Material Insights · Understanding the Material World

AI and digital platforms and Distribution grids, mini-grids and interconnectors

AI adds large, concentrated electric loads, from chip fabs to data centres, that local networks must connect and supply without interruption; transformer shortages and slow connections make the grid, not demand, the pacing factor.

AI and digital platforms (industry) and Distribution grids, mini-grids and interconnectors (energy): 4 relationships.

How they are linked

1. Reliable power for chip fabs

Physical flow · Strong · Grids → AI chip supply

AI accelerators are made in fabs that run continuously and use large amounts of electricity, so an unstable supply interrupts chip output; the site rates grid supply 5 of 5 on criticality for AI accelerator and chip designers.

  • 5 of 5 criticality for AI accelerator and chip designers (fab power)
  • about 8% TSMC's share of Taiwan's electricity use, as reported in 2024
  • 23.7% S&P projection of TSMC's share of Taiwan's power use by 2030, reported 2024

Advanced nodes need EUV lithography, which uses far more power than older tools, so each new node raises fab load. Taiwan's reserve margin was still below the government's 15% target in 2024, which makes supply stability a live risk for the world's main AI chip source.

Enterprise types that depend on it:

  • AI accelerator and chip designers: Fab power (criticality 5 of 5)

Sources: Material Insights; TrendForce

2. AI load on networks

Transition · Rising · AI growth raises network load

Data centres for AI are large single loads that arrive in a few clusters, so local networks must add substations, feeders and connections much faster than ordinary demand growth requires.

  • 17% growth in data centre electricity demand in 2025, against 3% for all electricity
  • 13.56 GW estimated data centre electricity demand in India by 2031 to 32

The IEA reports that slow grid connections are already delaying data centre projects. In India, Mumbai holds nearly half of operational data centre capacity, so the load falls on a few urban networks rather than spreading across the country.

Sources: IEA; W.Media (MeitY statement to Rajya Sabha)

3. Transformer shortage

Shared risk · Strong · Equipment supply limits new connections

Every new connection needs transformers, and factory capacity for them has not kept pace with demand from new loads, renewables and ageing networks.

  • 30% and 10% projected US supply deficits for power and distribution transformers, 2025
  • about 50% share of US distribution transformer supply expected from imports, 2025

The IEA also names transformers and gas turbines among the supply chains that tightened in 2025 as data centre demand grew. Where transformers are short, AI projects wait for connections or build their own generation.

Sources: IEA; Wood Mackenzie

4. Distribution reform in India

Policy · High in India · Discom finances shape readiness for new load

Indian distribution companies must fund the network upgrades that new data centre and industrial loads need, so their losses and finances decide how fast they can connect and supply them.

  • ₹3,03,758 crore outlay of the Revamped Distribution Sector Scheme, 2021-22 to 2025-26
  • 22.62% to 15.04% aggregate technical and commercial losses, 2013-14 to 2024-25
  • ₹2,701 crore discoms' aggregate profit after tax, 2024-25

Discoms moved from a loss of ₹67,962 crore in 2013-14 to a profit in 2024-25. Better finances make it easier to invest in substations and feeders for clustered loads, but the gains are recent and uneven across states.

Sources: Drishti IAS (Ministry of Power data); Angel One (Ministry of Power data)

For investors

Key learnings

  1. AI load arrives in a few dense clusters, so local network capacity and connection speed decide where AI capacity can be built.
  2. Transformers are short, with a projected 10% US deficit for distribution units in 2025, which favours equipment makers and sites with secured connections.
  3. AI chip supply depends on fab power: TSMC alone used about 8% of Taiwan's electricity in 2024 reports, so grid stability there is a supply risk for every AI platform.

What to track

  • Grid connection queues for large loads: Longer queues delay AI capacity and favour sites with power (Quarterly · grid operators, utility filings)
  • Transformer lead times and prices: Signals how quickly networks can add new connections (Quarterly · equipment makers, Wood Mackenzie)
  • Taiwan electricity reserve margin: A falling margin raises risk to AI chip output (Monthly · Taipower)

For policymakers

Key learnings

  1. India's data centre power demand, estimated at 13.56 GW by 2031 to 32, will land on a few urban networks such as Mumbai's, so distribution planning must be local.
  2. Discom losses fell from 22.62% in 2013-14 to about 15% in 2024-25 and the sector posted a ₹2,701 crore profit in 2024-25, which gives some room to fund upgrades for new loads.
  3. Transformer supply is a lever: the US expected about half its distribution transformers to be imported in 2025 and still faced shortages, so domestic factory capacity matters for how fast India can connect new loads.

What to track

  • Discom AT&C losses and profit: Shows capacity to fund network upgrades (Annual · Ministry of Power, PFC)
  • Data centre power demand by state: Shows where distribution upgrades are needed (Annual · CEA, MeitY)

India

India's discoms returned to profit in 2024-25 after RDSS and earlier reforms, just as data centre power demand is projected to reach 13.56 GW by 2031 to 32, concentrated in cities such as Mumbai.

Sources

  1. AI and digital platforms: materials, Material Insights (2026)
  2. TSMC's electricity demand could triple by 2030, raising concerns on Taiwan's power supply risks, TrendForce (2024)
  3. Data centre electricity use surged in 2025, even with tightening bottlenecks driving a scramble for solutions, IEA (2026)
  4. India's DC capacity almost touched 1.5 GW in 2025: IT Ministry, W.Media (MeitY statement to Rajya Sabha)
  5. Power transformers and distribution transformers will face supply deficits of 30% and 10% in 2025, Wood Mackenzie (2025)
  6. Reforming India's power distribution sector, Drishti IAS (Ministry of Power data) (2026)
  7. Power discoms net ₹2,701 crore profit in FY25, says ministry, Angel One (Ministry of Power data) (2026)