CryptoSlate
A single forecast of data-center growth added $5.7 billion to the modeled cost of keeping the US eastern grid powered, according to PJM's independent market monitor. The July auction for the 2028-29 delivery year produced roughly $16.4 billion in capacity revenue; rerun without forecast demand from new and expanding data centers, it falls to about $10.7 billion. Existing data-center load stayed in the counterfactual, isolating the price effect of expected AI growth alone.
The same pressure is showing up on store shelves. Australia's statistics bureau recorded a 6.1% quarterly rise in prices for audio, visual, and computing equipment, the largest in a series dating to 1980, and the Reserve Bank of Australia linked it to memory-chip costs driven by AI infrastructure demand. The RBA raised its cash rate to 4.85% on Nov. 3, citing AI-related investment among several sources of price pressure.
Why it matters
AI's inflation footprint has moved from investor presentations into the machinery of monetary policy and electricity markets. Central banks in Australia, the US, and the UK are now tracking component demand as a cost source, even as the San Francisco Fed calculates that information-processing equipment added just 0.2 percentage points to a 3.4% headline inflation rate. The promised productivity offset remains unproven: a survey of nearly 6,000 executives found 89% reporting no effect on firm productivity over three years, while the BLS's preliminary Q3 reading of 3.6% productivity growth cannot attribute any share to AI.
The costs also reach people who never buy AI services. Grid capacity is priced years ahead from demand forecasts, so expectations of data-center growth lift payments to power plants before any facility is built. Households connected to the same system have little control over what gets constructed or how those costs flow back through utility contracts and local regulation.
Market impact
Watch the next PJM auction and how regulators assign the bill. Virginia has already created a separate rate class for Dominion Energy's large-load customers, requiring 14-year minimum commitments for contracts from Jan. 1, 2027, an early template for making data centers carry more of their grid costs. Competing academic estimates, one finding data-center entry raised residential rates 2.1% and another finding a modest average reduction, suggest the local regulatory design, not a national average, will decide who pays. Until AI's productivity gains show up in prices consumers actually pay, the spending side of the ledger keeps compounding.
Frequently asked questions
-
How did a forecast add $5.7B to US power auction costs?
PJM's independent market monitor reran the 2028-29 capacity auction without forecast demand growth from new and expanding data centers, holding existing data-center load and other assumptions fixed. Modeled revenue fell from about $16.4 billion to $10.7 billion.
-
Why did computer prices in Australia rise 6.1% in one quarter?
The Reserve Bank of Australia linked the jump in audio, visual, and computing equipment prices to rising memory-chip costs driven by demand for AI infrastructure. It was the largest increase in the series since it began in 1980.
-
Did the RBA raise rates because of AI?
The RBA raised its cash rate to 4.85% on Nov. 3 citing AI-related investment as one of several sources of price pressure, alongside higher fuel costs and domestic capacity constraints. The bank said AI's effect on overall consumer prices remained modest.
-
Are data centers raising household electricity bills?
The evidence is mixed. An MIT working paper estimated data-center entry raised residential rates 2.1% from 2010 through 2024, while another study found a modest average reduction. Capacity payments are only one part of power costs, so auction increases do not translate directly into residential bills.
-
Has AI delivered measurable productivity gains yet?
A survey of nearly 6,000 executives found 89% reporting no effect on firm productivity over the previous three years, though some Fed-linked research found gains among adopters. The BLS's preliminary Q3 reading showed 3.6% productivity growth but cannot identify AI's contribution.
WuBlockchain
CoinTelegraph
TheBlock
Crypto News