68% plan to increase AI spend, only 18% say infrastructure is ready as sustainability concerns grow

CEOs Plan to Boost AI Spend-But Infrastructure and Sustainability Lag

A new NTT and WSJ Intelligence survey shows a clear signal from the top: 68% of global CEOs will increase AI investment over the next two years. The catch-only 18% say their infrastructure is highly optimized to support AI at scale. And 83% acknowledge the environmental cost of scaling fast.

The accompanying white paper frames the core challenge and opportunity for leaders building AI-ready enterprises. As NTT DATA Inc. CEO and chief AI officer Abhijit Dubey puts it: "AI scales not through advanced models alone, but also through the strength of the foundations beneath them." That means skilled talent, unified teams, resilient platforms, and trusted data.

Where the Money Is Going

Nearly half of executives expect to lift AI budgets by 11% or more. Two out of three CEOs prioritize predictive AI today, but most agree the best results will come from combining predictive, generative, and agentic systems. Strategy is shifting from isolated pilots to enterprise capabilities-if the plumbing can keep up.

What's Holding Back Scale

• Compute and network constraints: Access to accelerators, limited bandwidth, and legacy architectures are throttling performance and throughput.
• Data bottlenecks: Prep, quality, lineage, and secure access remain gating factors to enterprise-grade AI.
• Performance vs. planet tension: About 70% prioritize maximum or performance-first workloads; 75% still believe sustainable practices cut profitability.
• Shadow AI risk: Top concerns include data leakage (41%), data integrity (40%), security vulnerabilities (38%), and unreliable decisions (35%). 67% believe they have effective AI governance, but gaps persist.
• Scale pressure: 83% say leaders feel intense pressure to prioritize rapid benefits over energy use.

Photonics Moves From Idea to Roadmap

Interest in optical networking and computing is accelerating as AI workloads strain power and capacity limits. 91% of executives know photonics; 55% express strong adoption interest and 36% already use it.

NTT's IOWN initiative points to what's possible: up to 100x less power, a 125x capacity increase, and a 200x reduction in end-to-end delays across networks and data centers compared to conventional infrastructure. For leaders, photonics is no longer a science project-it's a lever for cost, speed, and sustainability.

The Executive Playbook: Make AI Scalable, Secure, and Sustainable

• Set a two-speed plan: Fund near-term AI use cases tied to clear P&L impact while running a parallel track for platform and infrastructure upgrades.
• Stand up an AI platform team: Centralize MLOps, LLMOps, feature stores, model registries, and security. Treat AI as a product with SLOs for cost, latency, and reliability.
• Audit infrastructure readiness: Map model classes to accelerator mix, storage tiers, and network bandwidth. Eliminate single points of failure and build for horizontal scale.
• Fix the data layer first: Establish data contracts, lineage, and quality SLAs. Create governed access patterns and a catalog so teams can find, trust, and ship faster.
• Contain Shadow AI: Require registration of AI tools, provide secure sandboxes, enforce DLP and secrets scanning, and review vendors for privacy and IP terms.
• Adopt a governance framework: Use the NIST AI Risk Management Framework to formalize risk controls across model lifecycle, monitoring, and incident response.
• Build GreenOps into FinOps: Track carbon intensity by workload and region, schedule jobs for low-carbon windows, and set targets for PUE, water use, and energy sourcing. See sector benchmarks from the International Energy Agency.
• Pilot photonics now: Start with optical interconnects in high-traffic links, benchmark latency and power, and update your 24-36 month data center roadmap.
• Modernize networks: Deploy high-speed fabrics, QoS for AI traffic, and micro-segmentation. Optical where it counts; cache and compress where it pays.
• Upskill for human-AI collaboration: Train teams on prompt patterns, review workflows, and risk flags. For role-based learning paths, see courses by job.
• Tie spend to value cadences: Use quarterly value reviews with cost-of-delay metrics. Kill slow movers, double down on workloads with measurable margin, revenue, or risk reduction.

Signals to Watch

Expect boards to push harder on AI guardrails and verifiable ROI. Energy markets and grid constraints will shape siting decisions for AI capacity, with photonics and workload scheduling becoming standard levers. Vendor SLAs will add sustainability clauses, and regulators will look closer at model risk, provenance, and data use.

Methodology Snapshot

Findings are based on a survey of 359 global CEOs. Companies had at least $1B in annual revenue (U.S.) or $500M (international). The research covered investment priorities, infrastructure maturity, risks, workforce expectations, sustainability, photonics, data governance, and AI regulation.

Bottom Line

AI investment is accelerating, but the winners will be the ones who fix foundations while they scale. The play is simple: fund outcomes, harden the platform, govern the data, and cut the energy curve as you grow. Do that, and AI becomes a disciplined capability-stable, secure, and profitable.