AI 研究的钱从哪来？Who Pays for AI Research?

Enormous capital sits behind every breakthrough AI model. Trace the money and you find Big Tech, venture capitalists and governments bankrolling the AI revolution through 2026.

A model that writes code, diagnoses illness or turns out photorealistic video invites admiration for the software. The genuine marvel, and the genuine constraint, is the money needed to produce it. Never before in the history of technology has wealth and investment been concentrated the way it is in AI right now. So who pays for AI research? Whose names sit on the checks covering the billions of dollars it takes to train these digital brains?

What answers it is a tangled ecosystem: tech giants, venture capitalists willing to take big risks, government agencies with strategic goals, and open-source communities operating from the ground up. Today the money is what we follow, in order to make sense of the economics driving the AI revolution.

01Just How Big the Investment Numbers Get

Grasping AI funding means first getting your head around the magnitudes involved. Global investment in AI across 2025 and 2026 climbed past the peak reached during the dot-com bubble. The sums are not in the millions of dollars — they run into the hundreds of billions.

Take the "Stargate" project, a huge joint venture that major tech leaders announced with the aim of building out AGI infrastructure. The opening commitment: $500 billion spread over four years. For scale, that figure exceeds the GDP of many European nations, and every dollar of it is aimed at server farms and microchips.

A winner-takes-all mindset is what sits behind spending at this pitch. Whoever builds the smartest models ends up holding the underlying infrastructure for computing's next hundred years. The question that follows, though, is a simple one: whose money is it?

02Big Tech: The Biggest Spenders by Far

The legacy giants — Microsoft, Alphabet (Google), Meta and Amazon — hold the title of AI's biggest funders without much argument. Their cash reserves stand comparison with national treasuries, and they are not sitting on them.

What Motivates the Spending

What justifies outlays this large? Selling a chatbot is a small part of it. AI has become the new cloud. OpenAI is bankrolled by Microsoft so that enterprises on Azure end up running Microsoft AI. Google puts billions behind DeepMind and Gemini to defend its search business. Meta channels money into Llama to keep its social platforms engaging and ad-relevant.

Almost every new AI research breakthrough you read about traces back to the compute infrastructure and the multi-billion R&D budgets of one of these giants. Losing money on AI for a decade is survivable for them, because what they are buying with that patience is total market dominance.

03Venture Capital: The Fuel for Risky Startups

Big Tech's walled gardens are not where all AI gets built. A lively startup ecosystem — Anthropic, Mistral and Cohere among them — runs on Venture Capital (VC) money. VCs supply the AI boom with its high-risk, high-reward engine.

Billions are flowing from venture capitalists into startups racing to settle what AGI is and whether it has been achieved. Big Tech can cushion AI work with ad revenue or cloud sales; startups cannot, and they burn cash at a rate that alarms observers. Server time and researcher pay alone can drain $50 million a month out of a single startup.

04Governments and Defense: Investing for Strategy

Market share is the tech industry's objective; national security is the government's. Foundational AI research therefore draws enormous sums from public bodies: in the US there is DARPA (Defense Advanced Research Projects Agency) and the NSF (National Science Foundation), while Europe has the European Union's Horizon programs.

Basic science that pays no immediate dividend yet matters strategically is a particular favorite of government funders. Defense agencies, for instance, put substantial money into work on what reasoning AI is and how it functions, with applications spanning autonomous logistics, cybersecurity and intelligence analysis.

05Open Source and Community: Funding From the Ground Up

Suits and officials are not the only source of AI money. A huge, decentralized funding engine exists in the open-source AI community. Hugging Face functions as something like an AI-era GitHub, hosting thousands of models kept alive by corporate sponsorships and community donations in combination.

Individual researchers and small labs frequently depend on GitHub Sponsors, Patreon and non-profit grants to get their findings published. Because these community-funded releases arrive so quickly, plenty of developers keep an eye on the AI research coming out this week to spot new open models that can stand up to the closed-source giants.

06Following the Dollars: Where They Land

Raise $1 billion for AI and what happens to it? Programmer laptops account for almost none of it. Energy and silicon are what define the economics of this field.

Hit the Wall: Data Shortages and Power Limits

Financially, the "Data Wall" is the hardest problem at the moment. Human data is what models learn from, and supplies of high-quality books, articles and code are running thin fast. Licensing fees paid to news publishers and academic journals now reach into the millions. Energy is the other squeeze: training runs need so much power that tech companies are outright buying nuclear plants and putting money into next-generation geothermal projects just to keep data centers humming.

07Questions Our Readers Send In