Mary Meeker's new report, "Trends — Artificial Intelligence," highlights the unprecedented pace of AI development and adoption, noting rapid user growth (ChatGPT reaching 800 million users in 17 months), decreasing usage costs (inference costs dropping 99% in two years), and intense competition driving feature parity. Despite massive VC investment and infrastructure build-out, the report cautions that the long-term profitability and market dominance of current AI companies remain uncertain, as significant cash burn persists across the sector.
Mary Meeker's new "Trends — Artificial Intelligence" report emphasizes the unprecedented speed of AI development, adoption, and associated spending, supported by extensive data. The report highlights ChatGPT's attainment of 800 million users in just 17 months and a dramatic 99% reduction in inference costs per 1 million tokens over two years, according to Stanford research, alongside the high training costs for models, which can reach up to $1 billion. Intense competition is driving rapid feature parity, with open-source options, particularly from Chinese developers, gaining traction. Major technology companies are making foundational investments in AI infrastructure; for instance, Nvidia's 2024 Blackwell GPU is reported to use 105,000 times less energy per token than its 2014 Kepler predecessor, while Google's TPUs and Amazon's Trainium chips are being developed at scale. Despite the rapid technological advancements and user adoption, which benefit consumers and enterprises through improved capabilities and lower costs, the report cautions that AI companies and cloud service providers are experiencing significant cash burn due to massive infrastructure investments. Consequently, the long-term profitability and identification of enduring market leaders among the current AI companies remain uncertain.
AI-powered research, real-time alerts, and portfolio analytics for institutional investors.
Request a DemoOverall Sentiment
moderately positive
Sentiment Score
0.50
Ticker Sentiment