Forrester: AI will make up roughly a fifth of enterprise IT spending by 2027

Forrester estimates in an analysis published on 21 October 2026 that AI will account for about 20% of enterprise technology spending in 2027. According to analyst Brian Hopkins, this is not the share that any single company would spend on AI out of its own IT budget, but an aggregate estimate that includes the infrastructure spending of technology vendors. According to Hopkins, the vast majority of the growth comes from infrastructure and compute capacity — servers, chips, cloud services, access to AI models, and software subscriptions.

At the same time, generative AI is changing the way companies pay for software. Instead of predictable annual licensing fees, vendors are introducing models based on token consumption, usage fees, and credits, with the costs of AI agents varying depending on the complexity of the task and the compute power required. According to Hopkins, companies are trying to cover rising AI costs by pushing for lower prices on IT services and by bringing development, support, or customer service back in-house, while security and the modernization of legacy systems such as ERP or Salesforce remain areas where, in his view, cuts would be a mistake — risks are also growing because attackers are using AI agents to find and exploit vulnerabilities at a larger scale than before.

Hopkins further distinguishes three time horizons for the return on AI investment: productivity (short-term, the easiest to measure, but with the smallest impact on financial results), engagement (medium-term improvement in interaction with customers and employees), and transformation (long-term reworking of processes and entry into new markets, riskier, but with the potential for a lasting competitive advantage). Companies focused only on productivity, he says, risk overlooking bigger opportunities in the other two areas.

Hopkins also mentions hidden costs he calls a "trust tax" — testing, red-teaming, guardrails, monitoring, a centralized agent registry, and identity verification during interactions with clients or partners. According to him, the industry is still working out how much companies should realistically expect such costs to be. He expects 2027 to be a difficult year, with some companies cutting AI budgets while a smaller group with a clear strategy continues to invest; the full financial impact of this split, he says, will only become apparent in 2028–2029.