According to ValueQuest Investment Advisors, the world may be entering a 'capital allocation supercycle' - an extended period of elevated investment in physical infrastructure across multiple industries and geographies. As reported by CNBC TV18, this supercycle could extend for decades rather than the typical few-year business cycles, driven by structural changes in the global economy. The current cycle is unusual because several long-term investment themes are unfolding simultaneously, potentially reinforcing each other rather than competing for capital. The first phase of the AI boom has funnelled an extraordinary amount of capital into a handful of US mega-caps, with NVIDIA, Microsoft and Alphabet dominating buy-lists and performance tables. However, UBS estimates that hyperscalers will spend the equivalent of 100% of operating cashflow in 2026, up from a 10-year average of 40%, meaning those supposedly asset-light companies are starting to look distinctly more capital-intensive.
The memory chip market is experiencing a supercycle driven by AI-optimised memory demand, with Bank of America estimating that AI-optimised memory requires three to four times the production capacity of conventional memory per unit. Following Micron's latest results, BofA has pushed its estimated timeline for the super-cycle out to the end of 2027, with a scenario extending as far as 2030. Multiple data points support this extended timeline: Micron's HBM capacity is sold out through 2027, while Kioxia confirmed in January that its entire 2026 NAND output had already been committed, with some hyperscale clients requesting supply agreements stretching into 2027 and 2028. SK Hynix's chairman has warned that global memory supply is likely to remain roughly 20% below demand through 2030. The industry's underlying cyclicality remains intact, but the mechanism has changed - long-term agreements now lock in pricing and allocation years in advance rather than quarter to quarter, giving buyers and sellers more visibility than in past cycles.
AI-driven demand has transformed memory chips into one of the market's strongest investment themes, with memory stocks delivering extraordinary returns ranging from 176% to over 4,000% in the past year. According to The Financial Express, Micron and SK Hynix have each delivered returns of more than 600% over the past year and have both crossed $1 trillion in market cap. SanDisk has returned roughly 4,000% since spinning off from Western Digital in February 2025, going from about $38.50 a share to north of $2,000. Seagate is up in the region of 400% in the last year, driven by demand for its enterprise hard drives. The Roundhill Memory ETF (DRAM) is up 176.3% year to date, providing broad exposure to the memory boom. As recently as 2022, data centres accounted for somewhere between 20% and 30% of global DRAM consumption, but IDC now expects that figure to hit roughly 70% in 2026, with TrendForce putting AI's specific share closer to 20%.
According to Ninety One's latest analysis, power is emerging as a key bottleneck for the AI/datacentre industry, with the hyperscalers increasingly turning to natural gas as a solution. Industry leaders at J.P. Morgan's Energy, Power, Renewables & Mining Conference in New York were surprised that the oil price has declined so rapidly following the biggest oil outage in history. This has been attributed to the Trump administration talking up peace hopes, rapid inventory draws in the US and China, and Chinese state-sponsored demand rationing. The gas-to-power (using natural gas to generate electricity) for data centres was a recurring theme, with Chevron signing a deal with Microsoft to power one of the largest US data centres in the Permian Basin to take advantage of cheap natural gas prices. Energy infrastructure company Williams appears to be making good progress on its behind-the-meter power solutions, while Baker Hughes is gaining momentum in the gas turbine market.
For nearly three decades after the 1990s, businesses focused on efficiency with global supply chains designed to minimize costs and manufacturing shifted to lowest-cost locations. However, according to ValueQuest's Ravi Dharamshi, recent global shocks have exposed these vulnerabilities. COVID-19 disrupted global supply chains, while the Russia-Ukraine conflict highlighted risks of relying on imported energy. Shipping disruptions and semiconductor shortages demonstrated how single bottlenecks could ripple across the global economy. These events have prompted governments and corporations to rethink decades of globalisation, focusing instead on building resilience, redundancy, energy security, and supply-chain security. The second wave of AI investment is taking place well beyond the mega-cap software names, with the International Energy Agency forecasting that data centre electricity demand will more than double by 2030, which exceeds Japan's current consumption.
While much of the current investment narrative revolves around AI, Dharamshi argues that focusing solely on AI misses the broader picture. According to CNBC TV18, AI should be viewed as one catalyst within a much larger industrial investment cycle rather than the cycle itself. The launch of ChatGPT in late 2022 demonstrated commercial potential, accelerating demand for computing infrastructure. However, deploying AI requires substantial investments in power generation, electricity transmission, cooling systems, networking equipment, industrial automation and advanced manufacturing. Dharamshi believes AI has accelerated investments that governments and companies were already compelled to make for infrastructure strengthening. Data centres have become the engine rooms of the AI economy, with McKinsey estimating that data centre capex could reach almost $7trn by 2030, including enough fibre-optic cable alone to circle the Earth 120 times. Data centres consume vast amounts of energy, with the cooling needed to manage heat-intensive AI chips often rivalling the power needed for the computing itself.
