OpenAI
Hybrid SaaS: consumer subscriptions (ChatGPT Plus/Pro/Team) + usage-based API + enterprise contracts; capped-profit structure under a nonprofit parent (converting to for-profit/PBC)
Several early marks (2023 $29B, 2024 $80B, 2025 $500B) are employee tender/secondary sales, not priced primary rounds; tagged accordingly. 2019 Microsoft ~$20B is an implied mark, never a formally disclosed post-money — omitted.
The thesis on this name
State of AI Compute
OpenAI is the frontier AI lab behind ChatGPT and the GPT/o-series models, monetizing via consumer subscriptions, a developer API, and enterprise deals while spending tens of billions on compute.
State of Frontier AI
Revenue-scale leader (~$20B ARR, $2B/mo, enterprise >40% of mix and approaching consumer parity) and the broadest consumer distribution in AI. The 'high potential' is real reach; the bucket is NOT durable_compounder because the unit economics don't yet work — ~33% gross margin and a -122% non-GAAP operating margin in Q1'26 mean every revenue dollar still loses money. Undervalued only IF margins inflect; a value trap if they don't.
Earnings, margins, COGS & capex
Hyper-growth top line ($3.7B FY24 → $13.1B FY25, ~$25B ARR by early 2026) against enormous losses: ~$20.9B operating loss and a headline $38.5B net loss in 2025 (the bulk a ~$41.6B non-cash charge from the nonprofit→for-profit conversion; adjusted net loss ~$8B). R&D alone (~$19B, much paid to Microsoft) exceeded total revenue. Confidential S-1 filed Jun 8 2026; Q4-2026 listing targeted but not locked.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~94¢ is cost of goods and ~0¢ operating expense, leaving ~6¢ of operating profit (~39¢ net).
Revenue trend
Margins
opaque; compute-driven
loss widening in absolute $
distorted by conversion charge
COGS structure
Cost of revenue ~$6.0B FY2025 (chiefly inference/training compute); much of total spend routed to Microsoft (~$17.2B in 2025 for Azure + R&D services).
Capex
Capex-like compute commitments are off-the-charts: Stargate ~$500B over 4yrs (SoftBank/Oracle/MGX/OpenAI); Microsoft $250B incremental Azure (2025-30); Oracle ~$300B (2027-31); AWS $38B (7yr); Nvidia up to $100B invest + 10GW; AMD ~6GW/~$90B; Broadcom custom silicon.
Latest earnings
n/a — no consensus; private company
Company plan: continued steep losses through ~2028, cash-flow positive ~2030; ~$27B cash burn projected 2026. IPO range floated $730B-$1T; Q4-2026 listing 'may be a while.'
- FY2025 revenue
- $13.1B
- FY2025 operating loss
- ~$20.9B
- FY2025 net loss (reported)
- ~$38.5B (incl. ~$41.6B non-cash conversion charge)
- ARR (Feb 2026)
- ~$25B
- Weekly active users
- ~800M-900M (press est.)
- Last primary valuation
- $852B (Mar 31 2026)
Growth drivers
- ChatGPT consumer subscriptions (~65% of revenue) — Plus/Pro/Team/Enterprise seats
- API / platform usage (~25%) — developers and agentic products built on GPT/o-series
- Enterprise contracts (now >40% of revenue, approaching consumer parity by end-2026)
- New surfaces: agents/operator, coding (Codex), Sora video, enterprise data connectors
- Partnerships/licensing (~10%) incl. Microsoft and distribution deals
Bull & bear
OpenAI is the AI-platform standard-bearer scaling revenue faster than any company ever, with a compute moat and an IPO that converts hype into permanent capital to reach 2030 profitability.
- ~$25B ARR and ~250% growth — the steepest revenue ramp in software history
- Consumer default (ChatGPT ~800M+ WAU) + fast-growing enterprise (>40% of revenue)
- Locked-in, multi-hundred-billion compute supply rivals can't replicate near-term
- Marquee strategic backers (Microsoft, Nvidia, Amazon, SoftBank) and $122B fresh capital
- IPO at $730B-$1T provides war-chest to outspend competitors through the build phase
A cash furnace burning ~$27B/yr with eroding share, supplier-concentration and circular-financing risk, and a valuation that prices flawless execution through 2030.
- ~$21B operating loss on $13B revenue; not cash-flow positive until ~2030 — years of dilution/capital need
- Compute obligations (Azure $250B, Oracle $300B, AWS $38B) dwarf revenue and assume sustained capital access
- Losing ground: consumer share 76.5%→54.7%; Anthropic leads enterprise API and wins most head-to-heads
- Nvidia-invests-then-OpenAI-buys-GPUs circularity invites scrutiny; AI-bubble unwind would be brutal
- $852B (and $730B-$1T IPO) is ~35x forward revenue — extreme multiple with negative margins and price competition
What it is worth
Last-round / secondary marks + reverse-DCF sanity on the implied multiple. Anchor: $852B post-money (Mar 2026); ~35x forward revenue (~$25B ARR). No public comps trade at this profile with negative margins.
~$300-500B
reverts toward the Oct-2025 secondary mark if growth decelerates, losses persist, or the AI-funding narrative cracks; multiple compresses sharply on negative FCF.
~$730B-$1T
IPO in the floated range; growth strong but share/price pressure caps the multiple near ~30-40x forward.
$1T-$1.3T+
IPO at top of range; ARR compounds to ~$45-60B by 2027 on agentic/enterprise wins, multiple holds as profitability path is believed.
Reverse-DCF read: an ~$850B-$1T mark on ~$25B ARR with deep losses requires sustained ~60-80%+ revenue CAGR for several years AND a credible path to 30%+ FCF margins by ~2030 — i.e., the price embeds near-flawless execution, durable pricing power despite a price war, and continuous capital access to fund the build. Any of (share erosion, margin compression, capital-market closure) breaks the case. Honest gaps: no GAAP filings, opaque gross margin, and enormous off-balance-sheet compute obligations not captured in a simple revenue multiple.
SWOT
Strengths
- Category-defining brand: ChatGPT is the default consumer AI, ~800M+ weekly users
- Fastest-scaling revenue in software history ($3.7B→$13.1B→~$25B ARR)
- Frontier model capability (GPT/o-series) and deep research/agentic lead
- Capital access: $122B committed in Mar-2026 round; backing from Microsoft, Nvidia, SoftBank, Amazon
- Massive locked-in compute supply (Stargate, Azure, Oracle, AWS) hard for rivals to match
Weaknesses
- Catastrophic losses and cash burn (~$21B op loss 2025; ~$27B burn 2026) — not FCF-positive until ~2030
- Compute purchase obligations (hundreds of $B) far exceed revenue and create circular financing risk
- Eroding consumer share (76.5%→54.7%) and trailing in enterprise API (Anthropic ~40% vs OpenAI ~27%)
- Heavy dependence on a few suppliers (Nvidia GPUs, hyperscaler clouds) and on continued capital inflows
- Governance/structure complexity from nonprofit→PBC conversion; Microsoft relationship renegotiated
Opportunities
- Enterprise + agentic AI monetization (workflow agents, coding, vertical copilots)
- Platform/ecosystem lock-in via API, GPTs, and app store-like distribution
- New modalities/products (Sora video, voice, devices) expanding TAM
- IPO unlocking permanent capital to fund the compute build-out
- Pricing power as models become mission-critical infrastructure
Threats
- Anthropic gaining enterprise share + winning ~70% of head-to-head deals; Google Gemini's distribution
- Commoditization / price war (Google slashing Gemini prices) compressing API economics
- AI-bubble / funding-environment risk if capital markets sour before profitability
- Regulatory, copyright, and safety/litigation exposure
- Supplier concentration and Nvidia-circular-financing scrutiny — open-weights models (Llama, DeepSeek, Qwen) undercutting
Moats, dependencies & bottlenecks
Moats
ChatGPT is the household AI name, but share is eroding to Gemini — distribution-led rivals are catching up.
Stargate + Azure/Oracle/AWS commitments give unmatched capacity, but it's bought with cash/equity, not earned cash flow.
Frontier lead is real but narrowing; Anthropic and Google match or lead on specific axes (enterprise, coding, multimodal).
Large developer base and GPTs, but switching costs are modest and open-weights alternatives exist.
Premier backers and IPO access — contingent on markets staying open and the AI narrative intact.
Dependencies
Supplier (GPUs) + investor Core training/inference silicon; Nvidia up to $100B invest tied to 10GW deploy — supply + circular-financing dependency.
Cloud + investor + partner ~$17.2B paid in 2025; $250B incremental Azure commitment; equity stake and renegotiated rights.
~$300B (2027-31) OCI capacity; Abilene flagship campus core to Stargate.
$38B 7-yr AWS deal (signed Nov 2025) diversifies compute; Amazon anchored the Mar-2026 round (~$50B).
~$27B/yr burn requires continuous external capital; IPO is the relief valve — and the single point of failure if markets close.
Supplier (alt silicon) ~6GW AMD (~$90B), Broadcom custom inference chips (H2-2026) to reduce Nvidia reliance and cost.
Advantages
- Fastest revenue ramp in software history (~$25B ARR)
- Consumer mindshare and ~800M+ weekly users
- Unmatched committed compute supply (Stargate + three hyperscalers)
- Tier-1 capital backers and IPO access
- Full-stack: consumer app + API platform + enterprise + research
Weaknesses
- Enormous, widening operating losses and multi-year cash burn
- Off-balance-sheet compute obligations far exceeding revenue
- Eroding consumer share and trailing enterprise-API position
- Supplier and capital-market concentration risk
- Extreme valuation (~35x forward revenue) with negative margins
Bottlenecks
- Compute/power availability — GPUs, datacenter capacity, and gigawatts of electricity
- Cash — sustaining ~$27B/yr burn until ~2030 profitability
- Enterprise sales execution vs Anthropic's momentum
- Margin structure — inference cost vs price competition from Google/open-weights
- Governance/structure overhang from the for-profit conversion and Microsoft terms
Top signals & trends
Top signals
Moves toward a Q4-2026 IPO and permanent capital; range floated $730B-$1T.
Validates funding access; Amazon/Nvidia/SoftBank anchors.
Crystallizes the burn; though ~$41.6B is a non-cash conversion charge (adjusted ~$8B).
Gemini and rivals closing the gap; consumer growth stalling per S-1 reporting.
Enterprise momentum shifting; Anthropic wins ~70% of head-to-heads.
Trends
Expands enterprise TAM and per-seat value if OpenAI leads on reliable agents.
Google slashing Gemini prices and open-weights models compress API margins.
OpenAI's scale advantage, but also the source of its cash burn and circular-financing risk.
OpenAI re-pivoting to enterprise while defending consumer share.
Public-market access for capital-intensive AI leaders.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Primary GPU supplier; up to $100B investment tied to 10GW NVIDIA-systems deployment.
Core cloud + investor; $250B incremental Azure commitment (2025-30).
~$300B OCI capacity (2027-31); Stargate Abilene flagship campus.
$38B 7-yr compute deal (GB200/GB300 EC2 UltraServers); round anchor.
~6GW commitment (~$90B revenue potential) diversifying silicon.
Custom inference ASICs (H2-2026) to cut Nvidia reliance and cost.
Dedicated GPU compute under a 5-yr contract worth up to ~$11.9B, expanded by up to ~$6.5B.
750MW of wafer-scale inference compute under a >$20B multi-year deal (Dec 2025; deploying 2026-28, option +1.25GW by 2030); co-designing models for Cerebras hardware.
>40% of revenue; Fortune-500 adoption for copilots, agents, coding.
API platform powering thousands of AI-native products.
~$65% of revenue; ~800M+ weekly active users.
ChatGPT integrated into Apple Intelligence (distribution channel).
Runs OpenAI GPT/Codex models as selectable options, with a dedicated Codex agent harness.
Notion AI features built on OpenAI GPT models.
Duolingo Max (Roleplay, Explain My Answer) powered by OpenAI GPT-4 under a licensed partnership.
Enterprise-API leader (~40% share), Claude/Claude Code momentum, ~$4B→up-to-$18B ARR path; wins ~70% of head-to-heads.
Gemini's distribution via Search/Android/Workspace; consumer share rising to ~27%; aggressive pricing.
Partner and investor, but Copilot + multi-model strategy makes it a coopetitor.
Open-weights Llama undercuts API pricing and seeds an alternative ecosystem.
Grok + Musk capital + X distribution; frontier ambitions.