OpenAI Filed. Their Financials Should Worry You.

OpenAI filed its confidential S-1 with the SEC on June 8, targeting a Q4 listing at an $852 billion to $1 trillion valuation with Goldman Sachs and Morgan Stanley leading. The April speculation was speculation. The actual filing is the receipt. The receipt says OpenAI lost $1.22 for every dollar of revenue in Q1, totaling roughly $6.95 billion in quarterly losses on a $25 billion annualized revenue base. That is a non-GAAP operating margin of negative 122 percent, and it is the number every enterprise buyer with an OpenAI contract needs printed on the renewal memo.

This filing landed exactly one week after Anthropic filed at $965 billion. Your two primary AI vendors are now simultaneously chasing Wall Street, and their financial profiles are opposite.

That divergence is the new variable in your 2027 contract calendar.

Quick Verdict

Question	The Answer
What got filed?	OpenAI confidential S-1 with the SEC on June 8, 2026
Underwriters?	Goldman Sachs and Morgan Stanley leading
Target window?	Q4 2026 public listing, contingent on SEC review
Target valuation?	$852 billion to $1 trillion
Non-GAAP operating margin?	Negative 122 percent in Q1 2026
Quarterly loss?	Roughly $6.95 billion
Annualized revenue?	$25 billion
The Anthropic contrast	Filed at $965B one week earlier, tracking first-ever operating profit
The adoption signal	Ramp data: more businesses paid Anthropic than OpenAI for the first time in May
Combined IPO ask (OpenAI + Anthropic + SpaceX)	More than $3.6 trillion against a 2025 US IPO market that raised $45B total
The buyer takeaway	OpenAI is the high-burn vendor; Anthropic is the structural counterweight

What the S-1 Actually Discloses, and What It Won’t Until Later

A confidential S-1 is not a public S-1. The SEC review happens first. The full risk factors, the customer concentration disclosure, the audited segment financials, the gross margin breakdown by product line, none of it is public yet. What is public is the fact of the filing, the target window, the underwriters, and the financial picture analysts have already stitched together from earlier disclosures and leaked operating data.

The headline number is the one Edward Zitron published from internal financial documents on Q1 2026 OpenAI performance. Non-GAAP operating margin: negative 122 percent. Revenue: $5.7 billion. Loss: roughly $6.95 billion. Annualize the revenue and you land at $25 billion, which matches the Sacra estimate for OpenAI’s annualized run rate in February 2026. ChatGPT consumer growth stalled in the quarter. The enterprise mix grew. The cost base grew faster than either.

The $852 billion valuation comes from OpenAI’s March 2026 secondary, and the IPO targeting range tops out near $1 trillion. The math is the part the buyer needs to internalize. A vendor losing $1.22 on every dollar of revenue cannot maintain that posture under public-market accountability. The post-IPO operating plan has to close the gap. The gap closes through some combination of compute cost reduction, pricing pressure on enterprise accounts, and revenue mix shift. Compute costs do not compress quickly. That leaves pricing and mix as the levers Wall Street will pull.

The buyer pays the difference. That is the only structural conclusion the math supports.

The -122 Percent Margin Is a Vendor Stability Signal, Not Just a Pricing Signal

Most CIO coverage of the filing has framed the operating loss as a pricing warning. That framing is incomplete. The number is also a vendor stability signal, and the stability signal is the one that should change the renewal conversation in any enterprise procurement office.

Software vendors that run at negative 122 percent operating margin do not stay private forever, and they do not stay solvent through public-market scrutiny without a credible path to break-even inside 36 months. OpenAI’s own internal forecast, published earlier this year, models continued losses through 2028 and a first profitable year not before 2029. The IPO has to bridge the funding gap between today and that projected break-even. The bridge runs through public-market capital that gets priced against the same operating loss the buyer sees in the S-1.

The buyer-side implication is not that OpenAI fails. The probability of OpenAI failing is low. The strategic shareholder positioning (Microsoft, Nvidia, the SoftBank stake) plus the consumer brand make solvency essentially underwritten. The implication is that the operating constraint will shape every renewal cycle, every model deprecation timeline, every standard MSA revision, and every customer support resource allocation for the next four to eight quarters.

What does that look like operationally? Faster compute cost pass-through into per-token pricing. Tighter usage caps on enterprise tiers. Compressed timelines on model deprecation. Reduced flexibility in custom terms. The same operating playbook I covered in Anthropic Is Going Public. Lock In Your Terms Now. applies to OpenAI with one critical difference. Anthropic is tracking toward its first operating profit. OpenAI is losing money at a pace analysts call unprecedented for a public-company candidate. The compression on the OpenAI side will run harder.

That is the vendor stability signal. Solvency is essentially underwritten. Contract posture and pricing under public-market pressure are where the buyer actually feels it.

The Anthropic Contrast Is the New Procurement Variable

Here is the part most buyer-side teams have not internalized yet. Your two primary AI vendors filed for IPO inside eight days of each other, and they look nothing alike on the financials.

Anthropic filed at $965 billion on June 1. The Q2 profit announcement that came earlier this year was the structural inflection I wrote up in Anthropic Just Turned Profitable. Now Negotiate. Anthropic is tracking toward its first-ever operating profit in 2026. The gross margin estimate sits in the low 40s, with a clear infrastructure path through the TPU and Trainium investments to push that toward 60 percent inside 18 months. The Series H closed in May. The S-1 followed. The operating posture going into the listing is closer to a high-growth profitable software company than to a venture-funded loss leader.

OpenAI filed at $852 billion on June 8. The same revenue base (roughly $25 billion annualized). A negative 122 percent operating margin. A 2029 projected first-profit year. A consumer business with stalled growth. An enterprise business growing fast, but inside a cost structure that bleeds harder than the revenue grows.

The same valuation. Two completely different operating profiles.

The procurement implication is that the two contracts behave differently under public-market pressure. Anthropic post-IPO has room to hold the line on pricing because the margin trajectory already runs toward expansion. OpenAI post-IPO has to compress the loss faster than the revenue base supports, which means pricing pressure, deprecation pressure, and MSA tightening will run harder and sooner on the OpenAI side.

The Ramp data published in May was the early signal. For the first time, more businesses paid Anthropic than OpenAI, 34.4 percent versus 32.3 percent across more than 50,000 companies. Twelve months earlier the gap was 23 points in OpenAI’s favor. The trajectory matters more than the crossover itself. I covered the broader divergence in ChatGPT Built the Market. Claude Is Winning the Enterprise. The S-1 filings on both sides this month sharpen the same picture into a procurement question.

If your AI vendor stack runs primarily on OpenAI, the next 18 months are the operating window where the loss compression playbook hits the buyer hardest. If it runs primarily on Anthropic, the pressure is real but the structural posture is friendlier.

If it runs on both with no portability between them, you pay both compression cycles in sequence.

What is a vendor stability signal in an AI contract context?

A vendor stability signal is any disclosed financial, operating, or strategic data point that predicts how a software vendor’s contract posture will shift under capital-market or competitive pressure inside the next four to eight quarters. Strong signals include operating margin trajectory, customer concentration, capital structure, and the timing of any public-market accountability event like an IPO or major refinancing. The S-1 filing is the single highest-quality vendor stability signal available to an enterprise buyer because it bundles the financials, the risk factors, the underwriting structure, and the operating plan into one disclosure the SEC has to vet. Buyers use vendor stability signals to decide which contracts to renegotiate early, which to add escape hatches to, and which to migrate workloads out of before the vendor’s posture forces the migration on a less favorable timeline.

The Combined IPO Pipeline Math Is the Macro Context Nobody Should Skip

The third number worth printing on the same memo is the combined IPO pipeline. OpenAI, Anthropic, and SpaceX together represent more than $3.6 trillion of equity value coming to public markets inside a 12 month window. The three listings together could demand north of $200 billion from public-market investors. The entire US IPO market raised $45 billion in all of 2025.

Bloomberg’s reporting puts the pipeline at 4x last year’s full-year IPO proceeds in a single concentrated wave. Goldman Sachs analysts had already projected 2026 IPO proceeds at roughly $160 billion, a quadrupling of 2025, before this wave fully materialized.

The macro implication for the enterprise buyer is the absorption math. Public markets can absorb a $200 billion ask in a single year, but only with help from the same pricing dynamics that will hit the buyer side. Stronger initial pricing on the IPOs, more aggressive post-listing margin expansion, faster operationalization of the public-company playbook. The buyer pays a portion of that absorption cost directly through contract repricing, indirectly through faster deprecation cycles, and structurally through the kind of multi-vendor risk concentration I covered in Your AI Stack Has an Expiration Date.

The IPO pipeline math is also a reminder that this is not idiosyncratic to OpenAI. The whole frontier AI vendor stack is converging on public-market accountability inside the same window. The procurement function inside any large enterprise has to treat that convergence as a category-level event. A vendor-level renewal lens misses the structural piece entirely.

Why the Anthropic and OpenAI Profiles Diverge Operationally

The reason the two financial profiles look so different is structural, not cyclical. Worth understanding before you write the next contract.

Anthropic optimized the inference cost side of the equation hard, partly through the Google TPU relationship and partly through aggressive model quantization on the Claude SKUs. The gross margin estimate in the low 40s today has a credible path to 60 percent by 2027 through infrastructure efficiency that is largely already committed capex. The enterprise mix is heavy, the consumer mix is light, and the consumer business is not bleeding the way OpenAI’s free ChatGPT tier bleeds.

OpenAI carries the cost structure of the consumer brand. Free ChatGPT users do not pay. They use compute. Compute costs money. The cost-of-revenue line is fatter on a per-dollar-of-revenue basis than on the Anthropic side. The enterprise business helps but does not close the gap because the enterprise business runs on the same underlying model serving infrastructure that the consumer business loads up.

Compute costs do not compress quickly. The data center capex cycle is multi-year. The chip supply is constrained. The optimization runway exists but does not run in two quarters. Which means the post-IPO pricing pressure on enterprise accounts is structural, not cyclical. The buyer who treats it as a temporary phase pays the difference for the duration.

The same compute-cost dynamic shows up on the agent side. I made the same call from a different angle in The AI Agent Runtime Cost Crisis when the early agent runtime billing data started coming in. The structural shape is the same. Compute is the bottleneck. The vendor that solves the bottleneck cheapest gets the margin. The vendor that hasn’t solved it pays the cost or passes it through.

OpenAI hasn’t solved it cheap enough yet. The S-1 is the receipt for that.

The Anti-Hype Read

Three honest cautions before this turns into a board memo.

The filing might not convert to a listing on the stated timeline. Confidential S-1s sit through SEC review for variable periods. The September to November window OpenAI is targeting is a target, not a guarantee. Market conditions in Q4 could push the print into early 2027. Plan against the stated window. Do not bet the procurement strategy on October as a hard date.

The negative 122 percent margin number is leaked operating data, not the audited S-1 disclosure. The actual figure could land slightly higher or slightly lower depending on how OpenAI presents non-GAAP versus GAAP adjustments. The directional read holds regardless. A vendor losing meaningfully more than it earns at $25 billion of annualized revenue is the structural picture, and that picture does not invert before listing.

The Anthropic financial profile is genuinely better, but Anthropic is not immune to the public-market playbook either. The Forrester forced migration cycle read I covered yesterday applies to Anthropic with real teeth. The contrast between the two vendors is real. The conclusion that Anthropic is the safe vendor is not the right read. The conclusion that Anthropic is the lower-pressure vendor of the two for the next 18 months is.

None of those cautions changes the procurement call. The OpenAI contract gets harder to negotiate this year than next year. The Anthropic contract gets harder to negotiate next year than this year. Both windows are open this quarter. Both will be closed inside the same 12 months.

Three Moves This Week

Sized for any CIO, CFO, or AI program owner running enterprise contracts on either or both vendors. Doable inside seven days.

1. Split the contract inventory by vendor and project the loss-compression timeline against each one. Tag every active OpenAI contract with its next renewal date and tag the workload running on it with portability status (portable, partially portable, locked). Do the same for Anthropic. The OpenAI contracts in renewal between Q4 2026 and Q2 2027 are the highest-risk procurement cycles in the portfolio because they hit the post-IPO compression window directly. The Anthropic contracts running through 2028 are the lower-pressure cycles, but the model-version guarantee gap I described in yesterday’s piece on Anthropic applies to them equally. Inventory both.

2. Open the early renegotiation conversation on every OpenAI contract with material spend. The named-account team will take the meeting because pre-IPO logo retention still drives the comp plan this quarter. Push for the same four clauses that apply to Anthropic: per-token rate lock with ceiling, model version guarantee, tightened data governance with audit rights, and credit portability. The OpenAI standard MSA is materially less buyer-friendly than the Anthropic one going into both IPOs. The pre-listing negotiation window is the only window where that gap narrows. After Q4, OpenAI’s public-company legal posture will widen the gap back out.

3. Build the portability path between the two vendors before the listings close. Stand up working API integration paths to both Claude and GPT for any production workflow that depends on either. The cost of running both in parallel is low this quarter. The cost of being locked to either one during the post-IPO compression cycle is high. The same logic I made in The OpenAI Killed Sora Vendor Risk applies here at the strategic level. Single-vendor exposure on either frontier model provider is a board-level risk question now. The dual-IPO event sharpens the question by a meaningful factor.

My Read

The OpenAI S-1 confirms what the April speculation could only model. The non-GAAP operating margin is negative 122 percent. The quarterly loss is $6.95 billion. The valuation target is $852 billion to $1 trillion. The Anthropic filing one week earlier landed at $965 billion against a first-operating-profit trajectory. Two valuations in the same zip code. Two operating profiles that look nothing alike.

The procurement call is not whether to use OpenAI or Anthropic. The call is whether to write the contracts on both sides with portability between them before the listings close. The OpenAI cost-compression playbook will hit harder and faster post-IPO than the Anthropic one will. The Anthropic contract gives you a structural counterweight if you write the clauses correctly this quarter. The OpenAI contract gives you a working primary vendor if you lock the terms before the public-market dynamics take over.

The combined IPO pipeline math is the macro that ties both calls together. More than $3.6 trillion of equity value coming to market inside 12 months against a US IPO market that raised $45 billion last year. The absorption mechanics force pricing pressure on every buyer in the category, on both vendors, simultaneously. The procurement function that treats this as a single-vendor renewal cycle misses the category-level event.

Pull the contract inventory this week. Project the renewal calendar against the listing windows. Open the OpenAI conversations first because the loss-compression curve is steeper. Open the Anthropic conversations next because the favorable-terms window closes second. Build the portability path between the two before either listing closes.

The CFO question in 2027 is not why you stayed on OpenAI or why you moved to Anthropic. The question is why you did not write portability into the contracts when both vendors were still pre-IPO. The honest answer right now is that most procurement teams are still treating this as a vendor-level conversation when the filings have already made it a category-level one.

Two filings. Eight days apart. One financial profile that worries you. One that worries you less. Both worth contracts that survive the dual compression cycle. Write the terms while the window is still open.

---

Related Reading:

• Anthropic Is Going Public. Lock In Your Terms Now.
• Anthropic Goes Public. Lock In Your Contracts Now.
• Anthropic Just Turned Profitable. Now Negotiate.
• ChatGPT Built the Market. Claude Is Winning the Enterprise.
• Your AI Stack Has an Expiration Date
• The AI Agent Runtime Cost Crisis
• The OpenAI Killed Sora Vendor Risk