Skanska signs additional contract for data center in Georgia, USA, worth USD 255M, about SEK 2.3 billion

Skanska signed a $255M (about SEK 2.3B) supplemental contract to build a data center in Georgia, USA, with the award booked in US order bookings for Q2 2026. The project covers 22,700 square meters of data center space, site work, underground utilities, and fitout of administration space and five data halls. Construction started in March 2026 and is expected to be completed in Q1 2028.

Analysis

This is less about one contractor’s backlog and more about the persistence of AI-related infrastructure capex. A new U.S. data-center award of this size reinforces that hyperscale demand is still outrunning available powered land and installed capacity, which should keep civil/utility-heavy contractors busy even if headline software spending softens. The second-order winner is the local ecosystem: electrical gear, switchgear, generators, cooling, and underground utility vendors tend to see higher mix and stickier pricing as projects move from site prep into fit-out.

The more important signal is duration. The work stretches into 2028, which means margin realization is gated by execution over multiple bidding cycles rather than one quarter’s bookings beat. That lowers the chance of immediate P&L upside but raises the probability of sustained revenue visibility; contractors with pre-existing client relationships and repeat awards should keep taking share because owners increasingly value schedule certainty over lowest bid. The main risk is that the data-center buildout can be delayed by power interconnection queues or permitting friction, pushing revenue out while costs on labor and materials remain sticky.

Consensus likely understates how capital intensive the AI buildout has become for the supply chain rather than just the end customer. If demand remains strong, the bottleneck migrates to grid equipment and electrical contractors, not the cloud/platform layer, creating a cleaner relative-value trade than chasing the obvious AI beneficiaries. Conversely, if hyperscalers pause spending, these long-dated projects can still convert slowly because sunk site work and client-specific fit-outs make cancellations costly, cushioning the near-term downside relative to more cyclical industrial end-markets.

The contrarian takeaway is that this is bullish for infrastructure execution names but not necessarily for broad construction beta. The market often treats data-center awards as a generic “AI trade,” yet the real alpha is in companies with power-delivery, underground utility, and mission-critical fit-out exposure, where pricing power is better and backlog quality is higher. Watch for any evidence that order flow is concentrated in a few repeat clients: that would imply customer concentration risk, but also a higher hurdle to dislodge incumbents once they are embedded in the build program.

Key Decisions for Investors

• Long a basket of U.S. infrastructure contractors with data-center exposure versus short a broad industrial ETF over 3-6 months; thesis is backlog durability and mix uplift, with downside limited if AI capex merely slows rather than rolls over.
• If accessible, buy call spreads on electrical equipment / power infrastructure names for 6-12 months; best risk/reward sits where order visibility is already high but margins are not fully priced for multi-year demand.
• Pair trade: long data-center enablement names (utility construction, switchgear, generators) against short general commercial construction names; the spread should widen as mission-critical work keeps price discipline while ordinary commercial capex remains soft.
• Avoid chasing headline construction stocks on the announcement alone; wait for confirmation that backlog converts into higher gross margin, because execution risk and working-capital drag can delay earnings upside by 2-4 quarters.
• Set a catalyst watch for any power/interconnection bottlenecks in Georgia over the next 6-18 months; if delays surface, fade the broader AI-infrastructure trade and rotate toward equipment suppliers that still get paid on long-lead orders.