President Trump’s sweeping new tariffs may not target AI chips directly, but they are already disrupting the materials, machinery and momentum behind AI infrastructure. As costs rise and uncertainty deepens, developers face a new era of complexity in building the physical backbone of artificial intelligence.
From server racks in Arizona to cooling corridors in Texas, the quiet hum of progress in the AI infrastructure boom just got a little louder, tinged with unease. President Trump’s announcement on Wednesday, 2nd April, of sweeping new import tariffs has injected a fresh wave of uncertainty into an industry already grappling with bottlenecks, energy volatility, and insatiable demand. This is no mere policy pivot for the hyperscale developers fuelling the digital backbone of artificial intelligence. It is a recalibration moment, and not just of budgets.
The long tail of President Trump’s trade strategy has always brushed against tech infrastructure. But this new round of tariffs hits with greater precision. Structured as reciprocal measures to counter perceived trade imbalances, the proposed levies, ranging from 10 to 25 per cent on key imports, are poised to affect everything from structural steel and server chassis to electrical components and industrial cooling systems. While high-value AI chips, such as GPUs from Taiwan or Korea, appear to be excluded from this initial wave of duties, their broader ecosystem is not. The materials and machines that house, cool and power those chips are now more expensive. The calculus just changed for the developers building the physical platforms that power AI.
Prometheus Hyperscale, one of North America’s most active and capitalised data centre developers, is already adjusting its playbook. “In the immediate term, we do not anticipate a dramatic pullback or cancellation of ongoing projects due to the tariffs,” Trenton Thornock, CEO at Prometheus Hyperscale, says. “Our current builds and power infrastructure projects are already ramping to get underway, and demand for capacity is pressing. Data centres have long lead times and significant sunk costs by the construction stage, so near-term projects will keep doing what they should do. The 10 to 25 per cent import tariffs introduce new costs and practical hurdles in the short run.”
Material impact
The implications are not uniform across the sector. Developers already in construction can absorb much of the impact within existing budget contingencies. However, new projects entering the planning or procurement phase will find their cost assumptions challenged. Thornock notes, “We expect a moderate rise in construction and equipment costs on projects starting or procuring materials post-tariff. CBRE estimates the tariff impact could increase commercial construction costs by 3 to 5 per cent. Fortunately, our liquid-focussed data centre ventures have a healthy internal rate of return buffer, so a mid-single-digit bump in CAPEX remains absorbable without jeopardising project viability.”
While Prometheus has hedged its risk through early procurement and domestic sourcing of long-lead items such as Caterpillar gensets and medium-voltage transformers, other players may not be so lucky. Smaller developers with tighter margins and less leverage with suppliers may struggle to maintain pace.
Hardware procurement is another pressure point. “Many data centre hardware components are imported and will now incur the 10 per cent base tariff, or 25 per cent if sourced from certain countries,” Thornock explains. “These end-deployment and workload-specific costs will largely be borne by our tenants since our demarcation point ends at the row-level CDU in a liquid environment. We are prepared to pay slightly higher prices in the short term or negotiate cost-sharing with vendors where possible rather than pause any development.”
This reinforces a critical nuance: while developers can cushion some of the tariff’s blow, the cost will ultimately ripple downstream to cloud hyperscalers and AI-native firms. In a market already defined by scarcity of compute and physical space, any additional cost will be passed on through delayed deployment, reduced performance, or higher prices for access to AI capacity.
Strategic uncertainty and economic noise
The tariff impact does not exist in isolation. It arrives into a macroeconomic environment already fraying at the edges. Caroline Simmons, Chief Investment Officer at Quilter Cheviot, warns, “If the full range of tariffs were implemented and maintained, global and US growth risk becomes more pronounced. Forecasting becomes challenging in this environment, but if the tariffs remain in place for the full year, we could see real US GDP growth lower by approximately 1 to 1.5 percentage points.”
That drop is material. Original 2025 forecasts put growth at 2.2 per cent, meaning the tariffs could effectively cut national economic expansion in half. The drag on business investment, especially in high-intensity sectors like infrastructure, manufacturing and energy, could ripple through capital markets and delay new project funding cycles.
Inflation, too, is becoming harder to ignore. “Estimates of the weighted average tariff rate for the new measures vary from 20 to 25 per cent,” Simmons adds. “That would represent a significant increase, higher even than the protectionist levels introduced during the 1930s under the Smoot-Hawley Act. These measures are a clear escalation from previous trade actions and will likely serve as a headwind to growth while also increasing inflationary pressure.”
The bond market has responded accordingly, with Treasury yields falling as investors flock to safe assets, and expectations of additional rate cuts from the Federal Reserve are rising. However, the central bank faces a bind. Stimulus to offset tariff-induced drag could worsen inflation, while inaction risks tipping the economy toward stagnation. Simmons observed, “There is growing concern about stagflation, where we see slowing growth accompanied by rising prices,” Simmons continues. “This puts into question the narrative of a smooth post-pandemic economic landing.”
Misguided reciprocity
The rationale for the tariffs has raised eyebrows among policy analysts and economists. Colin Grabow, Associate Director at the Cato Institute’s Center for Trade Policy Studies, argues that “tariffs are ultimately taxes on American consumers and companies, including those in the high-tech and IT sectors. These are among the most globally integrated industries, heavily reliant on complex international supply chains. Imposing tariffs disrupts access to critical inputs such as semiconductors, specialised components, and manufacturing equipment, increasing production costs and slowing innovation.”
Grabow dismissed the administration’s justification as incoherent and politically motivated. “The rhetoric around reciprocity increasingly appears to be a pretext for raising tariffs, not a means to secure better trade terms but an end in itself,” he continues. “Unless Congress intervenes, which seems unlikely, the result will significantly harm American businesses and consumers.”
Many executives are troubled by the lack of policy consistency. The scope, targets, and purpose of the tariffs appear to shift with each press conference. Simmons notes that “The fact sheet accompanying the tariffs states that the duties will remain in place until the President determines that the underlying issues are resolved or sufficiently mitigated. This wording leaves a great deal of room for interpretation and change.”
For investors, strategic planners, and CFOs charged with delivering the digital backbone that will support the demand from AI applications, ambiguity is not just inconvenient—it is a risk vector. Data centre investment decisions are made on multi-decade horizons. Projects take years to plan, zone, finance, and build. To inject this regulatory volatility into that cycle is to disincentivise long-term planning when AI infrastructure is central to national competitiveness.
Demand outpaces disruption
Despite these headwinds, the AI infrastructure surge shows little sign of slowing. Cloud hyperscalers remain on the hunt for capacity, and demand from AI-native firms continues to intensify. “Cloud operators and AI firms are currently constrained by data centre capacity,” Thornock adds. “With such customer demand, slowing our views on build schedule could mean forfeiting market opportunity. The frenetic pace of AI-driven data centre expansion is unlikely to slacken simply due to a tariff-induced cost uptick.”
Prometheus is not planning to delay or defer any projects. Instead, the company is revisiting project budgets, escalation clauses, and financing structures to absorb tariff-related costs without compromising delivery. “To be clear, the short-term priority is to stay on track: continue construction and deployment to deliver capacity on current timelines, albeit now with heightened cost oversight,” Thornock adds.
Power availability is now a greater constraint than materials. “Larger than the impact of the tariffs is the dislocation in the power generation equipment markets caused by nationwide grid power shortfalls,” Thornock continues. “For data centre developers who are not like us, having worked large renewable and natural gas microgrid and behind-the-meter co-development deals for several years, the outcome might be less rosy.”
This divergence in resilience could reshape the competitive landscape. Developers with integrated energy strategies, domestic procurement networks and early-stage tenant commitments will fare better. For others, the dual shocks of tariff-driven cost inflation and energy market disruption may lead to pauses, downsizing, or M&A activity as capital consolidates around the most robust operators.
Signals not strategy
What the data centre and AI infrastructure community now face is not a halt but a hesitation. Investment will continue, but with sharper pencils and narrower margins. Developers will push through, but the risk calculus has changed. Each decision on equipment sourcing, location planning, and energy provisioning now carries a policy filter that did not exist two weeks ago.
Grabow summed up the broader critique: “The likely outcome is a trade war where the United States inflicts self-harm while others maintain or escalate their barriers,” he says. “That is hardly a strategic win.”
Whether these measures are a prelude to negotiation or a permanent fixture in US trade policy remains to be seen. Simmons offered a note of caution and hope: “There is potential for tariffs to be scaled back over time as negotiations take place. “However, there remains a possibility that the situation escalates further before any meaningful resolution is achieved.”
For now, the sector marches on cautiously but determined. The AI era demands infrastructure, and infrastructure demands investment. That investment will come at a higher cost, with greater scrutiny, and against the grain of a policy environment that increasingly mistakes noise for leverage.
