{
  "id": 17972985,
  "title": "Who pays for AI’s $2.6 trillion build-out? Look to corporate budgets",
  "url": "https://urgent.news/2026/11/03/who-pays-for-ais-2-6-trillion-build-out-look-to-corporate-budgets",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-03T12:59:41.000Z",
  "source": {
    "name": "Fortune",
    "slug": "fortune",
    "url": "https://fortune.com/2026/11/03/who-pays-ai-2-6-trillion-build-out-look-to-corporate-budgets-cfo/"
  },
  "original_language": "en",
  "account": "By 2030, the AI ecosystem will require approximately $2.6 trillion in annual spending to sustain various aspects such as data centers, chips, model development, and investor returns, as per a BCG Institute analysis. Enterprises are projected to contribute the largest share of this total, accounting for $1.8 trillion, followed by consumers at $366 billion, hyperscalers' internal use at $204 billion, and emerging technologies like physical AI at $200 billion. These estimates encompass global spending, excluding China, and highlight infrastructure costs predominantly in the United States, where most new capacity is being developed. The $1.8 trillion figure represents what the AI ecosystem would need from enterprises, rather than an actual spending projection.\n\nThis $1.8 trillion can be achieved by enterprises reallocating $857 billion from their existing technology budgets and adding $937 billion in new spending, which would increase enterprise tech budgets from their current level to roughly 13% above International Data Corporation's IT spending forecast. In 2030, AI's share of these budgets is expected to rise from 5% today to 25%.\n\nThe growth of the AI industry heavily depends on corporate spending. For CFOs considering additional AI investment, this means presenting boards with evidence of productivity gains, cost reductions, new revenue, or competitive advantage, according to Clark O’Niell, BCG’s global lead for AI, infrastructure, and cyber. However, some strategic benefits may be challenging to quantify using traditional ROI calculations. O’Niell noted that the question extends beyond immediate financial returns to include the potential cost of falling behind competitors. While fear of falling behind can be a legitimate strategic consideration, it should not replace a business case.\n\nBCG envisions the AI build-out as more resilient than past infrastructure booms, primarily due to hyperscalers financing most of their investments through cash flow rather than debt. However, financing infrastructure does not ensure adequate returns. O’Niell identifies two indicators to monitor: how infrastructure investment is financed, with an increasing reliance on debt signaling higher risk, and whether deployed compute generates sufficient revenue. As frontier labs enter public markets, their disclosures could indicate if revenue is keeping pace with compute costs. The economics of this development leave little room for disappointment. BCG assumes AI chips have an average lifespan of five years and estimates that each gigawatt of capacity becoming operational in 2026 must generate around $15 billion in revenue to cover its costs. If demand for AI is weaker than expected, it could lead to price wars and losses for model providers and data-center operators. For corporate buyers, this implies avoiding long-term price-per-token commitments as AI prices continue to decline. Having flexible contracts could help CFOs capture future cost reductions. While the $2.6 trillion target is achievable, it is not guaranteed, and even if enterprises provide the necessary spending, the question remains whether they will capture enough economic value to justify it.",
  "summary": "BCG estimates companies would cover $1.8 trillion a year by 2030.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}