Average increase in DC energy cost
+ 18 %
Gas and oil volatility, plus logistics premiums, raise the base cost compared to the 2024 cycle.
Geopolitics
How sanctions, supply chains, and technological concentration influence the real cost of AI.
AISHA starts from a simple idea: AI no longer scales in a homogeneous global market. In April 2026, oil, maritime routes, advanced packaging, and technology sanctions affect the final cost as much as the model's own architecture.
Average increase in DC energy cost
+ 18 %
Gas and oil volatility, plus logistics premiums, raise the base cost compared to the 2024 cycle.
Logistics delay on GPU racks
14-21 days
Diversions via the Red Sea and Hormuz extend hardware delivery times from Asia to Europe and the U.S. East Coast.
Sub-7nm capacity still concentrated in Taiwan
86 %
Diversification exists, but critical dependence on TSMC and advanced packaging remains nearly intact.
While many analyses continue discussing PUE or per-token efficiency, the real picture is harsher: AI operates within an economy of trade war, strained energy, and regionalized hardware.
The result is a system less efficient than it appears on paper. The industry no longer optimizes solely for performance, but for access to firm electricity, secure routes, packaging capacity, and politically viable suppliers.
AISHA's conclusion is that geopolitics is no longer context. It is a variable of cost, timeline, and technical viability for AI deployment.
Military risk in the Middle East doesn't just move Brent. It also raises gas prices, lengthens maritime routes, and pressures data centers that still purchase part of their energy or backup in volatile markets.
AISHA compares the Brent trajectory with a DC operational index to show how the geopolitical premium is already entering the bottom line.
The Ukraine conflict reordered two things at once: the map of critical inputs for chip manufacturing and the geography where deploying electricity-intensive capacity still makes sense.
The 2022 crisis partially normalizes, but the new equilibrium remains structurally above the pre-war level.
Continental Europe loses relative advantage, while the Nordics, Texas, and the Middle East gain weight due to energy or industrial strategy.
Declining competitiveness
Electricity prices continue penalizing new AI-intensive projects and push capacity toward geographies with more stable energy.
Winners on cold and renewables
Natural cold, less cooling pressure, and cheaper electricity consolidate their attractiveness for strategic deployments.
Scale with volatility
Industrial proximity and infrastructure abundance help, but exposure to gas and maritime logistics remains high.
The neon bottleneck has eased relative to the 2022 peak, but the structural surcharge of reconfiguring refining and supply has not disappeared. The chain no longer returns to its starting point.
Something similar happens with palladium, titanium, and other materials where Russia and its intermediaries still carry weight. Although the flow isn't completely cut off, a hidden geopolitical tax emerges that ultimately reflects in more expensive hardware.
The most serious systemic risk remains the concentration of advanced nodes and CoWoS packaging. Diversification is progressing, but more slowly and at a higher cost than the public narrative usually admits.
Decentralization exists, but most of the useful capacity and critical know-how remains concentrated in Taiwan.
Moving capacity out of Taiwan reduces political risk, but still adds friction in price, yields, or energy consumption.
If an effective blockade of Taiwan occurred today, market inventories would last only a few months and the cost of critical hardware would spike immediately. The real bottleneck isn't just manufacturing chips: it's packaging them, moving them, and deploying them at scale.
There is no single trajectory. AISHA summarizes here three plausible paths and how each geopolitical shock activates different chains affecting energy, capex, and the pace of model development.
The same AI changes shape depending on the level of geopolitical escalation, energy cost, and continuity of advanced supply.
AISHA does not recommend betting on a single technological exit. AI resilience in 2026-2028 requires combining hardware, energy, and deployment decisions as if they were part of the same problem.
The core idea is simple: the best technical optimization loses value if it depends on a politically fragile supply chain or on energy you cannot lock in long-term.
Same category
Wed Apr 01 2026 00:00:00 GMT+0200 (Central European Summer Time)
Bottlenecks in data centers, energy, chips, and short-term deployment capacity.
Wed Apr 01 2026 00:00:00 GMT+0200 (Central European Summer Time)
La cara menos visible del stack de IA: backup energético, transporte pesado y fragilidad operativa.
Wed Apr 01 2026 00:00:00 GMT+0200 (Central European Summer Time)
Semiconductors, strategic minerals, and dependence on resources concentrated among few actors.