AI Server OEM Tariff Strategies in Turbulent Times: How Will (or Should) Vendors React?

The components that are needed to produce today’s most performant Artificial Intelligence (AI) servers rely on a supply chain that is notably more complex than for general-purpose infrastructure, with a narrower pool of suppliers concentrated in countries like Taiwan, South Korea, and Malaysia. Unsurprisingly, U.S.-headquartered AI server Original Equipment Manufacturers (OEMs) (e.g., Supermicro, Dell, Hewlett Packard Enterprise (HPE), Penguin Solutions, AMAX, etc.) were awarded a risk premium by the market in the aftermath of the Trump administration’s trade restrictions and tariffs. The supply chain for the semiconductors that go into AI servers is arguably the most complex in the world—this encompasses compute, such as Central Processing Units (CPUs) and Graphics Processing Units (GPUs) (from Intel, AMD, NVIDIA), as well as the networking, memory, and storage that support such AI hardware.

Stock markets were beset by panic at the beginning of April 2025: the NASDAQ 100 technology index shed 10% between April 1 and April 7, semiconductor indices lost around 15%, and the Hand Seng TECH Index, composed of Chinese hyperscalers like Baidu, dropped over 15%. Markets have recovered since then, but are still some way off pre-tariff levels. Server OEMs were similarly affected, shedding double-digit percentages due to concerns about procurement and demand for data center buildout.

It is also worth noting that hyperscalers, with Capital Expenditure (CAPEX) that underscores a large part of said server OEMs’ revenue, have been largely unaffected by the tariffs. Google has confirmed plans to spend US$75 billion on CAPEX this year, much of which will go to AI data center expansion (US$17 billion was committed in 1Q 2025 already). Microsoft’s president of cloud operations recently quelled fears by reiterating plans to spend over US$80 billion this year on infrastructure, while Amazon is expected to spend over US$100 billion, all of which trumps 2024 CAPEX.

How AI server OEMs react is uncertain, especially because the dust has yet to settle on how tariffs apply—some exemptions were carved out for semiconductors, as well as certain servers and components. But given the volatility of policy decisions, individual nations and industries may be granted concessions. Exemptions for the United States-Mexico-Canada Agreement (USMCA) are also beneficial for AI servers manufactured in Mexico, which has emerged as a significant hub alongside Taiwan. Nonetheless, even servers manufactured on U.S. soil, with components from globally-sourced U.S.-headquartered semiconductor vendors, may be impacted in the medium to long term by tariffs, even if exempt under today’s regime.

There are several strategies that AI server OEMs could implement in reaction to cost pressures from affected supply chains:

• Cost Pass-Through: Transfer costs to customers with higher Average Selling Prices (ASPs). This risks a dampening demand, although the necessity to fulfill the demand for AI compute—especially as workloads such as Agentic AI scale, which will require much more capacity—suggests that demand is more inelastic and will not be strangled by higher prices. However, we expect more elasticity from enterprise customers than hyperscalers. For OEMs passing through costs, any reduced demand may be offset by increased ASPs, but the risk remains that competitors absorb costs and gain market share.
• Margin Absorption: Take the hit and reduce profit margins with unchanged ASPs. By leaving ASPs largely unchanged, OEMs will decrease their profitability, but maintain their market share, assuming competitors adopt the same strategy. OEMs will also gain market share from rivals that opt for cost pass-through.
• Product Alterations: Favor “Made in USA,” friendshoring supplies, and/or introducing new Stock Keeping Units (SKUs). This could result in further disruptions to existing supply chains, with knock-on effects for delivery schedules. Radical product alterations are unlikely, given the foothold of NVIDIA compute hardware in today’s AI servers, for example. Once existing inventories are depleted, OEMs with less diversified supply chains will be at a disadvantage compared to those with more diversified existing supply chains.

Regardless of which strategy OEMs choose, there are several overarching dynamics:

• Competitive Landscape Disruption: Tariffs might not impact all OEMs equally. Those with more diversified assembly locations or pre-existing inventory might gain a temporary advantage. It could also accelerate shifts in manufacturing footprints if tariffs target specific countries of final assembly.
• Impact on Innovation Cycles: Uncertainty and higher costs could potentially slow down the adoption of cutting-edge AI chips if OEMs become more cautious about inventory risk, or customers balk at immediate price hikes for incremental performance gains.

• Inventory Management Challenges: Anticipating tariffs might lead to attempts at stockpiling (this is difficult, given the current shortages), while further sudden tariff changes can disrupt existing inventory flow and valuation.

The strategy that OEMs will choose, and the impact this will have on the market, is still unclear. Not all OEMs will be impacted equally. Those with more diversified or U.S.-based assembly locations might gain a temporary advantage—Lenovo has communicated to shareholders that its global manufacturing footprint could be flexed to offset the tariffs’ impacts. We could also see accelerated shifts in manufacturing bases if tariffs continue to apply unequally. Nevertheless, trade barriers add layers of administrative burden, customs complexities, and potential delays. OEMs selling into the United States must navigate new compliance requirements and face uncertainty regarding the scope, duration, and potential future expansion of tariffs, all of which adds complexity to procurement strategies.

To effectively manage tariff impacts and geopolitical risks, OEMs might consider the following options in their business strategies:

• Optimize Supply Chain and Consider Flexibility with Strategic Manufacturing Site Selection: Actively evaluate and invest in final assembly locations that are not subject to punitive tariffs on finished servers, such as Mexico (even if components originate from tariff-impacted regions, given the “substantial transformation” clause of U.S. trade rules). Considerations include logistics, labor costs, local skills, and geopolitical stability. Furthermore, by diversifying locations, OEMs can achieve flexibility with multiple assembly sites in different regions to hedge against future trade policy shifts.

• Refine Pricing and Product Communication: For OEMs implementing the above or with existing diversified supply chains, customers will value the reliability and resilience of optimized supply chains—especially important in the current Return on Investment (ROI) and Time to Market (TTM)-sensitive stage of the investment cycle. By clearly communicating the value proposition and highlighting supply chain resilience and future pricing stability, OEMs will be better placed to implement pass-through costs and updated pricing strategies without sacrificing market share.
• Engage in Effective Lobbying: Work through industry associations to advocate for clear, predictable, and internationally palatable trade rules. By lobbying against tariffs (or future restrictions) that could harm domestic access to critical AI infrastructure, OEMs can leverage their unique insight into global supply chains, and share information that may be unclear to civil servants and politicians with broader remits and less AI industry-specific knowledge.

Zooming out, the demand for data center compute may also be reduced for reasons not covered in this analysis. If critical components needed for data center power and cooling are affected, the knock-on effects will also be felt in the AI server market, as orders will be cancelled without the required infrastructure to support the servers. Tariffs may yet be scaled back or expanded, and ABI Research continues to monitor the entire supply chain and will update forecasts once significant information is crystallized in the market.