Strip away the ribbon cuttings and the renderings, and every data center is a procurement project: transformers, switchgear, power supplies, server boards, cooling controllers and thousands of electronic components that all have to arrive on time, meet spec and stay available for years of operation. When one of those parts disappears, the most ambitious construction program in modern industry stalls at the loading dock.

This is part of the artificial intelligence story, but it’s one that gets far less attention than it deserves. Because right now, everything around those parts is under strain at once.

The buildout itself keeps setting records. The largest technology companies spent more than $400 billion on capital expenditure in 2025, and the International Energy Agency expects that spending to jump another 75% in 2026.The same agency projects global data center electricity consumption will roughly double to 945 terawatt-hours by 2030.

The resistance is setting records too. Community opposition delayed or blocked delayed or blocked at least 75 U.S. data center projects worth about $130 billion in the first quarter of 2026, and more than 100 municipalities have imposed local moratoriums. Tariffs add a third pressure. Section 232 semiconductor tariffs began taking effect in January, 2026. The Computer & Communications Industry Association estimates that a 25% semiconductor tariff applied to data center equipment works out to a 15.6% tax on construction, roughly $90 billion per year in lost investment and GDP.

Those are the headline battles, and they will stay in the headlines. But conversations with engineering, procurement and quality leaders reveal that the fight that actually determines whether facilities and the hardware inside them ship on schedule happens at the component level. Two pressures deserve far more attention than they get: the true financial cost of obsolescence, and the difficulty of qualifying alternate parts.

AI demand is pulling electronic component supply toward hyperscale buyers. Semiconductor lead times reached 40 weeks in March, 2026, and lead times for high-capacity transformers now stretch as long as four years.

A shortage delays a build. Obsolescence forces a permanent decision. Too many teams budget for the two as if they were the same problem, and the difference is expensive. When a manufacturer discontinues a part, every design that depends on it needs a resolution, and each rung of the resolution ladder costs more than the last.

The price tag is well documented. In a March, 2026 survey of 439 engineering and supply chain professionals conducted by Fuld & Company, 85% of respondents across aerospace and defense, electronics, automotive, medical devices and industrial manufacturing reported design rework costs of up to $250,000 per obsolescence event. Run several design programs in parallel and those costs multiply. The parts themselves get pricier too Components that survive only on the secondary market typically sell for 10 to 15 times their original price.

Those figures cover a single component. A data center power system, server board or cooling controller carries hundreds of line items, and each obsolete part triggers its own resolution. Requalification and certification push the totals higher. In aerospace, defense and medical applications, requalifying a system after a redesign frequently costs more than the engineering work itself.

The structural problem is a lifecycle mismatch: Component technology cycles keep shrinking while the systems built on them are expected to run for a decade or more. AI hardware, with its rapid generational turnover, widens that gap further with every product generation. Teams that plan only for shortages are pricing the wrong risk. The budget conversation needs to move from expediting fees to redesign exposure.

Substitution looks like the cheap way out, and the industry leans on it heavily. In one study of obsolescence events on Boeing commercial aircraft, 67% were resolved through part substitution. The economics explain why. What teams underestimate is the burden of proof that comes with every swap: a dual challenge of design and compliance.

On the design side, an alternative must match the original in form, fit and function — electrical parameters, timing, thermal behavior, package footprint and temperature grade. A near-match that misses one parameter fails in qualification, or worse, in the field. Verifying equivalence requires complete, accurate technical data on the original part and every candidate replacement.

On the compliance side, the alternative must independently satisfy RoHS, REACH, and expanding PFAS restrictions, along with customer environmental, social and governance (ESG) requirements and the country-of-origin rules that now determine tariff exposure. A part that meets every electrical spec but fails a compliance check is still the wrong part. Most teams find out late: 62% percent of surveyed professionals discover compliance violations after the design phase, when fixing them costs the most. Every swap restarts the documentation trail.

Shortage conditions add a third hazard: counterfeits. When buyers leave authorized channels to find scarce parts, counterfeiters follow the demand. ERAI, which tracks nonconforming electronic parts, reported a 25% increase in suspect counterfeit and nonconforming parts in 2024, the highest total since 2015. Many were active components already on extended manufacturer lead times, exactly the parts data center supply chains are chasing hardest.

When supply chain and engineering leaders ask how to get ahead of this, the answer starts in the same four places:

Monitor lifecycle status continuously. Most surveyed professionals lack four or more months of visibility into component obsolescence, pricing trends and supply shifts. End-of-life notices carry last-time-buy deadlines, and the cheapest resolutions disappear when those windows close. Quarterly bill of materials reviews are too slow for this market.

Qualify alternatives before you need them. Approving second sources during design costs thousands. Doing the same work during a line-down event costs hundreds of thousands, plus weeks of schedule.

Consolidate component data. Resolution decisions stall when engineering specs sit in one system, compliance certificates in another, and lifecycle forecasts in a third. Decisions this consequential need supply chain intelligence built on accurate, complete component data in one integrated view, where a proposed swap can be checked against specifications, regulations, and sourcing risk at once.

Make it cross-functional. Obsolescence lands on procurement's desk, but the resolution belongs equally to engineering, quality and compliance. Shared data and shared ownership shorten every resolution.

The data center boom will keep straining power grids, community patience and trade policy. No company controls any of that. What a company does control is how early it sees component risk and how fast it acts. The organizations that are expected to come out of this boom strongest are the ones treating obsolescence as a design-stage financial decision, and verifying every alternative against complete data before a line goes down. Everyone else will pay redesign prices.

Sarita Benjamin is general manager, supply chain intelligence at Accuris.