AI will eat a fifth of tech budgets in 2027, but don't expect it to pay for itself

Forrester expects more than half of CIOs to miss AI-related ROI targets next year; only 22% of surveyed AI decision-makers said AI had a positive impact on EBITDA over the past 12 months. The report forecasts AI will account for 20% of business technology spending next year, while open-weight models will handle half of AI tokens but represent just 5% of enterprise AI spend. Forrester says isolated use-case gains may not deliver firmwide returns without organizational changes, increasing pressure to manage costs, infrastructure and risk.

Analysis

The investable signal is a change in the mix of AI spend, not a collapse in spending: enterprise budgets may keep growing while value shifts from premium model access toward low-cost inference and implementation. That threatens pricing power for frontier-model APIs and vendors whose economics rely on high per-token fees. It also raises the bar for application vendors: isolated productivity wins may not support broad seat expansion unless customers redesign workflows and show EBITDA impact.

GOOG is comparatively two-sided. Its Gemini offering faces the same substitution pressure as other frontier models, but open-weight adoption is not automatically negative if workloads move to managed cloud or other paid infrastructure. The key unknown is where inference runs and who captures the margin; open-weight usage could instead encourage self-hosting and dilute external cloud economics. The article provides no evidence yet on either outcome, so the company-level read remains neutral.

Near term (days to weeks), treat this as a valuation/multiple headwind for AI narratives that equate spend with monetization, not as proof of an imminent budget cut. Over 1–3 months, watch enterprise guidance and cloud disclosures for evidence of AI workload growth translating into revenue rather than only infrastructure cost. Over 6–18 months, open-weight models could commoditize routine tasks while concentrating premium pricing in high-stakes workloads. The thesis weakens if customers demonstrate sustained, firmwide EBITDA gains and continue paying for proprietary models at scale.

Key Decisions for Investors

• Do not treat rising AI budgets as a standalone bullish signal for software or model vendors. Prefer evidence of paid usage, renewal/expansion, and measurable customer productivity; otherwise, keep broad AI-monetization exposure sized cautiously.
• For GOOG, stay neutral rather than shorting on this survey alone. Reassess on quarterly cloud growth, AI-related monetization, and capex commentary: improving revenue conversion without disproportionate infrastructure spending would support the upside case; rising costs without corresponding monetization would undermine it.
• Set an alert—not a trade yet—for signs that open-weight adoption is reducing frontier-model pricing or external inference volumes. Confirmation would support underweighting premium API-dependent businesses; evidence that open models are predominantly hosted on paid cloud infrastructure would favor cloud platforms, including GOOG.
• Falsifiers: credible disclosures of broad-based AI-driven EBITDA improvement and continued premium-model usage would challenge the commoditization thesis; conversely, weaker AI monetization alongside escalating infrastructure commitments would strengthen the caution.