Key Takeaways
- • Global venture capital totaled about $300 billion in Q1 2026, up more than 150% year-on-year and quarter-on-quarter, with AI companies receiving roughly $242 billion (about 80%), according to Crunchbase data cited in the report.
- • Epoch AI estimates inference demand is growing about 10-fold annually at current prices versus roughly 3.4-fold annual growth in inference supply, leaving supply, not demand, as the industry's main constraint.
- • Advanced packaging capacity is effectively sold out through the end of 2027, and the most advanced high-bandwidth memory is fully allocated for 2026 and largely booked for 2027.
- • Robotics and physical AI carry the highest venture momentum index at 1.73, while AI applications and agents account for 25% of investor attention, roughly twice the second-ranked category.
- • Anthropic faces valuation pressure ahead of its IPO, with secondary-market quotes below $1.7 trillion and some buy-side institutions saying pricing above $1.5 trillion may struggle to attract demand.
Summary
Hong Hao, managing partner at Lianhe Asset Management, and Lars Tvede, author of "The Business Cycle" and "Supertrends," jointly released a report titled "Where Has the Smart Money Gone? An Economic Map of the Global Artificial Intelligence Ecosystem," tracking capital flows through the AI supply chain. The report concludes that supply rather than demand is the main constraint on the industry, and that profit pools are migrating from upstream compute toward downstream applications and scarce physical resources.
Citing Crunchbase statistics, the report says global venture capital reached about $300 billion in Q1 2026, with AI companies taking about $242 billion. OpenAI completed a funding round with total committed capital of $122 billion, while Anthropic and xAI closed $30 billion Series G and $20 billion Series E rounds, respectively. Separately, a TrendForce August forecast projected that capital expenditure by the nine largest cloud service providers will grow about 90% year-on-year in 2026 to more than $886.7 billion. Goldman Sachs estimates about $7.6 trillion could be invested globally in AI infrastructure cumulatively from 2026 to 2031.
The report identifies interconnected bottlenecks spanning GPUs, high-bandwidth memory, advanced packaging, power, transformers, grid connections, networking and cooling. It divides the profit pool migration into five stages: compute scarcity (from 2026), intelligence commercialization (around 2028), embodied intelligence (around 2029), autonomous enterprises (around 2030) and ubiquitous intelligence (around 2035). Roughly half of tracked venture signals are directly or indirectly AI-driven, and robotics and physical AI show the strongest momentum. The report also states that AI applications and agents account for 25% of investor attention, AI infrastructure 12% and stable power 11%.
The report argues that as intelligence becomes abundant, competitive advantage will shift toward proprietary data, customer relationships, workflows, distribution, brands and ecosystems. Citing William Nordhaus research, it notes innovators captured only about 2.2% of U.S. nonfarm business social surplus from 1948 to 2001. It also flags an "impossible trinity" of margins across model companies, compute suppliers and end customers.
Cited reporting from 36Kr says Anthropic faces valuation pressure ahead of its IPO roadshow, with secondary quotes below $1.7 trillion. The company reported Q3 revenue of $16.8 billion and adjusted profitability for the first time, though sequential growth slowed to 45% from 147%. The report notes that Anthropic's gross-basis revenue recognition versus OpenAI's net basis can create a 30% to 40% reported revenue gap, and that Anthropic has committed at least $518 billion in cloud and compute offtake over the next decade, about 80% non-cancelable. CoreWeave shares fell about 25% over one week, with some long-term bond yields near 15%.
Citing Crunchbase statistics, the report says global venture capital reached about $300 billion in Q1 2026, with AI companies taking about $242 billion. OpenAI completed a funding round with total committed capital of $122 billion, while Anthropic and xAI closed $30 billion Series G and $20 billion Series E rounds, respectively. Separately, a TrendForce August forecast projected that capital expenditure by the nine largest cloud service providers will grow about 90% year-on-year in 2026 to more than $886.7 billion. Goldman Sachs estimates about $7.6 trillion could be invested globally in AI infrastructure cumulatively from 2026 to 2031.
The report identifies interconnected bottlenecks spanning GPUs, high-bandwidth memory, advanced packaging, power, transformers, grid connections, networking and cooling. It divides the profit pool migration into five stages: compute scarcity (from 2026), intelligence commercialization (around 2028), embodied intelligence (around 2029), autonomous enterprises (around 2030) and ubiquitous intelligence (around 2035). Roughly half of tracked venture signals are directly or indirectly AI-driven, and robotics and physical AI show the strongest momentum. The report also states that AI applications and agents account for 25% of investor attention, AI infrastructure 12% and stable power 11%.
The report argues that as intelligence becomes abundant, competitive advantage will shift toward proprietary data, customer relationships, workflows, distribution, brands and ecosystems. Citing William Nordhaus research, it notes innovators captured only about 2.2% of U.S. nonfarm business social surplus from 1948 to 2001. It also flags an "impossible trinity" of margins across model companies, compute suppliers and end customers.
Cited reporting from 36Kr says Anthropic faces valuation pressure ahead of its IPO roadshow, with secondary quotes below $1.7 trillion. The company reported Q3 revenue of $16.8 billion and adjusted profitability for the first time, though sequential growth slowed to 45% from 147%. The report notes that Anthropic's gross-basis revenue recognition versus OpenAI's net basis can create a 30% to 40% reported revenue gap, and that Anthropic has committed at least $518 billion in cloud and compute offtake over the next decade, about 80% non-cancelable. CoreWeave shares fell about 25% over one week, with some long-term bond yields near 15%.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.