost intelligence as a commercial strategy while its gross margin is already showing measurable improvement.

One important current-data warning applies. OpenAI's indexed pricing pages were showing inconsistent Terra and Luna prices during the verification pass, with some pages reflecting older prices and others reflecting the announced July 30 reductions. The price-cut announcement itself is verified, but any publication relying on the exact live billed price at a later date should recheck OpenAI's current pricing directly.

This Is the Core OpenAI IPO Question: Can Unit Economics Beat Infrastructure Economics?

That is where the entire story converges. At one level, OpenAI is pushing the cost of intelligence downward. At another, the absolute cost of building the infrastructure required to produce more intelligence is moving upward.

The company expected to spend approximately $50 billion on computing power in 2026 alone, according to testimony from OpenAI co-founder Greg Brockman reported by Reuters in May. Earlier in 2026, Reuters reported a roughly $600 billion compute-spending outlook through 2030. By July, The Wall Street Journal reported that OpenAI's planned cloud-computing spending had increased to roughly $750 billion through 2030. The change from $600 billion to $750 billion is not merely a correction to an old headline. It is part of the story: as OpenAI's commercial business expands, its expected infrastructure bill is also expanding.

That raises the defining economic question. Can OpenAI make each unit of intelligence cheaper faster than total demand for compute causes its aggregate infrastructure spending to rise? If yes, the company could build extraordinary economics at enormous scale. If not, rapid revenue growth could coexist with persistent capital consumption for much longer than investors accustomed to conventional software economics expect.

$750 Billion Does Not Mean OpenAI Is Writing One Check

The cumulative infrastructure figures also require discipline. The reported approximately $750 billion is planned cloud spending through 2030. It should not be interpreted as one current capital expenditure. The Information separately reported that OpenAI held roughly $665 billion in spending commitments at the end of 2025 extending through 2030. The graph treats those figures as potentially overlapping but not established as identical accounting concepts. Adding $665 billion and $750 billion together to produce a $1.4 trillion obligation would therefore be unsupported. The cleaner near-term figure is Brockman's approximately $50 billion expected 2026 compute spending. The larger figures show the prospective scale of the multi-year infrastructure program.

How Can OpenAI Keep Spending Like This?

Capital. OpenAI announced on March 31 that it had closed a funding round with $122 billion in committed capital at an $852 billion post-money valuation. That is a first-party OpenAI disclosure and the canonical valuation figure.

The size of the financing changes how the cash-burn discussion should be interpreted. OpenAI ended the first quarter with more than $73 billion in cash and marketable securities, up from roughly $40 billion at the end of December. The increase was driven primarily by financing rather than internally generated cash. That does not make cash burn irrelevant. It means OpenAI has so far demonstrated extraordinary access to outside capital while burning it. The company is financing an industrial-scale buildout before becoming a public company. The eventual IPO question is therefore partly about whether public markets are willing to continue underwriting a strategy that private investors have already funded at historic scale.

Is OpenAI Actually Going Public?

OpenAI has taken the formal preliminary step. On June 8, the company confirmed that it had confidentially submitted a draft S-1 to the U.S. Securities and Exchange Commission. That is not speculation. The company filed.

What remains uncertain is timing. OpenAI's own statement was unusually explicit: it had not decided when to proceed and said going public "may be a while" because certain things may be easier to accomplish while remaining private. The filing gives the company the option to move sooner if that becomes preferable. That statement should govern the factual baseline.

Separately, Reuters reported that a source said an offering could occur as early as September 2026 and could value OpenAI at as much as $1 trillion. Those are not equivalent claims. OpenAI confirmed the S-1. Reuters reported a possible date and valuation based on external sourcing. And the timetable has continued moving. The Financial Times reported on August 15 that an IPO once expected in 2026 may now be more likely in 2027 amid broader organizational changes. The durable conclusion is that the confidential S-1 exists and the timing remains unsettled.

Could OpenAI Really Be Worth $1 Trillion?

Mathematically, a $1 trillion valuation against a $40 billion current annualized revenue run rate equals approximately 25 times current run-rate revenue. That calculation is correct. Its interpretation is more complicated.

An investor evaluating an eventual OpenAI IPO does not have to price the company only against its August 2026 run rate. A valuation could be evaluated against trailing revenue, current run rate, forward-one-year revenue, forward-two-year revenue, expected gross margins, expected future cash generation, growth trajectory, infrastructure position, competitive advantage, or combinations of those factors. A company whose revenue is changing as quickly as OpenAI's can produce dramatically different valuation multiples depending on which period becomes the denominator. That means "25 times revenue" is simultaneously mathematically correct and economically incomplete. If growth decelerates sharply, a trillion-dollar valuation could look extremely expensive. If revenue continues compounding at extraordinary rates and gross margins continue improving, the same valuation could appear very different against future revenue. The IPO will force public investors to decide how much future success they are willing to purchase in advance.

The $852 Billion Private Valuation Already Sets an Extraordinary Starting Point

OpenAI does not have to reach $1 trillion to become one of the largest technology listings ever contemplated. Its March financing already established an $852 billion post-money private valuation. A future IPO would not begin from a venture-stage valuation and suddenly discover public-market pricing. OpenAI is approaching public markets after private investors have already assigned the company a value approaching a trillion dollars.

That creates an unusual challenge. The company must persuade public investors not merely that OpenAI is important, which is increasingly obvious, but that enormous future economic value still remains above a valuation that already incorporates extraordinary expectations.

What Would OpenAI Investors Actually Be Buying?

They would be buying a company with several seemingly contradictory attributes at once. On the revenue side: extraordinary growth, improving gross margins, and rapid model-price compression. On the cost side: extremely large operating losses, extremely large R&D spending, large cash burn, and hundreds of billions of dollars in expected infrastructure needs. Surrounding it all: enormous financing capacity and a commercial market that is still developing quickly enough that today's economics may look materially different several years from now.

That is why reducing the IPO debate to "profitable or unprofitable" misses the harder question. The real question is whether OpenAI can convert technological scale into economic scale. The two are related. They are not identical.

OpenAI's 2025 Losses Show Why Accounting Precision Matters

The same distinction appears in OpenAI's full-year 2025 reporting. OpenAI's reported 2025 net loss was approximately $39 billion. But roughly $30 billion of that figure was associated with a non-cash structural-conversion charge. Excluding that accounting effect, the loss was closer to approximately $8 billion, against roughly $34 billion in total 2025 spending. Those figures are full-year 2025 reporting and must be kept chronologically separate from the Q1 2026 shareholder-document figures, not blended into one undated financial picture.

That does not make the company profitable. It demonstrates why headline loss figures can badly distort an AI company whose capital structure, compensation and corporate conversion create enormous accounting charges. Public-market investors will eventually receive more standardized disclosure if OpenAI completes an IPO. Until then, each financial number needs to be classified by what it actually measures.

Why OpenAI's Revenue Growth Is Not Enough by Itself

If OpenAI were a conventional software company with little incremental infrastructure cost, the revenue trajectory might answer much of the investment question. It is not. OpenAI is selling access to computational intelligence whose production requires enormous physical infrastructure. The business has software-like characteristics at the customer interface and infrastructure-like characteristics underneath it, and that hybrid structure may prove central to how the company is valued.

The customer experiences software. OpenAI has to procure compute. The customer pays per subscription, token, seat or service. OpenAI has to fund model research, training, inference and physical capacity. That means revenue can scale extraordinarily quickly without automatically producing the capital-light economics historically associated with the strongest software businesses. The IPO will test whether investors believe that changes over time.

The Price Cuts Are More Than a Competitive Move

External reporting naturally places OpenAI's GPT-5.6 price cuts within an increasingly aggressive market involving Anthropic and lower-cost Chinese models. Axios explicitly framed the reductions within broader competition even while reporting OpenAI's stated efficiency explanation.

But there is another reason the pricing matters to OpenAI's IPO economics. Falling inference prices can reveal technological improvement. If OpenAI can deliver equivalent or greater capability with fewer underlying resources, then a lower customer price can coexist with stronger internal economics. This is what makes gross-margin improvement so important. Without the margin data, price cuts could look simply like competitive pressure forcing OpenAI to sacrifice economics. With gross margin rising from 33% to 39%, a more nuanced possibility emerges: OpenAI may be lowering the market price of intelligence while simultaneously lowering its own cost of producing it. That is precisely the type of operating dynamic that could eventually support a very large public valuation. Whether it can continue at sufficient scale remains unresolved.

Anthropic Shows Why Current Revenue Multiples May Become Misleading

OpenAI is not approaching public markets alone. Anthropic has also confidentially filed IPO paperwork. Bloomberg estimated that Anthropic had reached an approximately $47 billion annualized revenue run rate by May 2026, while preserving a caveat that OpenAI and Anthropic may calculate run rate differently, making direct comparison imprecise.

Reuters has separately reported that Anthropic has pitched projected 2028 revenue of roughly $190 billion to $200 billion in valuation discussions. That future figure should not be treated as realized revenue. Its significance is methodological. AI companies and investors are increasingly discussing valuation against what these businesses might generate several years from now rather than simply what they generate today. OpenAI's eventual IPO may therefore become a referendum on forward AI economics more than a traditional valuation of current financial results.

OpenAI Is Also Reorganizing Leadership Before a Possible IPO

Financial scale is not the only transition underway. Chief Revenue Officer Denise Dresser is departing less than a year after taking the role and is being replaced by Dali Rajic, formerly an operating executive at Wiz. Longtime OpenAI executive Brad Lightcap also announced his departure to start a new venture. Importantly, Lightcap had already moved out of the chief operating officer role earlier in 2026 and into a special-projects position before announcing his August departure. Describing Lightcap as the sitting COO suddenly leaving in August would misstate the sequence.

The Financial Times and Axios have described a broader reorganization involving multiple senior departures and Greg Brockman taking on greater operating responsibility. That does not establish that OpenAI is unstable. Executive turnover is not self-proving evidence of corporate distress. The fair conclusion is that OpenAI is undergoing substantial leadership reorganization during one of the fastest periods of commercial expansion in its history and while evaluating a transition to public markets. That context belongs in the IPO story without becoming the story by itself.

What Does Profitability Even Mean for a Frontier AI Company?

This may eventually become the most important analytical question. A frontier AI company could reach accounting profitability by reducing investment. That may also weaken its future competitive position. OpenAI is operating in a market where the next generation of products may require enormous research and infrastructure commitments before their future revenue exists.

R&D can therefore appear simultaneously as an expense reducing current profit and an investment attempting to create future economic capacity. The $8.6 billion first-quarter R&D figure makes that tension visible. If OpenAI dramatically reduced that spending, its current financial statements might improve. The strategic consequences could be very different. Public investors will therefore have to decide how much current loss they are willing to tolerate in exchange for continued frontier development. That trade-off is not unique to OpenAI. The scale is.

The OpenAI IPO Could Test a New Type of Public Company

Historically, public markets have valued large technology companies after their core economic engines were relatively mature. OpenAI may reach public markets while the basic economics of artificial intelligence are still moving quickly. Model performance changes. Token pricing changes. Infrastructure projections change. Competitive positioning changes. Revenue run rates change. The cost of producing intelligence changes. Even the expected IPO timing is changing.

That means an OpenAI prospectus could eventually expose public investors to something unusual: an extremely large company whose fundamental unit economics and capital requirements are still evolving rapidly at the same time. The traditional question, "When will OpenAI become profitable?" may therefore be too shallow. A more useful set of questions would be: Can gross margins continue improving? Can prices fall while aggregate revenue rises? Can utilization grow faster than infrastructure obligations? Can R&D spending eventually become a smaller share of revenue without sacrificing technological leadership? Can OpenAI finance hundreds of billions of dollars of compute without permanently depending on capital markets? And can the company convert a $40 billion annualized commercial engine into enough durable cash generation to justify a valuation already approaching one trillion dollars?

Those questions are different versions of the same problem.

What Would Make the OpenAI IPO Thesis Stronger?

The current financial record already contains one encouraging signal: gross-margin improvement. A second would be sustained revenue growth. A third would be continued reduction in the cost of delivering intelligence. The strongest future evidence, however, would come when those improvements begin reaching the bottom of the financial structure: lower cash burn relative to revenue, greater operating leverage, R&D spending growing more slowly than commercial output without technological deterioration, infrastructure commitments producing enough monetizable capacity to justify their cost, and ultimately positive cash generation.

OpenAI has not demonstrated that final state yet. It has demonstrated enough movement in the underlying economics that dismissing the possibility would be premature.

What Would Make the OpenAI IPO Thesis Weaker?

The opposite evidence would matter just as much. Infrastructure spending rising faster than revenue would undermine the core efficiency argument. Gross-margin improvement stalling or reversing, persistent large cash burn despite enormous scale, price competition outpacing cost improvements, R&D requirements remaining structurally larger than the business can economically support, or a repeated need for extraordinary capital raises simply to maintain frontier position would each represent a meaningful challenge to the IPO thesis.

None of those outcomes is sealed as OpenAI's future. They are the economic risks embedded in the current architecture.

The Real Economics of an OpenAI IPO

OpenAI has already accomplished something extraordinary commercially. A company that ended 2025 at more than a $20 billion annualized revenue run rate is now running at roughly $40 billion. Its gross margin improved six percentage points year over year in the first quarter. It has shown enough serving efficiency to reduce the price of important GPT-5.6 models dramatically. It raised $122 billion at an $852 billion valuation. And it has submitted the confidential S-1 that preserves the option to enter public markets.

At the same time, the company burned $3.7 billion of cash in one quarter. Its operating loss was approximately $9.3 billion. Its R&D spending exceeded its quarterly revenue. Its planned cloud-computing spend has risen toward $750 billion through 2030. And its own public statement still says the timing of an IPO remains undecided.

That is neither a celebration story nor a collapse story. It is an economics story. OpenAI is attempting to prove that intelligence can become simultaneously cheaper per unit, more capable, more widely consumed and more profitable even while the physical cost of creating the next generation of that intelligence reaches infrastructure scale. If it succeeds, the company may justify financial assumptions that look extreme by historical software standards. If it does not, extraordinary revenue growth alone will not resolve the economics.

That is the real question behind an OpenAI IPO: can OpenAI turn rapidly improving economics per unit of intelligence into durable company-level profitability before the cost of building the next generation of intelligence outruns even its extraordinary revenue growth? In August 2026, the evidence finally allows the question to be asked seriously. It does not yet provide the answer.

Fact Summary

Is OpenAI generating $40 billion in annual revenue? Not in the accounting sense that phrase normally implies. Bloomberg reported that OpenAI's annualized revenue run rate has reached roughly $40 billion. A run rate annualizes a recent revenue pace and is not the same as audited trailing-twelve-month revenue.

How quickly has OpenAI's revenue pace grown? OpenAI ended 2025 at more than a $20 billion annualized pace, was generating roughly $2 billion per month by March 31, 2026, and had reached approximately a $40 billion annualized run rate by August 2026.

How much revenue did OpenAI actually generate in Q1 2026? Approximately $5.7 billion, according to The Information citing shareholder documents.

How much cash did OpenAI burn in Q1 2026? Approximately $3.7 billion. That should not be confused with its approximately $9.3 billion operating loss or more than $21.3 billion net loss, which each measure different things.

Why was OpenAI's net loss so much larger than its cash burn? The more than $21.3 billion net loss included an approximately $12.4 billion non-cash accounting charge associated with revaluation of convertible interests and warrant liabilities. The operating loss also included non-cash expenses such as stock-based compensation.

Are OpenAI's underlying economics improving at all? Yes, at the gross-margin level. Q1 2026 gross margin increased to approximately 39% from 33% a year earlier, indicating that the economics of delivering its products improved year over year.

How much did OpenAI spend on R&D in Q1 2026? Approximately $8.6 billion, more than the quarter's $5.7 billion in revenue. This explains a substantial share of why operating losses remain large despite improving gross margins.

How much capital has OpenAI raised? OpenAI announced on March 31, 2026 that it closed its latest round with $122 billion in committed capital at an $852 billion post-money valuation.

Has OpenAI officially filed for an IPO? It has confidentially submitted a draft S-1 to the U.S. Securities and Exchange Commission. OpenAI confirmed the submission on June 8, 2026.

When is the OpenAI IPO? There is no confirmed date. Reuters reported that an offering could come as early as September 2026, while OpenAI itself said timing remains undecided and going public may be a while. Newer reporting from the Financial Times suggests the expected timetable may be moving toward 2027.

Could OpenAI be valued at $1 trillion at IPO? Reuters reported an externally sourced possibility of a valuation as high as $1 trillion. OpenAI has not announced a $1 trillion IPO valuation. Its most recent confirmed private valuation was $852 billion.

What would a $1 trillion valuation equal against the current $40 billion run rate? Approximately 25 times current run-rate revenue. That is mathematically correct but is not necessarily the revenue period investors would use to value an eventual IPO.

How much does OpenAI plan to spend on compute? Greg Brockman said OpenAI expected approximately $50 billion in computing expenditure during 2026. Earlier reporting placed cumulative spending around $600 billion through 2030, while The Wall Street Journal subsequently reported that projected spending had risen to roughly $750 billion through 2030.

Did OpenAI cut GPT-5.6 prices? Yes. On July 30, 2026, OpenAI announced an approximately 80% price reduction for GPT-5.6 Luna and approximately 20% for Terra, citing efficiency improvements. GPT-5.6 Sol pricing was unchanged.

Does lower AI pricing mean OpenAI's gross margins must fall? Not necessarily. Lower customer pricing and lower underlying serving costs can occur simultaneously. Q1 gross-margin improvement provides evidence that delivery-level economics were improving before the July pricing changes, but the long-term effect of those price cuts on company profitability remains unresolved.

Why is OpenAI still losing money if gross margin is improving? Company-wide expenses remain enormous, particularly R&D, infrastructure, compensation and expansion spending. Q1 R&D spending alone exceeded the quarter's entire revenue.

What is the central OpenAI IPO question? Whether OpenAI can convert improving unit economics and extraordinary revenue growth into durable company-level profitability before frontier-model and infrastructure costs consume those gains.