X
Private giants & exposure vehicles
OpenAI, Anthropic, xAI — the AI cap tables without tickers, and the listed and quasi-listed paths into them.
The most valuable assets in the AI compute buildout do not trade. OpenAI ($852B post-money in March 2026) and Anthropic ($965B in its May 2026 Series H, with a ~$1.0–1.25T IPO base case) are the demand engines that pull GPUs, power, and datacenters through the entire physical stack — yet a public-equity investor can only reach them through a parent's balance sheet, a foreign ADR, or a listed fund wrapper. This chokepoint is the cap table itself: who owns the equity, and what each listed path actually conveys.
Who owns it
The model labs are owned by their cloud sponsors. Microsoft holds ~27% of OpenAI on an as-converted basis, crystallized at 26.79% in the October 2025 recapitalization into OpenAI Group PBC. Amazon's Anthropic position is worth roughly $74B on paper — $42.2B in convertible notes plus $32B in nonvoting preferred — with up to $25B more committed against an Anthropic commitment to 5GW of Trainium. Alphabet holds ~14% of Anthropic, contractually capped at 15%. SoftBank reached ~11% of OpenAI after funding its $40B commitment, and anchors the $500B Stargate buildout. Below the labs sit the neoclouds: CoreWeave is the most prominent neocloud to list (NASDAQ: CRWV, March 2025), turning OpenAI and Microsoft contracts into a $99.4B backlog.
What breaks it
Two things. First, circular financing. Nvidia's headline "up to $100B" OpenAI commitment shrank to a ~$30B equity stake; Jensen Huang conceded the $100B was "probably not in the cards". The same loop runs through CoreWeave, where Nvidia took a $2B equity stake plus a $6.3B take-or-pay backstop — vendor financing that can inflate apparent demand. Second, concentration: CoreWeave's Customer A was 45% of Q1-26 revenue, and the listed-fund wrappers trade at violent premiums to the very assets they hold — DXYZ once printed a >2,000% premium to NAV.
What forces the reprice
The IPOs. Anthropic confidentially filed its S-1 around June 1, 2026; OpenAI confidentially filed around June 8, 2026, though it has not committed to timing (it said "it may be a while"); sources put a possible window around Sep–Q4 2026 at $852B–$1T. Each debut converts a balance-sheet footnote into a marked, tradable line — and forces every proxy (MSFT, AMZN, GOOGL, SFTBY, VCX, DXYZ) to be re-derived against an actual price rather than a private mark. Earnings are the interim tape: Microsoft, Amazon, and Alphabet now disclose lab marks that swing reported income.
Be honest about size and earliness
The revenue behind these stakes is still thin relative to the narrative. Microsoft's equity-method line is a loss-share, not a revenue stream: its Q3-FY26 OpenAI loss was just $14M (vs. $583M a year earlier), and the nine-month figure was a $5.9B gain driven almost entirely by a one-time recap dilution gain — accounting, not cash. Fortune flagged that roughly half of Amazon's and Google's "blowout AI profits" came from Anthropic stake markups, not operating AI revenue. The revenue-share economics are being capped, not expanded: OpenAI capped Microsoft's cut at $38B through 2030, down from a projected $135B. And the cleanest retail wrappers are the most dangerous: VCX (Fundrise Innovation Fund) and DXYZ have traded at multiples of NAV, so the proxy can fall even if the underlying mark rises. For a mega-cap, the lab stake is a small slice of enterprise value; for a fund, it can be the whole thing at the wrong price. This chokepoint is about access structure and catalysts, not disclosed operating revenue — size the position to the wrapper's distortion, not to the headline valuation. None of this is investment advice.
Who owns the choke
Amazon.com, Inc.
Guided to ~$200B capex for 2026 (vs ~$132B in 2025). Project Rainier activated with nearly 500,000 Trainium2 chips — the largest commercial hyperscaler ASIC fleet by unit count — with Anthropic scaling toward a >1M-Trainium2 target.
- Bull
- If internal silicon scales, Amazon's ~$200B 2026 capex, ~500k-chip Project Rainier Trainium fleet and ~$74B Anthropic stake give it the deepest vertically-integrated AI position.
- Bear
- AI is a fast-growing slice of a vast retail/cloud P&L, and roughly half of recent 'AI profits' were Anthropic stake markups, not operating revenue.
Anthropic PBC
Confidentially filed an IPO S-1 around June 1, 2026. Its most recent priced round was the $65B Series H at a $965B post-money valuation (May 2026); banker IPO base case is ~$1.0–1.25T. Verified holders include Amazon (~$74B paper position in convertible notes and nonvoting preferred) and Alphabet (~14% equity, capped at 15%); Anthropic committed to up to 5GW of AWS Trainium compute.
- Bull
- If the IPO prices near the ~$1.0-1.25T banker base case, Anthropic's June 2026 S-1 marks a $965B private asset into tradable equity, lifting listed holders AMZN, GOOGL and VCX.
- Bear
- As a private company reachable only via AMZN, GOOGL or the premium-prone VCX fund, no clean tradable line exists and IPO timing/valuation are unconfirmed.
CoreWeave, Inc.
GPU neocloud serving OpenAI, Microsoft and Nvidia; IPO'd on NASDAQ in March 2025 at $40/share. Q1-2026 customer concentration was 45% from Customer A and 20% from Customer B; Nvidia holds a ~$2B equity stake plus a $6.3B take-or-pay capacity backstop running six years.
- Bull
- If neocloud demand persists, CoreWeave's OpenAI/Microsoft/Nvidia contracts and Nvidia's $6.3B take-or-pay backstop convert a $99.4B backlog into recurring revenue.
- Bear
- Customer A was 45% of Q1-26 revenue and Nvidia is both equity holder and backstop, so concentration and circular vendor financing inflate apparent demand.
Alphabet Inc.
Alphabet holds roughly 14% of Anthropic in straight equity, contractually hard-capped at 15% per court documents, and committed billions more in 2026. Fortune reported that about half of Alphabet's and Amazon's 2026 'AI profits' came from markups on their Anthropic stakes rather than operating AI revenue.
- Bull
- If Anthropic IPOs near base case, Alphabet's ~14% capped equity stake plus internalizing Ironwood TPU give it both proxy upside and merchant-layer disintermediation.
- Bear
- About half of recent 'AI profits' were Anthropic stake markups, not operating revenue, and the lab stake is a small slice of a vast enterprise value.
Microsoft Corporation
Guided to ~$190B capex for calendar 2026 (confirmed on its fiscal-Q3 call, Apr 29, 2026; ~$90B in 2025) — roughly tied with Alphabet at the top of hyperscaler spend. Builds the in-house Maia AI accelerator but remains heavily reliant on NVIDIA GPUs; its deepest AI exposure is the OpenAI relationship, not owned merchant silicon.
- Bull
- If OpenAI lists near base case, Microsoft's ~27% as-converted stake plus ~$190B 2026 capex give it the deepest proxy exposure to the leading private lab.
- Bear
- The OpenAI line is a loss-share, not revenue; the recap gain was one-time accounting, and Microsoft's revenue cut is capped at $38B through 2030.
OpenAI Group PBC
Reorganized into OpenAI Group PBC in October 2025; raised at an $852B post-money valuation by March 2026 and confidentially filed an IPO S-1 around June 8, 2026. The company has not committed to listing timing (it said "it may be a while"); sources put a possible window around Sep–Q4 2026 at an implied $852B–$1T. Verified economic holders include Microsoft (~27% as-converted), SoftBank (~11%) and Nvidia (~$30B equity).
- Bull
- If it lists near the implied $852B-$1T window, OpenAI's June 2026 S-1 marks the supply chain's largest demand engine into tradable equity, repricing MSFT, SFTBY and VCX.
- Bear
- Timing is uncommitted ('it may be a while'), and public investors reach it only via MSFT, SFTBY or the premium-prone VCX fund, never a direct line.
SoftBank Group Corp. (unsponsored ADR)
OTC-traded unsponsored ADR (each represents 0.50 ordinary shares) of SoftBank Group, which reached about 11% ownership of OpenAI after funding a $40B commitment and anchors the ~$500B Stargate datacenter program. Management has signaled it may borrow against its OpenAI shares to finance Stargate and robotics expansion.
- Bull
- If OpenAI lists strongly, SoftBank's ~11% stake and anchor role in the ~$500B Stargate buildout give leveraged proxy exposure to the leading private lab.
- Bear
- It is an unsponsored OTC ADR (0.50 shares each) of a holding company planning to borrow against its OpenAI shares, adding leverage and wrapper risk.
The Fundrise Innovation Fund
NYSE-listed, non-diversified closed-end fund that began public trading in March 2026. Anthropic is reported at roughly 21% of the portfolio; OpenAI has been a core holding since 2023, and the fund participated in OpenAI's $122B round in 2026. Top-three holdings (Anthropic, Databricks, OpenAI) were reported near 48% of assets; shares have traded far above the disclosed ~$18.97 NAV.
- Bull
- If Anthropic and OpenAI IPO strongly, VCX's ~21% Anthropic weighting and core OpenAI holding give listed access to the otherwise-untradable private labs.
- Bear
- Shares trade far above the ~$18.97 NAV, so this closed-end fund can fall even if the underlying marks rise — the wrapper's distortion is the risk.
Catalyst calendar
- 2026-07-28Microsoft FY2026 Q4 earnings (after close)mediumUpdates the OpenAI equity-method mark and revenue-share-cap commentary, the largest disclosed listed window into OpenAI's economics.
- 2026-08-07SoftBank Group Q1 FY2026 results (quarter ended June 2026)mediumMarks SoftBank's ~11% OpenAI stake and Stargate funding posture, the cleanest non-US listed read on OpenAI exposure and AI-buildout leverage.
- 2026-09-30OpenAI IPO listing window opens (Q4 2026 target)mediumOpenAI confidentially filed its S-1 around June 8, 2026 but has explicitly not committed to timing. A listing — whenever it lands — would be the largest tech IPO ever and reprice every private-AI exposure wrapper. The date here is a speculative placeholder, not a confirmed event.
- 2026-10-01Anthropic IPO pricing window (later-2026 target after June 1 S-1)mediumAnthropic filed confidentially around June 1, 2026; trade-press expectation is a later-2026 listing, but no date is set. Pricing would mark the second mega-cap AI pure-play into public markets. The date here is a speculative placeholder.