AI, MANBRIC/Biotech, Robotics & Smart Cities: the physical capex supercycle

The “AI bubble” debate is secondary; what is observable is the largest private investment cycle in modern history, whose spillover effects ripple across energy, industrial real estate, semiconductors, robotics, and cities. The supercycle figures. The five major hyperscalers (Microsoft, Alphabet, Amazon, Meta, Oracle) will commit $660-725 billion in capex by 2026, nearly double that of 2025; Morgan Stanley projects over $800 billion annually for all major cloud providers; Gartner: global IT spending will reach $6.37 trillion in 2026 (+14.2%), with data centers growing +62.5% to $822 billion; IDC: AI infrastructure ~$497 billion in 2026, exceeding $1 trillion in 2029. Physical bottlenecks: electricity (data centers 415 TWh in 2024 → ~945 TWh by 2030, ~3% of global consumption; +130% US demand by 2030), memory (DRAM +158% in 2026), and installed capacity (23 GW under construction; 92 GW data center demand by 2027 according to Goldman Sachs). A maturity signal: revenues are following — OpenAI ~$20 billion ARR by end of 2025, Anthropic >$9 billion run-rate in January 2026 — but remain a fraction of capex: the revenue/capex ratio is the key risk variable of this cycle.

MANBRIC/Biotech. The AI-biology convergence (accelerated drug discovery, bioproduction, multi-omic data) is the secondary beneficiary of compute capex: the same GPUs training LLMs reduce the marginal cost of molecular simulation and screening. The 2026-2035 investment thesis: pharma margins will shift toward AI-native discovery platforms and proprietary data (biobanks, cohorts) — the rare asset is not the model, but exclusive biological data. Robotics. Embodied AI (humanoid robots, logistic AMRs) is entering an industrial deployment phase, driven by declining inference costs and labor shortages. Key business model: Robotics-as-a-Service (RaaS) — robots are no longer sold but rented by the hour, transforming robotics into a rent-bearing asset… tokenizable (cf. Idea 3).

Smart cities & digital twins. Smart city market: $2.8 trillion (2026) → $15.9 trillion (2035), CAGR ~21.6%; digital twins: $33.97 billion in 2026 → $384.79 billion projected by 2034. References: Virtual Singapore ($73 million), NEOM ($1.8 billion), 28 Chinese municipal pilots (48 billion yuan), UK transport twin (£30 million), NYC 3D Underground ($10 million). Measured results: -19% district energy in Helsinki (Kalasatama), -14% congestion in Dubai. The bottleneck is not technological but semantic: 59% of experts cite the lack of a shared definition of “digital twin” as an obstacle.

OVERWEIGHT the physical value chain of AI (energy, cooling, memory, industrial real estate) and biotech platforms with exclusive data; the infrastructure compute rent is the emerging asset class of the decade.