Vendors project nearly 7 trillion dollar in AI infrastructure spending by 2030 while AI datacenter revenues are projected at 3,6 trillion dollar by 2030
‘This will lead to a massive reduction in value,’ Gartner said during the opening keynote at the IT Symposium in Barcelona, ‘but it will also create opportunities through incentives to promote token consumption and the proliferation of agents. However, lower per-token costs do not necessarily reduce total spending. “Convenient or inexpensive entry points can encourage unplanned consumption.” Gartner’s chief of research compared it to going into a shop for milk and bread and coming out with a kayak.’
Memory, servers, storage and energy all add to the cost. Organisations should not only forecast whether costs might become unsustainable; they should ensure that AI creates enough value to cover them.
Gartner advocates an AI ‘pin-up’ dashboard for tracking who consumes tokens and why. ‘Use consumption information to turn off “token sinks” and prioritise “value vendors” that generate meaningful outcomes’.
