The $1 trillion AI 'arms race' is fueling a hidden supercycle and it's not just chips

The $1 trillion-plus annual AI capex “arms race” is quietly powering a broad capex supercycle that reaches far beyond chips, into energy, infrastructure, networking hardware and new “safe haven” asset classes, according to DBS Bank.

Global markets are being reshaped by two investment turbines: massive AI capex and a parallel energy capex build-out. DBS Bank notes that combined capex at Alphabet, Amazon, Meta and Microsoft has surged about 200% since the launch of ChatGPT, with 2026 guidance lifting to around $725 billion, nearly double 2025 levels. On top of commercial cloud demand, governments are racing to build “sovereign AI” as they treat AI infrastructure and semiconductor capability as strategic national assets, with McKinsey estimating a $500–600 billion total addressable market for sovereign AI by 2030.

Chief Investment Officer Hou Wey Fook said the world is moving “from a savings glut regime to one of massive capex boom,” with AI optimism pushing equity investors “off to the races” even as inflation risks are under‑priced. In his words, “the AI capex cycle had room to run and with it, driving corporate earnings higher,” but investors must “watch your downside” as markets price an orderly end to the Iran war and weaker oil, while risks point the other way.


Also Read | The AI boom won't burst all at once. It will pop in 'rolling bubbles': Macquarie

Beyond Chips: The New ‘Pick‑and‑Shovel’ Winners

Far from being just a semiconductor story, the report by DBS frames the AI build‑out as a “pick and shovel” opportunity across hardware, infrastructure and energy services. Semiconductor manufacturers and designers have become “the primary gatekeeper of the AI era,” capturing roughly 60% of total data‑centre capex as hyperscalers race to assemble massive compute clusters, according to McKinsey quoted in the report.

World Semiconductor Trade Statistics data cited by DBS shows the global semiconductor market on track to reach $975 billion in 2026, with memory expected to grow about 39% as high‑bandwidth memory (HBM) turns into a “critical and high‑margin bottleneck” for large language models.

These capital flows are “funnelled directly through foundries like TSMC,” ensuring that as long as hyperscaler capex keeps rising, the broader semiconductor ecosystem remains “the single largest beneficiary.” But networking and specialised hardware are also emerging as major winners as cloud giants move from standalone GPUs to “mega‑cluster” AI supercomputers in their data centres.

To avoid data bottlenecks, hyperscalers are pouring money into high‑speed Ethernet and InfiniBand fabrics, lifting demand for sophisticated switching and routing solutions and for original equipment manufacturers (OEMs) that can integrate proprietary designs into high‑density, rack‑scale systems.

Also Read | Chris Wood’s big warning: The specific risk that will finally trigger the end of AI trade

Energy Security: The Hidden Backbone Of AI

The report argues that the AI arms race is inseparable from a brewing energy supercycle, driven by years of underinvestment and a renewed focus on energy security. ESG pressure and prior boom‑bust cycles pushed oil producers to favour capital discipline and shareholder returns over production growth, leaving capex per barrel of oil equivalent about 49% below 2014 peaks and shrinking integrated oil majors’ average reserve life to around 10 years by 2025. “As a result,” DBS writes, “capex per barrel… remains 49% below 2014 peak levels, while integrated oil majors’ average reserve life has declined,” forcing a rethink as geopolitical risk climbs.

Recent years have seen roughly 90% of oil and gas capex directed at replacing declining output from existing fields and only 10% devoted to expanding supply, underscoring the need to rebuild longer‑term reserves through higher exploration and greenfield activity. Oil services and equipment providers are therefore “well positioned to benefit from the capex cycle,” with spending skewed to brownfield developments that offer faster production and shorter payback, but a growing requirement for longer‑duration, security‑driven investment across the value chain.

DBS links this directly to AI: soaring power and electricity demand from data‑centre construction, plus a “return of energy security,” is expected to support both traditional hydrocarbons and renewables, as well as storage, grid networks and nuclear power.

Inflation, Hawks And The Search For New Safe Havens

The AI arms race, coupled with war‑driven energy shocks and China’s exit from deflation, is feeding a regime of structurally higher inflation and the “return of central bank hawks,” according to the CIO team. The report points out that producer prices for electronic components and accessories have already surged 37.7% since October 2020, while intensive data‑centre construction has helped push up electricity prices. At the same time, US fiscal deficits, the Iran war’s estimated total cost of roughly USD1 trillion, and defence spending rising towards USD1.5 trillion are amplifying inflation tail risks that investors “are under‑pricing.”

DBS warns that the traditional 60/40 equity‑bond portfolio is vulnerable because equity‑bond correlations have risen sharply since the post‑Covid inflation spike, eroding diversification benefits. Using five‑year weekly data, the report finds bonds’ correlation with equities at 0.47, versus much lower correlations for gold, commodities and China A‑shares, in the 0.18–0.25 range.

As a result, CIO asset allocation favours a “barbell” strategy: staying structurally overweight technology to capture AI capex, while adding commodities, gold and China A‑shares as “new safe havens” to manage volatility in an environment of rising bond yields and sticky inflation.

Hou Wey Fook urges investors to “stay invested, stay diversified,” noting that the bank’s barbell strategy has delivered 9.1% annualised net returns since 2019 across multiple shocks, including Covid‑19, the Fed’s 500‑basis‑point hiking cycle in 2022, and the recent Liberation Day sell‑off. At the same time, he stresses that “portfolio downside protection matters” as equity markets increasingly price a benign resolution to geopolitical tensions and a moderation in oil prices that may not materialise.

Positioning For The AI Arms Race Era

Looking ahead, DBS’s 3Q26 CIO asset allocation remains tactically bullish on equities, but emphasises bottom‑up selection over top‑down country calls, given how AI‑linked momentum stocks have vastly outperformed low‑volatility names. The US remains “the only game in town” in terms of equity inflows, with AI‑related earnings strength driving a 26% surge in US earnings in the latest quarter and pushing the market’s price‑to‑earnings multiple to 22 times, above its historical 18‑times average but justified by growth expectations in the bank’s PE/G framework.

Also read: India's next stock market headache isn't oil but a bigger storm brewing in the skies

Within this AI arms race backdrop, DBS highlights several portfolio principles for the coming quarters. First, “riding the AI theme with downside protection” via a barbell between high‑growth AI plays and lower‑energy‑intensity sectors that are less exposed to fuel‑cost shocks. Second, favouring low energy‑intensity industries and companies to avoid margin compression if oil prices remain “higher‑for‑longer” due to disruptions at the Strait of Hormuz and broader Middle East tensions.

Third, using commodities and China A‑shares alongside gold to rebuild diversification as central banks shift from a dovish bias to a cautious, hawkish stance in response to persistent inflation pressures.

(Disclaimer: Recommendations, suggestions, views and opinions given by the experts are their own. These do not represent the views of The Economic Times)