LAUNCH Founder Jason Calacanis: Smart Commoditization Will Drive Intelligent Inclusivity
Jason Calacanis suggested that once intelligent capabilities transition from scarce expert services to standardized, interchangeable, on-demand commodities, their prices will decrease, and the barriers to access will also lower; here, inclusivity refers to the diffusion of usage rights, not the dispersion of model ownership.
The inference cost of models reaching GPT-3.5 level has dropped from about $20 per million tokens to $0.07, a reduction of over 280 times in about 18 months. The performance gap among leading models continues to narrow, with competitive variables shifting from "who is smarter" to price, reliability, and effectiveness in vertical scenarios.
Buyers include enterprises, developers, and individual users, while sellers are model companies and cloud service providers; funding is not flowing in due to a single event but is shifting from labor costs and traditional software licensing to token usage, GPU computing power, and data services. Application developers and users benefit, while the premium on general models is under pressure, and companies that control computing power, proprietary data, and distribution channels gain more bargaining power.
Generative AI is expected to cover about 53% of the population within three years, with 88% of surveyed institutions having adopted AI, and the estimated annual surplus value for U.S. consumers reaching $172 billion; however, the U.S. has 5,427 data centers, and advanced AI chip manufacturing is highly dependent on TSMC, with the diffusion on the usage side and concentration on the supply side occurring simultaneously.
ABAB AI Insight
Jason Calacanis has founded Weblogs, Inc. and LAUNCH, transitioning from content platform entrepreneurship to early-stage investment. LAUNCH's public investment portfolio includes companies like Uber, Robinhood, and Calm, supporting over 1,000 startups; his long-term path has been to lower the barriers for entrepreneurs to access capital, media, and distribution resources.
Smart commoditization has changed the funding pathway: startups no longer need to build their own model teams but can purchase APIs, rent cloud GPUs, and call upon open-weight models, transforming fixed R&D costs into variable costs. Meanwhile, private AI investment in the U.S. has reached $285.9 billion, 23 times that of China's private investment; the lowered usage threshold has increased the number of startups but has allowed underlying capital to remain concentrated in model labs, cloud platforms, and chip supply chains.
AWS turned servers from fixed corporate assets into pay-as-you-go services in 2006, lowering the threshold for internet entrepreneurship but concentrating infrastructure profits in cloud platforms; Linux provided a free open operating system, while Red Hat established commercial value through enterprise services. AI is repeating this path: foundational intelligence is entering a standardization phase, and industry competition is shifting from training general models to controlling workflows, customer relationships, and proprietary data.
Essentially, this is a transfer of pricing power: as model capabilities converge, model vendors can no longer rely on being "smarter" to maintain high prices, and pricing power will shift to GPU capacity, energy, data, and user access. Intelligent inclusivity disperses the ability to call upon intelligence while concentrating the capital and infrastructure needed to produce intelligence.
ABAB News · Cognitive Laws
The more widespread the capability, the more scarcity migrates to the entry point.
As tools approach free, those who control the tools become more expensive.
Inclusivity disperses usage rights while scale concentrates ownership.