Reported $1.2T round valuation
~41% above $852B
Per Financial Times, Sep 15, 2026; OpenAI confidentially filed its S-1 earlier in 2026
OpenAI annualized revenue (Aug 2026)
~$40B
Reported 20% MoM surge tied to GPT-5.6 launch; Yahoo Finance / FT reproduction, Sep 15, 2026
Implied $1.2T price/sales
30×
Versus Anthropic at ~20.5× ($965B / $47B ARR); Yahoo / FT, Sep 15, 2026
2025 net loss
$38.5B
Vs. $5.1B in 2024; TradingKey citing OpenAI disclosures, Jun 16, 2026
2025 payments to Microsoft
$17.2B
Largest single cost line; TradingKey, Jun 16, 2026
Oracle Stargate cloud commitment
$300B / 5 yrs
Starts 2027; WSJ, Sep 10, 2025
A 41% Step-Up on the Same Revenue Base
Six months after OpenAI closed $122 billion at an $852 billion post-money valuation, the company is in early discussions for another private round at approximately $1.2 trillion, according to Financial Times reporting cited by Yahoo Finance on September 15, 2026. CEO Sam Altman confirmed three days earlier that OpenAI will not pursue a public listing in 2026, citing AI-safety work that still needs to be done — a sequence that explains why a fresh private ceiling is on the table before the S-1 ever goes effective.
| Date | Round size | Post-money | Lead backers | ARR at round |
|---|---|---|---|---|
| Feb 2024 | ~$6.6B | ~$86B | Thrive Capital, MSFT, others | ~$2B |
| Mar 2025 | ~$40B | $300B | SoftBank, MGX, Thrive | ~$10B |
| Feb 27, 2026 | $110B | $730B pre | SoftBank $30B, NVIDIA $30B, Amazon $50B | ~$20B |
| Mar 31, 2026 | $122B | $852B | SoftBank co-led w/ a16z, TPG, MGX, T. Rowe Price | ~$24B |
| Sep 2026 (reported) | TBD | ~$1.2T | Discussions ongoing | ~$40B |
The Math Has Already Broken the Public-Market Comparables
At $1.2 trillion on ~$40 billion in annualized revenue, OpenAI would carry an implied 30× price/sales — roughly 80% richer than Anthropic's 20.5× at its $965B post-money mark on $47B ARR. The premium widens further against the public-market AI complex.
| Company | Mkt cap | TTM revenue | P/S (TTM) | EV/EBITDA (TTM) |
|---|---|---|---|---|
| NVIDIA | $5.14T | $302.97B | 16.8× | 22.4× |
| Microsoft | $3.72T | $331.84B | 11.3× | 17.6× |
| Amazon | $2.68T | $775.68B | 3.5× | 11.3× |
| Broadcom | $1.62T | $89.10B | 18.5× | 33.7× |
| Micron | $1.04T | $90.27B | 11.6× | 15.8× |
| AMD | $819.5B | $41.30B | 19.5× | 77.8× |
| Oracle | $416.9B | $71.78B | 5.8× | 14.8× |
| Constellation Energy | $93.3B | $31.27B | 3.0× | 15.2× |
| Equinix | $99.4B | $9.90B | 10.0× | 27.7× |
| CoreWeave | $44.4B | $7.59B | 6.0× | 24.1× |
| OpenAI (private, implied) | ~$1.2T | ~$40B | ~30× | n/m (negative EBITDA) |
Why a Private Round Now — Capital, Capacity, and the IPO Off-Ramp
OpenAI's burn is no longer a rounding error. The 2025 disclosures show a $38.5 billion net loss — nearly eight times 2024's $5.1 billion — driven by $19.2B in R&D, $7.5B in cost of revenue, and $5.7B in sales and marketing. TradingKey, citing OpenAI's filings, June 16, 2026, notes that payments to Microsoft alone totaled $17.2B in 2025 across cloud computing, model-training credits, and joint go-to-market spend.
That capital intensity is the operational reason behind the timing. The Stargate build-out — anchored by a $300B, five-year cloud-compute commitment to Oracle starting in 2027 (WSJ, Sep 10, 2025) — does not pay for itself on $40B of ARR. A private round at $1.2T lets existing backers (SoftBank, Thrive Capital, the a16z/T. Rowe Price/BlackRock-affiliated cohort from the March round) expand their equity footprint without forcing a 2026 listing into a window where the AI safety narrative is suppressing investor appetite.
OpenAI burn trajectory, 2024 → 2025
Net loss widened nearly 8×; operating losses more than doubled
Unit: $B
2024 net loss
Per OpenAI disclosures, Jun 2026
5.1
2025 net loss
Per OpenAI disclosures, Jun 2026
38.5
2025 R&D
+146% YoY
19.2
2025 cost of revenue
+178% YoY
7.5
2025 S&M
+418% YoY
5.7
2025 payments to Microsoft
Single-largest counterparty
17.2
The Supply Chain Map: Who Gets Paid Before the S-1
The transmission mechanism from a $1.2T OpenAI round to listed earnings is mechanical: every additional dollar of equity OpenAI raises is, in practice, pre-committed to the Stargate compute stack — and most of that spend is concentrated in a handful of public suppliers.
- NVIDIA takes the largest cut — already a $30B OpenAI equity backer from the February round and the de facto chip supplier for both the Azure and Stargate training clusters; trailing twelve-month revenue of $303B and 75% gross margin (NVIDIA Q2 FY26 10-Q, ended Jul 31, 2026) implies OpenAI's incremental GPU spend falls straight to NVIDIA's top line.
- Oracle converts $300B of Stargate cloud commitments into a multi-year backlog — already visible in Oracle's Q1 FY27 results, ended Aug 31, 2026, which show $71.78B TTM revenue at 5.8× price/sales and a market cap of $416.9B that bakes in only part of the Stargate ramp.
- Broadcom supplies the custom-networking and inference-ASIC silicon behind both Stargate and Anthropic's rival build — its $89.1B TTM revenue and 33.7× EV/EBITDA (Broadcom Q1 FY26 10-Q, ended Aug 3, 2026) price in continued ASIC share gains.
- Micron and the broader HBM cohort are the bottleneck. Micron's Q3 FY26 results, ended May 31, 2026, show 72.6% gross margin and $68.2B TTM EBITDA on $90.27B in revenue — the HBM3E/HBM4 allocation to OpenAI is the swing factor for 2027 gross margin.
- AMD is the alternate-supply lever. AMD's Q2 2026 results, ended Jun 28, 2026, show 50.1% YoY quarterly revenue growth on $41.3B TTM; an OpenAI MI355/MI400 design win is the open catalyst.
- CoreWeave, Equinix, and Constellation Energy sit downstream — CoreWeave is a direct GPU-cloud counterparty (TTM revenue $7.59B, EV/EBITDA 24.1× per Q2 2026 10-Q); Equinix and Constellation provide the datacenter shells and baseload power that turn OpenAI's checkbook into deployed compute.
The key non-obvious read: the $1.2T round tightens the supply chain before OpenAI ever files an S-1. Each new dollar of OpenAI equity locks in years of GPU, HBM, ASIC, datacenter, and power commitments — meaning the supply-chain revenue tail is more visible than the OpenAI revenue line itself, which is precisely why listed AI infra trades at 11–20× sales while OpenAI commands 30×.
Short-Term Catalysts vs Long-Term Risks
Days-to-quarters: the $1.2T round's confirmation, terms sheet, and lead-investor lineup are the next market-moving events. The OpenAI $122B round's investor roster (OpenAI press release, Mar 31, 2026) included SoftBank (co-lead), a16z, D. E. Shaw Ventures, MGX, TPG, T. Rowe Price, BlackRock affiliates, Blackstone, Coatue, D1, Dragoneer, Fidelity, Insight, Paragon, Sands, Sequoia, Temasek, Thrive Capital, and UC Investments — a syndicate broad enough that a $1.2T reset will move public wealth managers' private-allocation budgets materially.
One-to-three years: the long-term risk is structural, not financial. Anthropic's late-May 2026 round at $965B on $47B ARR and its reported October Nasdaq filing at a $2T target mean the AI private-premium has a credible competing benchmark. If Anthropic prices into a public-market correction first, OpenAI's $1.2T mark becomes the anchor — and a 41% step-up in six months on flat-to-down revenue is the kind of trajectory that draws regulatory scrutiny on AI infrastructure concentration (Constellation Energy, Microsoft Three Mile Island PPA, Oracle Stargate land/permit) before the S-1 ever reaches the SEC.
| Horizon | Catalyst / risk | Most exposed |
|---|---|---|
| Days–quarters | Round confirmation; lead-investor mix; GPT-5.6 ARR continuation | NVIDIA, Oracle, AMD, CoreWeave |
| Days–quarters | Stargate site buildout (4.5 GW capacity) | Constellation Energy, Equinix, Microsoft |
| 1–3 years | Anthropic IPO at ~$2T resets the comparable | Microsoft, Oracle, Amazon |
| 1–3 years | 2025 burn rate ($34B) holding vs $40B ARR | NVIDIA, Micron, Broadcom |
| 1–3 years | Regulatory scrutiny on AI infra concentration | Sector-wide multiple compression |
Investable exposure to the OpenAI private round
- Already a $30B OpenAI equity backer and the de facto GPU supplier for both Azure and Stargate; the new round pre-funds multi-quarter Blackwell/Blackwell-Ultra orders.
- TTM revenue $303B and 75% gross margin (Q2 FY26 10-Q) mean incremental OpenAI dollars fall straight to operating profit, with the Stargate ramp most visible in FY27 datacenter revenue.
- Short-term (days–quarters): round confirmation is a sentiment catalyst; long-term (1–3 years): the 30% AI compute share makes NVIDIA the index proxy for the entire OpenAI trade.
- The $300B Stargate cloud commitment converts directly into RPO; Oracle's $416.9B market cap on $71.8B TTM revenue (Q1 FY27 10-Q) prices in only a fraction of the ramp.
- Stargate's 4.5 GW capacity build gives Oracle a multi-year competitive moat against AWS and Azure in dedicated AI hosting.
- Short-term: Stargate site announcements are the near-term catalyst; long-term: the question is whether OCI revenue inflects above 25% growth to justify the 5.8× P/S.
- OpenAI's $17.2B of payments to Microsoft in 2025 makes the partnership a tailwind — but also the largest single counterparty concentration risk in Microsoft's cost base.
- Azure's AI workloads now anchor a $3.72T market cap on $331.84B TTM revenue (Q4 FY26 10-K), so the question is whether OpenAI diversifies to Oracle/Amazon over time.
- Short-term: any softening of the Microsoft-OpenAI partnership narrative is a sentiment headwind; long-term: the $1.2T private ceiling lifts the eventual IPO back-solve value of Microsoft's stake.
- An OpenAI MI355/MI400 design win is the binary catalyst not yet confirmed; the new round creates the capital OpenAI needs to diversify its silicon base.
- AMD's 50.1% YoY quarterly revenue growth (Q2 2026 10-Q) on $41.3B TTM is already priced in at 19.5× P/S and 77.8× EV/EBITDA — leaving little room for disappointment.
- Short-term (days–quarters): watch for an OpenAI-AMD order announcement alongside the $1.2T round; long-term: AMD's 2027+ share depends on execution, not valuation.
- Custom-networking and inference-ASIC exposure ties Broadcom to both OpenAI's Stargate build and Anthropic's competing compute stack — a rare cross-bet in the AI infra complex.
- TTM revenue $89.1B at 67.7% gross margin (Q1 FY26 10-Q) means every dollar of OpenAI Stargate capex is a high-margin revenue event.
- Short-term: the 18.5× P/S already reflects strong expectations; long-term (1–3 years): the ASIC share gain against merchant GPUs is the structural upside.
- HBM3E/HBM4 supply allocation to OpenAI is the swing factor for 2027 gross margin — Micron's 72.6% TTM GM (Q3 FY26 10-Q) already reflects HBM tightness, but the $1.2T round adds visibility.
- Net debt/EBITDA of -0.27× and operating margin of 80.4% make Micron one of the most leveraged plays on AI memory demand.
- Short-term (days–quarters): round confirmation tightens 2026 HBM allocation; long-term (1–3 years): HBM4 ramps and any pricing reset are the key milestones.
- CoreWeave is a direct GPU-cloud counterparty to both OpenAI and Microsoft; its $7.59B TTM revenue (Q2 2026 10-Q) prices in continued capacity expansion at 6.0× P/S.
- Debt/equity of 10.27× means the equity is a leveraged option on continued GPU demand — a $1.2T OpenAI round reduces perceived customer-concentration risk.
- Short-term: any new OpenAI-anchored capacity contract is the catalyst; long-term (1–3 years): the question is refinancing and unit economics as capex normalizes.
- Stargate's 4.5 GW capacity build and Microsoft's Three Mile Island PPA framework make Constellation the baseload-power beneficiary of the entire AI compute build-out.
- TTM revenue $31.27B at EV/EBITDA 15.2× (Q2 2026 10-Q) reflects a regulated utility profile — the upside is volume, not multiple expansion.
- Short-term (days–quarters): new hyperscaler PPAs are the catalyst; long-term (1–3 years): permitting and grid interconnection are the binding constraints, not demand.
