a16z Co-founder Marc Andreessen: Free AI Capabilities Will Bring Unprecedented Consumer Surplus
Marc Andreessen, co-founder of a16z, quoted a "sidebar remark" from Silicon Valley, stating, "I am extremely bullish on AI capabilities, so consumer surplus will be ridiculously high—like giving away the power of gods at below cost, leading to a long-term bullish market overall." He described AI as a super technology dividend that is "sold at far below its true value." This viewpoint continues his core narrative in podcasts, articles, and interviews: AI is rapidly lowering the "unit intelligence cost," with a divergence between model capability curves and price curves, resulting in ordinary users receiving benefits far exceeding the revenue scale of related companies. Thus, AI becomes a new engine for economic expansion and asset revaluation.
Multiple studies from the Stanford Digital Economy Lab, WSJ, and economists provide supporting evidence for this claim: Estimates based on large sample user surveys show that U.S. adults have achieved an annual consumer surplus of hundreds of billions of dollars from generative AI chat tools like ChatGPT, Gemini, Claude, and Copilot, far exceeding the generative AI revenues disclosed by companies like OpenAI, Anthropic, Google, and Microsoft during the same period. This indicates that most benefits are captured by users who "use for free or at low prices," rather than reflected in corporate revenues and GDP statistics. Some analyses liken this to the early internet—where large-scale free services were statistically underestimated but continued to change the economic structure in terms of behavior and productivity. Andreessen extrapolates this model to a new stage where "intelligence itself is distributed cheaply."
Source: Public Information
ABAB AI Insight
This statement's key point is defining the economic size of the AI cycle using "consumer surplus" rather than "company profits." Traditional tech bull markets often revolve around revenue and profit growth, while in the AI narrative, Andreessen and a group of digital economists emphasize that most of the value created by AI will be delivered to end users in the form of "free or very low-priced tools"—users are willing to pay far above current prices for these tools, but in reality, they are rarely charged, leading to a significant underestimation of AI's real impact on macro statistics (GDP, corporate revenue). This structure is highly similar to internet services like search, email, and maps, with the "underestimated item" shifting from information distribution to orchestrated intelligent labor.
The phrase "the power of gods given away below cost" views AI as a form of "intelligent infrastructure" that is subsidized in the short term and may have marginal costs approaching zero in the long term. With rapidly declining costs for model training and inference, and open-source models continuously approaching closed-source levels, the marginal cost per invocation is being compressed to nearly the "penny-level computing price" seen in cloud computing, while users perceive it as "a high-level assistant available 24/7." When this high-value—low-price divergence continues, the real restructuring is not just within a single industry, but rather "who can still profit from selling labor and time differences": many intermediate services primarily selling "intellectual time" (low-end consulting, basic content production, some back-office positions) will be forced to align their prices with AI, while end consumers enjoy a substantial increase in purchasing power without any change in nominal income.
From an asset pricing perspective, the logic of "Long the market" views AI as a broad positive externality rather than a single sector opportunity: if consumer surplus far exceeds AI company revenues, then the most direct beneficiaries are not the model companies themselves, but all entities that can translate AI into reduced costs, improved efficiency, and enhanced pricing power—ranging from office software to e-commerce, from financial services to industrial enterprises. Those who can embed "low-cost intelligence" into their processes will be able to generate higher cash flows with the same capital and labor. In this framework, "bullish on the market" means betting on the entire profit base being lifted by AI, rather than just betting on the valuations of a few pure-play AI companies continuing to rise.
However, this optimistic narrative also implies a set of assumptions regarding distribution and power structures: a large consumer surplus does not automatically translate into a safe landing for all workers and small businesses. AI providing "god-like capabilities" at low prices will compress the pricing power of many intermediary labor positions, intensifying price competition for mid-skill jobs; meanwhile, the entities that can truly integrate these capabilities into large-scale processes are the leading companies that control capital, data, and brands. The result may be a scenario where, on a macro level, "consumer welfare skyrockets, and the stock market benefits overall," while on the other hand, income and bargaining power further concentrate towards "AI + capital-intensive platforms." Andreessen's choice to use "consumer surplus" and "long the market" to tell the story emphasizes total volume and indices; structurally, what AI brings is likely a new normal of "a larger total pie, stronger leaders, and a squeezed middle."