Skip to content
AI IPOs and valuations: OpenAI, Anthropic, and the model race
Skip to content
Scanner
News
LIVEEconomy & Finance· August 3, 2026

AI IPOs and valuations: OpenAI, Anthropic, and the model race

AI IPOs and valuations: OpenAI, Anthropic, and the model race

Key highlights

  • Timing and level are two separate trades. Kalshi runs monthly ladders on when OpenAI and Anthropic each announce a listing, while Polymarket carries contracts on whether the listing itself completes before 2027.
  • Anthropic is priced ahead of OpenAI on the contracts that ask which of the two reaches the public market first, as of August 2026. That ordering is what every valuation contract below it is calibrated against.
  • Valuation contracts settle on two different clocks: private marks reached at any point before December 31, and the market cap an OpenAI listing closes at on its first trading day.
AI IPOs and valuations trade as two separate questions now, and keeping them apart is most of the edge. OpenAI and Anthropic are both still private, so one set of contracts prices timing (when each lab announces a listing, and whether the listing completes) while a second set prices level (where each private mark lands before the year ends, and where an OpenAI debut would close on day one). Anthropic sits ahead of OpenAI on the who-goes-first contracts as of August 2026, which matters beyond the pair: the first frontier lab to price sets the comparison every later listing gets read against. Around that sit the two races that decide whether the capital keeps arriving. One asks which company holds the best model when December closes; the other asks which public company ends the year largest by market cap. Nvidia is the hinge between them, since it supplies the accelerators the frontier labs depend on and its own share price is what the largest-company contract settles against.

What will move the AI listing odds.

Several open questions will decide how these markets resolve:
  • Whether Anthropic announces before OpenAI, since the first listing sets the comparison every later one is priced against in the AI IPO and valuation markets
  • How public investors price revenue multiples far above precedent for companies this large that are not yet profitable
  • Where each lab's private mark lands before year end, which is what the valuation ladders settle on whether or not a listing happens
  • Whether the best-model leaderboard changes hands again before December, since these contracts read the benchmark rather than the marketing
  • Nvidia's position as Alphabet ramps custom tensor processing units and other hyperscalers test alternatives to GPU compute

How prediction markets price the AI economy.

AI Model Race: Both venues ask which company holds the leading model, and both resolve against a public leaderboard rather than a vendor claim. They are not the same contract: Polymarket asks who is on top when December closes, Kalshi asks who reaches the top at any point before 2027, and the two list different rosters of companies. Read the rules before comparing the two prices.

IPO Pipeline: Announcing a listing and completing one are different events with different dates, and the contracts split along that line. Kalshi prices the official announcement in monthly steps running into 2027; Polymarket prices the completed listing at fixed cutoffs. A lab can clear the first and miss the second inside the same quarter.

Valuations & Market Cap: Two ladders run in parallel for each lab, one upward and one downward, so a level can pay on a markup or a markdown. Separately, a bracket set asks where an OpenAI listing closes on its first day, which needs two things to land at once: that it lists at all, and that it prints inside the range. The largest-company contract is the public-market version of the same question.

AI Wild Cards: A small set of contracts prices the structural surprises: an AGI announcement by fixed deadlines, an acquisition of a frontier lab, and a trillion-dollar OpenAI debut. Any one of them reprices most of the rest of the board, which is why they trade cheap and stay liquid.

Catalysts to Watch: An S-1 becoming public, a funding round closing, and the next GPU cycle shipping are the three events that move several of these contracts at once. Each of them lands on a known calendar rather than arriving as a surprise, which is what makes the ladders worth watching between resolution dates.

Total Volume
Active Markets
Platforms

Key indicators

Nvidia (NVDA)
Google (GOOGL) YTD+

AI industry timeline.

Feb 12
Anthropic $30B round closes at $380B
Market impact: Earlier valuation benchmark; later eclipsed by the $900B round
Mar 31
OpenAI closes record $122B round at $852B post-money
Market impact: Amazon $50B, Nvidia $30B, SoftBank $30B; the mark the valuation ladders start from
Apr 30
Anthropic begins $50B round talks near a $900B valuation
Market impact: Would more than double its Feb mark and pass OpenAI
May 18
Jury dismisses all claims in Musk v. OpenAI (statute of limitations)
Market impact: Removes a legal overhang; Musk vows to appeal
Sep 1
First Kalshi cutoff on an Anthropic IPO announcement
Market impact: Earliest rung of the announcement ladder settles
Sep 30
First Polymarket cutoff on a completed Anthropic IPO
Market impact: Announcing and completing settle on different dates
Oct 31
Second Polymarket cutoff on a completed Anthropic IPO
Market impact: Middle rung of the completion ladder
Nov 1
Kalshi cutoff on an OpenAI IPO announcement
Market impact: Announcement cutoffs then run monthly into 2027
H2 '26
Nvidia Vera Rubin chips begin shipping
Market impact: Next-gen GPU cycle; reinforces or disrupts the market cap lead
2026-27
Alphabet TPU external sales ramp
Market impact: Could erode Nvidia GPU dominance in AI compute
Dec 31
Best model, largest company, and IPO-before-2027 contracts resolve
Market impact: Model race, market cap race, and the listing question settle together
Jan 1
Year-end valuation ladders on OpenAI and Anthropic settle
Market impact: Level questions resolve whether or not either lab has listed
On IPO
OpenAI closing market cap brackets resolve
Market impact: Needs both a completed listing and a first-day close inside the range

AI industry reference data.

MetricValueContext
OpenAI Valuation$852B (post-money)$122B round closed Mar 2026; Amazon $50B, Nvidia $30B, SoftBank $30B
Anthropic Valuation~$900B (in round)$50B round in progress; up from $380B in Feb 2026
OpenAI Revenue~$24B run-rate~$2B/month; $13.1B full-year 2025; not yet profitable
Anthropic ARR~$30BAs of end of March 2026
Databricks Valuation$134B$5.4B ARR; investors include Microsoft, JPMorgan, a16z
Mistral AI Valuation~$13.7B (€11.7B)ASML lead investor; Nvidia and a16z participating
Nvidia Market Cap~$5.2TWorld's #1 by market cap; ~80-90% AI accelerator share
Hyperscaler 2026 Capex~$650BMeta, Microsoft, Amazon, Alphabet combined
US Data Center Power Share~4.4%Of total US electricity (2023 DOE); projected ~12% by 2028

How AI models are ranked.

BenchmarkWhat It MeasuresHow It WorksCurrent LeadersWhy Traders Care
Chatbot Arena ELOHuman preferenceUsers vote blind between anonymous models; millions of votesGPT-5 (~1561), Claude Opus 4.6, Gemini 3.1 ProPrimary metric for "best model" resolution
MMLU-ProBroad knowledge14,000 expert-level multiple choice across 57 subjectsTop models: 85-92%General intelligence proxy
GPQA DiamondExpert science reasoningPhD-level questions designed to stump non-specialistsGemini 3.1 Pro (94.1%), GPT-5.4 (92.0%)Tests deep reasoning, not pattern matching
SWE-bench VerifiedReal-world codingCan the model resolve actual GitHub issues end-to-end?Claude Opus 4.6 (80.8%), GPT-5.2 (80.8%)Proxy for enterprise coding value
Humanity's Last ExamFrontier capability3,000+ questions from 1,000+ expertsBest models: ~25-35%Measures distance to AGI-level capability
AIME 2025Mathematical reasoningCompetition-level multi-step math problemsClaude Opus 4.6 (99.8%), GPT-5.2 (~100%)Tests reasoning depth, not memorization

AI company financial profiles.

CompanyValuationARR (Mar 2026)Rev. MultipleBusiness MixKey Investor
OpenAI$852B (post-money)~$24B~36x (on ARR)Consumer 60%, Enterprise 40% (900M+ WAU)Amazon ($50B), SoftBank ($30B), Nvidia ($30B)
Anthropic~$900B (in round)~$30B~30x (on ARR)Enterprise-heavy (~80% of rev)Amazon, Google, plus $50B round in progress
Google DeepMindPart of AlphabetN/A (internal)N/AGemini integrated; TPU + CloudPublic company (GOOGL)
Mistral AI~$13.7B (€11.7B)~€300M (Sept '25)~39xAPI + enterprise licensesASML (lead), Nvidia, a16z
Databricks$134B$5.4B~25xData + AI platformJPMorgan, Insight, a16z, Microsoft

The compute arms race.

MetricValueContext
Nvidia H100 GPU price$25,000-$40,000 eachWorkhorse of current AI training clusters
Nvidia B200 (Blackwell)2.5x inference throughput of H100At 1.4x the power draw; shipping now
Nvidia Vera RubinNext-gen, arriving H2 2026Expected to extend Nvidia's lead vs. custom silicon
Cost to train GPT-4~$79M (compute only)25,000 A100 GPUs; ~50 GWh electricity consumed
Cost of "GPT-4 equivalent" today~$5-10MDown from $79M in 2023 due to efficiency gains
Frontier model cost by 2028$10B+ projectedPer Anthropic CEO Dario Amodei
Single GPU rack power draw80-140 kWvs. 3-5 kW for standard server racks (20-40x more)
10,000-GPU cluster power10-15 MWEnough to power a small town
Hyperscaler 2026 capex (Big 4)~$650B combinedMeta, Microsoft, Amazon, Alphabet

Where the money goes: AI spending breakdown.

Cost CategoryShare of SpendWhat It CoversWhy It Matters
GPU/Chip Hardware50-60%Nvidia GPUs ($25-40K each), HBM memory, networkingNvidia's revenue directly tracks this spend
Power Infrastructure15-20%Grid connections, substations, backup generatorsWait times now 3-5 years in Virginia
Cooling Systems5-10%Liquid cooling ($500K-$2M per MW), HVACDense GPU racks need 20x traditional cooling
Construction/Land10-15%Buildings, land acquisition, permitting$10.7M per MW construction cost (2026 avg)
Talent/Stock CompensationSignificant (off-balance)Core engineers command $10M+ signing bonusesOpenAI projects $50B in employee stock value by 2030
Inference at Scale80-90% of operational energyServing billions of daily queriesRevenue depends on inference cost efficiency

The AI industry arc: training to agents.

PhaseEraKey CharacteristicWho BenefitsWhere We Are
Foundation Training2020-2024Race to build largest, most capable base modelsLabs with most compute: OpenAI, Google, AnthropicMature; costs $100M+
Inference Scaling2024-2025"Test-time compute": models think longer on hard problemsLabs with efficient architecturesActive; token cost down 280x in 2 yrs
Coding Agents2025-2026AI writes, tests, and deploys code autonomouslyAnthropic (Claude Code $2.5B ARR), CursorExplosive growth; 4% of GitHub commits
Enterprise Agents2026-2027AI handles end-to-end business workflowsCompanies with enterprise trustEarly; driving Anthropic's 80% enterprise mix
Industry-Specific AI2027+Custom models for healthcare, legal, financeVertical specialists with domain dataNascent; requires regulatory approval
Autonomous Systems2028+Persistent AI agents operating independentlyUnknown; depends on AGI progressSpeculative; connects to AGI markets

How each contract family resolves.

Contract FamilyWhat It AsksWhat Settles ItWhat A Trader Is Pricing
Best model at year endWhich company holds the leading model when December closesThe public leaderboard named in the market rules, read on the resolution dateWhether the current leader keeps the top slot through one more release cycle
Top-ranked model before 2027Whether a company reaches the top slot at any point before January 1The same class of leaderboard, but touched once rather than held at a dateA brief spike counts here and does not count in the year-end contract
IPO announcement dateWhether a lab has officially announced a listing before each monthly cutoffA public announcement, not a completed listingHow quickly a filing becomes public news
IPO completion dateWhether the listing itself has happened by the cutoffShares trading on a public exchangeThe gap between announcing and actually pricing
Year-end valuation ladderWhether a private mark reaches a level, upward or downwardA reported valuation from a funding round or secondary saleRound timing as much as round size
IPO closing market capWhere a first trading day closes, if a listing happens at allMarket cap at the close on IPO dayTwo things at once: that it lists, and where it prints
Largest companyWhich public company ends the year biggest by market capShare price on the resolution dateWhether a rival overtakes the incumbent inside the window
Acquisition and AGI wild cardsWhether a structural surprise lands before a deadlineA completed acquisition, or an official capability announcementTail risk that would reprice most of the rest of the board

Mega-tech IPO precedents.

CompanyIPO YearIPO ValuationRevenue at IPORev. MultipleFirst-Year ReturnProfitable?
Google2004$23B$3.2B7.2x+80%Yes
Facebook2012$104B$5.1B20.4x-30% (Y1)Barely
Alibaba2014$231B$12.3B18.8x+46%Yes
Snowflake2020$33B$592M55.7x+112% (day 1)No
Arm Holdings2023$54B$2.68B~20x+25% (day 1)Yes
OpenAI (pending)Not announced$852B-$1T+~$13B ('25)65-77xTBDNo
Anthropic (pending)Not announced~$900B~$4.5B ('25)~30x (on ARR)TBDNo

All related markets.

events · markets
POLY
AI Model Race markets
Best AI model at the end of
Anthropic
KLSH
AI Model Race markets
Top-ranked AI model before
OpenAI
POLY
IPO Pipeline markets
AI companies to IPO before
Databricks
KLSH
IPO Pipeline markets
When OpenAI announces an IPO
Before Jan,
KLSH
IPO Pipeline markets
When Anthropic announces an IPO
Before Sep,
POLY
IPO Pipeline markets
Anthropic IPO completed by
September, +
KLSH
IPO Pipeline markets
Which lists first, OpenAI or Anthropic
Anthropic
POLY
Valuations & Market Cap markets
Anthropic valuation by December
High: T-
POLY
Valuations & Market Cap markets
OpenAI valuation by December
High: -
POLY
Valuations & Market Cap markets
OpenAI IPO closing market cap
Above T
POLY
Valuations & Market Cap markets
Largest company by market cap on December
NVIDIA+
KLSH
AI Wild Cards markets
When OpenAI announces AGI
By end of
POLY
AI Wild Cards markets
AI companies acquired before
Perplexity AI
POLY
AI Wild Cards
OpenAI T+ IPO before ?
Yes-
Ben L.
Ben L.

Founder & CEO

Former gaming executive, multiple-time founder, and public and private market investor. Brings the operator perspective most sites can't: why platforms make the decisions they do, and what that means for you.

More news

More in Economy & Finance.

SpaceX prediction markets: Starship, launch cadence, and the Moon
Economy & Finance

SpaceX Prediction Markets 2026: Starship, Launches, Moon

The SpaceX IPO has priced and settled. What still trades: Starship reusability and launch cadence, docking milestones, NASA Moon timelines, and the Tesla merger question, live on Polymarket and Kalshi.

Bitcoin price prediction 2026: live odds and crypto forecasts
Economy & Finance

Bitcoin Price Prediction 2026: Odds & Forecasts

BTC price thresholds, altcoin returns, crypto regulation, and institutional stress contracts on Polymarket and Kalshi.

Gold, oil, and silver: 2026 year-end price predictions and odds
Economy & Finance

Gold, Oil & Silver: 2026 Year-End Price Predictions and Odds

The mid-year gold, oil, and silver contracts settled in June. What still trades: year-end gold touch and level ladders, the WTI maximum and minimum ladders, crude record-high contracts, and the cross-asset return races on Polymarket and Kalshi.

Explainers

Economy & Finance explainers.

Small business owner reviewing costs at the counter of a neighborhood shop
Strategy

How Small Businesses Hedge Risk With Prediction Markets

A bar, an ice cream shop, and a conference organizer all hedged real business risk with prediction market contracts. How the hedges worked, what the arithmetic looks like, and where the limits sit.

Beginner12 min read
How Prediction Market Fees work dashboard
Fees & Costs

The 6 Fees That Eat Your Prediction Market Profits

Six fee categories, one $100 trade. See what Kalshi, Polymarket, and FanDuel Predicts actually take.

Beginner11 min read

Track markets you care about. Set up free alerts in 30 seconds.

Last updated