AI investment is likely to keep climbing in 2027, even as it contributes less to US growth than it does today, ING economists argue. Their report also points to a second, longer-lived wave of spending on power and grids, and a shift towards riskier financing.
Watch Desk analysis
What happened
ING estimates that investment in US data-centre structures, computers and software reached 5.6% of GDP in the second quarter of 2026, up from 3.8% in 2022. Its economists expect hyperscaler capital spending to exceed US$1 trillion in 2027, increasingly directed towards AI infrastructure.
That does not mean the same spending will keep adding to growth at the same rate. ING says the investment’s contribution to US growth is probably near its peak: as the spending base gets larger, even continued increases add less to growth. The report also describes financing gradually moving beyond hyperscalers’ cash flows towards debt, private credit and more complex structures.
Key findings
- AI investment keeps rising
ING expects heavy spending in 2027, with chips, power, permits and skilled workers limiting how quickly projects can proceed. - The growth contribution may peak
Investment can reach new highs while adding less to annual growth, simply because the starting base is larger. - Power becomes a major part of the build-out
ING estimates that if half of expected 2027 hyperscaler spending went to AI compute, it could imply more than 12 gigawatts of additional power demand. - The infrastructure has different lifespans
ING contrasts GPUs, which it says may have an economic life of three to six years, with grid assets that can last for decades.
Why it matters
The AI build-out is no longer just a story about chips and data centres. Power generation, transmission, cooling and the cost of financing them could shape how quickly projects proceed, who benefits and who bears the costs. ING notes that grid investment may be more widely shared across suppliers, while higher electricity demand can also feed into household bills.
The report’s forecast is not a guarantee of what companies will spend or what returns they will earn. Its central distinction is useful, though: investment can keep rising even as the economic boost from each additional dollar fades.
Our read
The most interesting question is not whether AI spending hits another record. It is whether AI revenues and productivity gains catch up with the infrastructure being built around them. ING’s base case is a plateau rather than a bust, but points to financing risk if expected returns fall short. Follow the bottlenecks and the bills, not just the spending headlines.
What to watch
- Whether hyperscalers keep raising capital-spending guidance.
- Signs of cancelled projects or shorter lead times for chips and equipment.
- How much AI infrastructure is funded through debt and private credit.
- Whether grid constraints and data-centre demand push electricity costs higher.
Discussion spark: If AI investment keeps rising but contributes less to growth, should governments and companies keep prioritising the build-out, or wait for clearer evidence of returns?
Sources and evidence
- AI in 2027: 20 questions for the global economy | ING THINK (5 November 2026, 00:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.