Global AI spending, including on software, hardware, and services, is forecast to grow 26.9% YoY to $154B in 2023 and companies slow to AI “will be left behind”
Context & Ripple Effects
IDC's $154B global forecast is a milestone in a spending curve it has been charting for years — back in 2019 it projected Asia Pacific AI spending would reach $5.5B that year, up 80% YoY, en route to $15B by 2022. The 2023 figure extends that regional tracking into a worldwide number spanning hardware, software, and services.
The 'left behind' warning proved prescient in a specific way: when the boom matured, most of the money flowed to infrastructure rather than applications, leaving software vendors like Salesforce struggling to capture AI budgets — and later surveys found fewer than half of enterprise AI projects generating returns above cost.
First-order effects
- Enterprises face immediate budget-allocation decisions across the three categories IDC counts, with hardware and cloud infrastructure positioned to absorb the largest share of new AI spending.
- Vendors selling AI infrastructure and services get a demand tailwind validated by a third-party forecast, strengthening their case in enterprise procurement conversations.
Second-order effects
- As spending concentrates in compute and cloud rather than application software, application vendors are forced into defensive repositioning — the squeeze Bloomberg documented at Salesforce shows where that budget skew lands.
- Rising outlays raise ROI scrutiny inside adopters, which is exactly the tension Teneo's CEO survey later surfaced: majority plans to spend more alongside a majority of projects not yet paying for themselves.
Third-order effects
- If the spending pattern holds without matching revenue, the gap between compute investment and monetization widens — the shortfall Bain sized at roughly $800B against the $2T annual revenue needed by 2030 to fund projected compute demand.
- 'Adopt or fall behind' hardens from vendor marketing into a structural expectation, pushing boards to fund AI programs whose business cases remain unproven.
The trend: Enterprise AI is moving through an infrastructure-led supercycle in which spending commitments outrun demonstrated returns, shifting value toward whoever owns compute and distribution.