Anthropic, OpenAI and SpaceX: Three Paths to Public Markets That Could Reshape the Global Economy
- Andrea Viliotti
- 18 ore fa
- Tempo di lettura: 15 min
IPOs, public capital, industrial AI and scenario analysis to 31 December 2026
17 June 2026 Scenario analysis, not investment advice
ANTHROPIC Confidential draft S-1 No public price range or timetable | OPENAI Confidential draft S-1 Timing not decided | SPACEX Public S-1 and priced IPO SPCX trading from 12 June 2026 |
Reader note This article reconstructs the history, competitive position and valuation of Anthropic, OpenAI and SpaceX, and treats their moves towards public markets as a stress test for the global economy. Values to 31 December 2026 are rounded scenario corridors - not target prices, calibrated probabilities or trading signals. Unless otherwise stated, all figures are in US dollars; one trillion equals 1,000 billion. |
The argument in 90 seconds
During 2026, artificial intelligence and the space economy have started moving from private markets dominated by specialist funds towards public capital that can be held by millions of investors. SpaceX is the most advanced case: it has a public S-1 registration statement, a priced IPO, the ticker SPCX and an implied equity value that moved above US$2 trillion in its first trading sessions. Anthropic and OpenAI remain one stage earlier. Both announced confidential submissions of draft S-1 registration statements, but neither had published a prospectus with a final share count, price range or definitive timetable as of the data freeze. [S1][S3][S7][S8]
The central question is not simply how much these companies may be worth. It is what kind of market emerges when frontier AI models, satellite networks, cloud infrastructure, computing capacity, software agents and robotics become parts of the same investable system. If the market absorbs these offerings without damaging market breadth, credit conditions or volatility, the signal would extend beyond a successful IPO. It would suggest a system-level rebound: a move from fear about the cost of AI towards acceptance of AI as economic infrastructure.
Current values need to be read through two different lenses. Anthropic raised US$65 billion in a Series H round at a reported post-money valuation of US$965 billion and reports annualised run-rate revenue above US$47 billion. OpenAI announced US$122 billion of committed capital at a post-money valuation of US$852 billion and said revenue had reached US$2 billion per month in March. SpaceX raised approximately US$75 billion in its IPO at US$135 per share; at US$190.90, its implied equity value was approximately US$2.50 trillion. These are extraordinary figures, but they do not automatically translate into margins, returns or stability. [S2][S4][S7][S8]
The thesis These are not merely stock-market events. They represent an attempt to bring the core of a new economic infrastructure - AI models, computing capacity, orbital connectivity and software automation - into public ownership. The opportunity is faster investment in the real economy; the risk is that liquidity prices scarcity before profitability. |
Three histories, three routes to market
Anthropic: the enterprise trust proposition
Anthropic was built around the goal of developing reliable, interpretable and controllable AI systems. By 2026 it was no longer simply a competitor to OpenAI. It had become an enterprise platform offering Claude models, coding tools, agents and cloud integrations. Its distinguishing proposition is the combination of safety credibility, product capability and corporate demand. [S19]
The US$65 billion Series H at a reported US$965 billion post-money valuation, together with annualised run-rate revenue above US$47 billion, shows that investors are valuing Anthropic as AI infrastructure rather than as an application start-up. The confidential draft S-1 gives the company the option to list when market conditions are favourable, but it does not yet establish a price, share count or timetable. [S1][S2]
Anthropic's economic value will depend on the quality of enterprise revenue: renewals, multi-year contracts, deployment in critical workflows, inference costs and continued credibility in a market that demands both capability and control. Trust is its strength; the risk is that trust alone may not protect margins if model prices fall.
OpenAI: the platform that created a mass market
OpenAI is the company that turned generative AI into a mass-market consumer and enterprise category. Its strength is the width of its distribution funnel: ChatGPT, APIs, developer products, coding tools and enterprise deployment. The company describes itself as an AI research and deployment organisation with a mission to ensure that artificial general intelligence benefits humanity. [S18]
I
n 2026, its financial story entered a new phase: US$122 billion of committed capital at a reported US$852 billion post-money valuation, monthly revenue of US$2 billion reported in March, and a confidential draft S-1 announced on 8 June. The fragile side is cost. Reuters, citing The Information, reported first-quarter 2026 revenue of US$5.7 billion and cash burn of US$3.7 billion; Reuters said it could not independently verify the figures. [S3][S4][S5]
OpenAI's economic value depends on converting scale into margins. If mass usage becomes a stable enterprise platform, very high valuation corridors may be defensible. If computing costs and competitive pressure rise faster than revenue per user, public investors will demand a level of financial discipline that private capital has so far been able to defer.
SpaceX: from reusable launch systems to a space-and-AI platform
Space Exploration Technologies, or SpaceX, was founded in 2002 to reduce the cost of access to space and transform launch technology. Its industrial trajectory runs from Falcon and Dragon to Starlink and then to a broader system in which connectivity, satellites, AI and computing capacity converge. Unlike Anthropic and OpenAI, SpaceX was already exchange-listed at the data freeze. [S20]
The S-1 filed on 20 May 2026 is public, the IPO was priced at US$135 per share, the ticker is SPCX and trading began on 12 June. On 16 June, the company also disclosed an agreement to acquire Cursor/Anysphere at an implied equity value of US$60 billion, signalling that AI software is moving into the post-IPO strategic perimeter. [S6][S7][S9][S21]
SpaceX's value rests on two axes. The first is Starlink: recurring revenue, global customers and connectivity for underserved locations. The second is industrial capability: launches, Starship, data centres, vertical integration and public-sector contracts. The risk is that markets capitalise mature activities and expensive options together. The strength is that very few companies can combine hardware, orbit, networks and software at comparable scale.
A compressed chronology: from mission to public capital
Anthropic's history is short but dense. Markets are already valuing it as if it has moved beyond the pure laboratory stage: enterprise demand, cloud distribution, developer tools and a reputation for safety place it in the category of trusted infrastructure. A listing, if it occurs, would mainly finance computing capacity, global commercial reach and credibility with regulated customers.
OpenAI followed a different trajectory. It first created the mass market for generative AI and then turned that market into a platform for developers and enterprises. Its strength is distribution; its weakness is the cost of maintaining the research frontier and serving inference at global scale. Its confidential S-1 is therefore best understood as a strategic option, not a fixed listing date.
SpaceX is the oldest and most industrial of the three. More than two decades of launches, reusability, government contracts, Starlink and vertical integration created operating credibility. The IPO changes the perimeter: investors are not only buying a space company, but a platform attempting to combine orbital transport, global connectivity, data centres, AI and software. That ambition explains both the enthusiasm and the risk of valuing activities with very different maturity and margin profiles.
Competitive positioning and what the valuations really mean
Anthropic and OpenAI compete at the frontier of generative AI: reasoning, coding, agentic tools, safety, enterprise distribution and access to computing capacity. Their competition will not be decided by benchmark scores alone. It will be decided by the cost of serving each request, reliability in production, integration into business workflows and the quality of governance.
SpaceX occupies a different position. It does not sell only software or models; it operates physical and orbital infrastructure. Starlink converts satellite capacity into recurring revenue, while launches and Starship remain capital-intensive industrial assets with technical risk. The AI component could turn the company into a vertically integrated platform in which networks, energy, data centres and software reinforce one another.
The economic value of the three companies is not the same as their headline valuation. A valuation records what capital is prepared to pay today for the possibility of owning tomorrow's infrastructure. Real value will depend on four tests: durable revenue, margins after compute and capital expenditure, the ability to monetise enterprise automation, and geopolitical defensibility in a world in which the United States and China are pursuing different technology architectures.
Company | Public-market status | Observable current value | What the value is really pricing |
Anthropic | Confidential draft S-1; no direct public S-1 with pricing | US$965bn post-money; annualised run-rate revenue above US$47bn | Value depends on enterprise demand for Claude, trust and control of compute costs. [S1][S2] |
OpenAI | Confidential S-1; timing not decided | US$852bn post-money; US$2bn monthly revenue reported | Value depends on global distribution, APIs, ChatGPT, coding products and the conversion of usage into margins. [S3][S4] |
SpaceX | Public S-1; priced IPO; SPCX trading | US$1.77tn at IPO; about US$2.50tn implied at US$190.90 | Value depends on Starlink, launches, AI integration and the ability to sustain very high capital expenditure. [S7][S8][S10] |
Table 1 - Current value: observable figures and interpretation. Private post-money valuations are not directly equivalent to listed market capitalisation.
Scenario analysis to 31 December 2026
The values to 31 December 2026 should not be read as point forecasts. They are scenario corridors. The downside case assumes that investors conclude that revenue and margins do not justify scarcity premiums. The base case assumes that the IPO process or preparations advance and AI remains central to corporate spending. The upside case assumes that public markets accept these assets as a new layer of economic infrastructure and that major indices begin to absorb them.
Company | Downside31 Dec 2026 | Base31 Dec 2026 | Upside31 Dec 2026 | Main corridor drivers |
Anthropic | US$0.55-0.75tn | US$0.75-1.35tn | US$1.35-1.60tn | Enterprise demand, model pricing, compute costs, IPO timing and safety perception. |
OpenAI | US$0.55-0.80tn | US$0.80-1.35tn | US$1.35-1.70tn | Distribution scale, revenue per user and enterprise, cash burn, governance and compute partnerships. |
SpaceX | US$1.40-2.10tn | US$2.10-3.20tn | US$3.20-4.00tn | Starlink, capital expenditure, AI integration, launches/Starship, post-IPO volatility and index entry. |
Table 2 - Scenario corridors to 31 December 2026. These are not target prices and do not carry calibrated probabilities.
In the downside case, the shared mechanism is normalisation: revenue may continue to grow, but investors reject extreme multiples if computing costs, price competition or regulation reduce margins. In the base case, the companies remain at the centre of AI spending and investors treat 2026 as a transition from private to public ownership. In the upside case, the listings become a new benchmark: AI is no longer priced merely as a theme, but as listed infrastructure.
Effects on stock markets, indices and capital flows
The first effect is scale. A primary offering of approximately US$75 billion, such as SpaceX's, is not only corporate fundraising; it is a redistribution of market liquidity. If investors absorb that supply while demand for semiconductors, cloud providers, software and credit remains stable, the market is demonstrating resilience. If demand for mega-IPOs is financed by selling other AI leaders, the market is not broadening; it is concentrating. [S7][S9]
The second effect concerns indices. Nasdaq updated its Nasdaq-100 methodology to respond to the emergence of mega-cap IPOs. Russell US indices add eligible IPOs quarterly. S&P Dow Jones Indices chose not to create a market-capitalisation-only exception, preserving financial eligibility and trading-history requirements. This creates three different routes: potentially faster entry into some benchmarks, scheduled quarterly entry into others, and a longer path to the S&P 500. [S11][S12][S13]
The third effect is behavioural and structural. While Anthropic and OpenAI remain private, retail investors and many passive funds gain AI exposure indirectly through Nvidia, cloud companies, semiconductor producers, enterprise software and thematic exchange-traded funds. Once frontier-model developers list, investors can buy direct exposure to the model provider rather than only to its suppliers. Capital moves, but risk moves as well: public portfolios become more exposed to training and inference costs, model regulation and Chinese competitive pressure.
Signals of a system-level rebound
In this analysis, a rebound is not a promise that prices will rise. It is the market's ability to absorb a large new block of capitalisation without losing balance. The observable indicators are market breadth, volatility, credit spreads, the response of semiconductor shares, the relative performance of AI companies not involved in the IPOs, demand in the order book and the stability of passive funds.
A constructive pattern would combine strong primary-market demand with broad participation across the rest of the market. A fragile pattern would show headline indices rising while most securities weaken, credit spreads widen or investors sell existing AI leaders to fund the new issues. The distinction matters because a listing can be successful for the issuer while still reducing resilience elsewhere in the system.
Signal | Constructive reading | Risk reading |
IPO absorption | Strong primary demand without a broad sell-off | Demand is financed by selling other AI leaders |
Market breadth | The advance includes many sectors and securities | The index rises while most securities weaken |
Volatility | High IPO volatility remains contained | Volatility spreads to Nasdaq, credit and infrastructure assets |
Indices and ETFs | Passive flows are orderly and progressive | Forced buying meets low free float and creates crowding |
Credit and private markets | Financing for compute and data centres remains open | Venture, private credit and infrastructure debt reprice sharply |
Real economy | AI budgets move from pilots into governed production | AI spending is cut because of cost, trust or governance problems |
Table 3 - Indicators of proximity to a system-level rebound. These are monitoring variables, not trading signals.
United States, China and Europe: different consequences
United States: public capital for AI infrastructure
For the United States, these listings have geoeconomic significance. They make publicly financeable the assets that previously lived mainly inside venture capital, private equity, hyperscalers and strategic agreements. This strengthens the US advantage in closed frontier models, computing capacity and capital markets. It also increases concentration: if a small number of securities absorb global savings, the system becomes more powerful and more fragile at the same time.
China: lower-cost models and embodied AI
China's competitive response does not need to reproduce the US model. Pressure comes from open-weight or lower-cost systems associated with laboratories and companies such as DeepSeek, Qwen and Kimi, and even more importantly from the connection between AI and manufacturing. [S14]
The central signal is embodied AI - artificial intelligence operating through robots and other physical systems. China combines industrial robotics, humanoid programmes, factory data and supply-chain integration. MERICS notes that China already has the world's largest installed base of industrial robots and is actively exploring humanoid robotics. The International Federation of Robotics reports that robotics is at the heart of China's industrial strategy, with an operational stock of around two million units. [S15][S16]
Europe: dependency risk, but room for trust and industrial depth
For Europe, the three public-market transitions are a mirror. The continent risks paying two rents: US models and cloud infrastructure on one side, and Chinese hardware and robotics on the other. The opportunity is not closed. Europe can compete where trust, industrial data, compliance, specialised robotics, energy systems and vertical automation matter more than the largest general-purpose model. The political and economic question is who owns the operational data of European companies and who captures the margin created by automating it.
Intelligent enterprise automation: where the real impact appears
The destination is not the IPO. It is the company that changes how work is organised. Intelligent automation does not mean replacing every employee with an agent. It means redesigning processes, responsibilities and controls. The nearest-term effects are likely to appear in assisted coding, customer operations, internal research, procurement, compliance, finance, logistics and maintenance. The deeper effects begin when agents do not merely answer questions but execute parts of a workflow with defined identities, permissions, audit trails and accountability.
The limiting factor is organisational. McKinsey describes a market in which the use of AI and agentic AI is increasing, yet the transition from pilot projects to scaled value remains difficult. The winners will not be only those that purchase the strongest model. They will be the organisations that build data foundations, governance, operating models, human validation and risk measures. These public-market transitions can accelerate that process by placing public capital behind tools and infrastructure. They can also inflate expectations and create disappointment if measurable returns do not follow. [S17]
Impact on the global financial economy
These public-market transitions change the financial cycle of AI. Until now, a large share of the value has remained in private markets accessible to venture capital, sovereign funds, strategic partners and growth investors. With SpaceX listed and Anthropic and OpenAI preparing optional routes to market, AI risk begins to enter public portfolios: index funds, pension funds, retail accounts, thematic ETFs and global active strategies.
This produces three effects. First, it can lower the cost of capital for the leading companies because liquid public equity can support capital expenditure, acquisitions, debt financing and employee compensation. Second, it increases discipline because margins, cash burn, capital expenditure and risk factors become visible on a recurring basis. Third, it transfers volatility from private funding rounds to listed markets. AI becomes not only a private valuation story but also a component of global financial stability.
The systemic risk is not the failure of a single IPO. It is crowding. If too much passive capital pursues a small number of shares with limited free float, prices may reflect scarcity more than profitability. A correction would then extend beyond three names to semiconductors, cloud infrastructure, private credit, data centres, energy and software valuations. The healthier signal is a market that absorbs SpaceX and prepares for Anthropic and OpenAI without weakening the wider ecosystem.
Impact on the global real economy
In the real economy, public ownership can accelerate a shift already under way: turning AI from experimentation into operating capability. Public capital can finance data centres, chips, electricity supply, agentic software, satellites, user terminals, networks and acquisitions. The effect is not automatic. Technology must enter production processes rather than remain a demonstration dashboard.
In the United States, the most direct effect is the expansion of an AI-financial-industrial complex linking models, cloud computing, chips, networks, defence, space and automation. In China, the parallel effect is the strengthening of a manufacturing route based on efficient models, robotics, factories, operational data and physical automation. In Europe, the outcome depends on whether regulation and industrial expertise can be turned into products: vertical AI, industrial data, applied robotics, energy systems and compliance as a source of competitive advantage.
For labour markets, the issue is not a simple one-for-one substitution. It is the recomposition of tasks. Companies that deploy agents without governance may create additional workload, errors and control costs. Companies that redesign processes, permissions, audit trails and accountability may reduce cycle times, improve decision quality and automate bottlenecks. Large listings can make this transition faster; they do not make it simpler.
What to monitor through 31 December 2026
· SPCX: the relationship between price, trading volume, volatility and the expansion of free float.
· Nasdaq and Russell: the actual timing of any index inclusion and the scale of resulting passive purchases.
· S&P 500: no shortcut if profitability and trading-history requirements remain unmet.
· Anthropic and OpenAI: publication of an S-1, share count, price range, use of proceeds and risk factors.
· AI revenue: not only headline growth, but the quality of enterprise revenue, retention and contract duration.
· Cost of compute: margins after training, inference, electricity and data-centre expenditure.
· Chinese competition: releases of open-weight models and pressure on model pricing.
· Embodied AI: robotics and manufacturing contracts, industrial pilots and integration into production processes.
· Credit: spreads on data-centre financing, private credit and infrastructure debt.
· Real economy: movement from AI proof of concept to governed production.
The common thread
The common thread is straightforward: SpaceX, Anthropic and OpenAI are not three separate stories. They are three ways of financing the same transformation - computing capacity, intelligence, networks and automation. If public markets absorb them, AI moves beyond private promise and becomes listed infrastructure. If markets reject them, 2026 becomes the year in which public capital asks AI to demonstrate margins, not only capability.
For the United States, this is a test of technological and financial leadership. For China, it is a signal that the response will increasingly combine open or low-cost models, manufacturing and robotics. For Europe, it is a warning that regulating AI is not enough; Europe must own part of its operational value. The final question is not whether these companies can command very high valuations. It is whether the world can convert financial value into real productivity, better-organised work, more efficient infrastructure and more accountable decisions.
Technical annexes
Essential source ledger
ID | Source | Date / vintage | Load-bearing evidence | Link |
S1 | Anthropic, announcement of confidential draft S-1 | Anthropic, 1 June 2026 | Confidential S-1 submitted; share count and price not set | |
S2 | Anthropic Series H | Anthropic, 28 May 2026 | US$65bn raised; US$965bn post-money; annualised run-rate revenue above US$47bn | |
S3 | OpenAI confidential S-1 | OpenAI, 8 June 2026 | Confidential S-1 submitted; timing not decided | |
S4 | OpenAI funding round | OpenAI, 31 March 2026 | US$122bn committed; US$852bn post-money; US$2bn monthly revenue reported | |
S5 | OpenAI Q1 revenue and cash-burn report | Reuters, 16 June 2026 | Q1 2026 revenue of US$5.7bn and cash burn of US$3.7bn; Reuters said it could not independently verify the figures | |
S6 | SpaceX S-1 filing | SEC EDGAR, 20 May 2026 | Form S-1, Accession No. 0001628280-26-036936, file 333-296070 | |
S7 | SpaceX pricing announcement | SpaceX, 11 June 2026 | 555,555,555 shares at US$135; trading expected 12 June; 83,333,333-share over-allotment option | |
S8 | SPCX market quote | Runtime snapshot, 17 June 2026 19:58 UTC | SPCX US$190.90; intraday high US$213.65, low US$187.14 | Runtime quote snapshot |
S9 | SpaceX SEC free writing prospectus - pricing terms | SEC EDGAR, 11 June 2026 | Offering size US$74,999,999,925; ticker SPCX; trade date 12 June 2026 | |
S10 | Summary of SpaceX S-1 financials | Via Satellite, 20 May 2026 | 2025 revenue US$18.7bn; operating loss US$2.6bn; adjusted EBITDA US$6.6bn | |
S11 | S&P mega-cap consultation result | S&P Global, 4 June 2026 | No exception based only on market capitalisation; financial and trading-history requirements preserved | |
S12 | Nasdaq-100 methodology update | Nasdaq, 8 May 2026 | Methodology updated in response to the emergence of mega-cap IPOs | |
S13 | Russell US IPO additions | FTSE Russell / LSEG | Eligible IPOs can be added quarterly in March, June, September and December | |
S14 | DeepSeek and Chinese AI competition | Reuters, April/June 2026 | DeepSeek, Qwen and Kimi illustrate model and pricing pressure; DeepSeek funding above US$7bn was reported | |
S15 | Embodied AI in China | MERICS, 2026 | China has the world's largest installed base of industrial robots and is actively exploring humanoid robotics | |
S16 | Robotics as a Chinese national strategy | International Federation of Robotics, 5 May 2026 | Robotics placed at the centre of industrial strategy; operational stock around two million units | |
S17 | Enterprise and agentic AI adoption | McKinsey, 2025/2026 | Adoption is rising; scaled impact requires governance, data and operating-model change | |
S18 | OpenAI company and mission | OpenAI About | AI research and deployment company; mission focused on broadly beneficial AGI | |
S19 | Anthropic company and mission | Anthropic Company | Reliable, interpretable and steerable AI systems | |
S20 | SpaceX mission | SpaceX Mission | Founded in 2002 to transform space technology | |
S21 | SpaceX 8-K on Cursor/Anysphere agreement | SEC EDGAR, 16 June 2026 | Agreement to acquire Cursor/Anysphere at an implied equity value of US$60bn; closing expected Q3 2026 |
Commenti