# Who is actually paying for the AI boom? Date: 2026-10-31 · Author: Robert Scales · https://guidetoartificialintelligence.com/en/learn/who-funds-ai-map/ **Answer:** Six groups pay for most of it: cloud and chip companies investing in their own customers, venture and growth funds, sovereign wealth funds, public stock markets, bond and private credit investors, and governments. The biggest single source remains the large tech companies' own cash flow, though that is no longer enough, which is why debt and share sales have grown so quickly in 2026. AI funding headlines tend to merge very different kinds of money. A pension fund buying a bond, a chipmaker buying stock in its customer and a government buying a stake in a national champion all count as "investment in AI," but each has different motives and different risks. Here is a map of the money in late 2026, using filings and company announcements where possible. Figures are in US dollars unless noted. ## The map | Who | What they want | Examples | |---|---|---| | Tech giants, from their own cash | Capacity for their own products and clouds | About $720–740B of [planned capex](/en/learn/ai-capex-explained/) at four companies in 2026 | | Suppliers and clouds, as investors | Secure customers and share in their growth | Amazon, $50B in OpenAI; Nvidia, $99B of equity holdings in total | | Venture and growth funds | A financial return | Altimeter, Dragoneer, Greenoaks and Sequoia leading Anthropic's Series H | | Sovereign wealth funds | Return, and a strategic position | MGX, GIC, Temasek; UAE investors in OpenAI's current talks | | Public markets | A financial return, with shares that can be sold | SpaceX, about $86B; Cerebras, $5.55B; Alphabet's share sales | | Bonds and private credit | Steady interest payments | PIMCO in Meta's Hyperion financing; Meta's $25B of notes | | Governments | Sovereignty, jobs, research | Canada's stake in Cohere | A sovereign wealth fund is a state-owned investment fund, often built from oil revenue or trade surpluses. MGX is based in Abu Dhabi, GIC and Temasek in Singapore. ## Private rounds keep growing A "round" is a single fundraising, usually labelled Series A, B and so on. Valuation is the price the round implies for the whole company. "Post-money" includes the new cash, and "pre-money" does not. | Date | Company | Amount | Post-money valuation | |---|---|---|---| | Mar 2025 | OpenAI | $40B | $300B | | Sept 2025 | Mistral AI | €1.7B | €11.7B | | Feb 2026 | Anthropic | $30B | $380B | | Mar 2026 | OpenAI | $122B committed | $852B | | May 2026 | Anthropic | $65B | $965B | | Oct 2026 | Cohere | Up to C$7B target | About C$36B | | Oct 2026 (in talks) | OpenAI | Up to $20B reported | $1.25T reported by the FT | Two details matter. "Committed" is not the same as paid: Amazon's filing shows its $50 billion arriving in stages through the third quarter. And headline totals can include old money: Anthropic says its $65 billion includes $15 billion committed earlier by large investors. Its filing shows about $45.7 billion of sales in the main class of Series H stock. The OpenAI talks show how quickly these things move. CNBC reported up to $30 billion on October 22, then confirmed on October 31 that the goal was down to about $20 billion. The round has not closed, and the terms could change. Ordinary people are starting to appear in this list too. CNBC reports that about $3 billion of OpenAI's March round came from individual investors through bank channels. ## Public markets opened in 2026 | Date | Company | Exchange | Raised | |---|---|---|---| | Mar 2025 | CoreWeave (GPU cloud) | Nasdaq | About $1.5B | | Jan 2026 | Zhipu (Z.ai) | Hong Kong | About HK$4.35B | | Jan 2026 | MiniMax | Hong Kong | About HK$4.82B | | May 2026 | Cerebras (AI chips) | Nasdaq | About $5.55B | | June 2026 | SpaceX, with xAI | Nasdaq | $86.25B gross, about $85.7B net | Anthropic has filed publicly but has not priced its shares, and OpenAI has made a confidential filing. Established companies can tap the market too. Alphabet raised about $50 billion, net, in June 2026, including a $10 billion private sale to a Berkshire Hathaway affiliate. Not every listing goes smoothly. CoreWeave priced below its planned range, and CNBC reports that Firmus, a GPU cloud backed by Nvidia, withdrew its IPO plans in October. ## Debt is the fast-growing piece Debt means the money must be paid back on a schedule, whatever happens to the business. The Bank of England says the five largest AI hyperscalers held just 3% of outstanding US investment-grade debt at the end of 2025, yet accounted for over 15% of new issuance in 2026 by early May. "Investment grade" means a credit rating that marks the borrower as relatively safe. Debt also reaches people who never chose an AI investment. Pension funds and bond funds can hold it. Private credit, meaning loans from funds rather than banks, is a growing lender too. For the structures involved, see [our explainer on SPVs](/en/learn/circular-financing-spv-ai-explained/). Even strong companies use it: Meta sold notes for $24.9 billion in net proceeds in May 2026. Weaker borrowers pay more. CNBC reports that neoclouds, smaller GPU-rental firms without investment-grade ratings, are finding lenders more cautious. ## Governments and national strategies Public money is small next to the private totals, but it comes with different goals. Canada's commitment of up to C$1.6 billion to Cohere buys shares along with promises about jobs and Canadian computing capacity. Canadian investors are also present in private rounds; Bell's venture arm and Inovia Capital joined Cohere's Series E. Other governments are pursuing their own strategies, in some cases mainly through their sovereign funds. Germany's Schwarz Group, which leads the Cohere round, is a further category of its own: a retail group investing in AI for strategic reasons. > **Try it:** Take the next funding headline you read and place it on the map above. Ask who the money belongs to, what they expect in return and whether it arrives all at once. ## Whose money is at risk if the boom slows? It depends on the position in the queue. If a company gets into difficulty, lenders are generally paid before shareholders, and shareholders absorb losses first. So venture funds and public shareholders carry the most immediate price risk, including workers holding company stock and anyone whose index fund holds the big tech names. Bondholders are safer as long as the borrower can pay. Their risk rises where debt is tied to a single data centre that depends on a single tenant. Suppliers carry a double exposure: they can lose a customer and the value of their stake at the same moment. The IMF's latest stress test found banks broadly resilient under an AI slowdown, with weak spots. Taxpayers carry a smaller share, through direct stakes such as Canada's in Cohere, and through any public response if a downturn spreads. This map describes who is exposed. It makes no recommendation about any investment. ## Sources - Anthropic raises $65B Series H at $965B valuation: https://www.anthropic.com/news/series-h - OpenAI closes funding round at an $852 billion valuation: https://www.cnbc.com/2026/03/31/openai-funding-round-ipo.html - Financial Stability Report, July 2026: https://www.bankofengland.co.uk/financial-stability-report/2026/july-2026 - Space Exploration Technologies Corp. Form 10-Q, quarter ended June 30, 2026: https://www.sec.gov/Archives/edgar/data/0001181412/000162828026052535/spcx-20260630.htm --- From the Guide to Artificial Intelligence, by Robert Scales, published by Holistic Hybrid Inc. Share and adapt it under CC BY-NC-SA 4.0 (https://creativecommons.org/licenses/by-nc-sa/4.0/): credit the guide, link to https://guidetoartificialintelligence.com/en/learn/who-funds-ai-map/, no commercial use without permission, and keep the same licence.