Big Tech spend set to near $700B by 2026 on AI buildout

AI infrastructure capex near $700B: estimates vary, scope matters

Big Tech is projected to pour close to $700 billion into AI investments by 2026. In practice, AI infrastructure capex spans accelerators and GPUs, high-speed networking, data center buildout, and power systems, and ranges largely depend on how much of this stack is included. The current debate centers on hyperscaler spending and whether compute expansion can be matched by power and utilization at scale.

According to Goldman Sachs, 2026 capital outlays tied to AI could reach about $527 billion, while MLQ reports a combined outlook for Alphabet, Amazon, Meta, and Microsoft in the $650–$700 billion range, primarily for data centers, GPUs, and power systems. These figures are often framed as directional and may diverge based on scope, timing, and what portion of multiyear build programs is recognized in a given calendar year.

Who is spending: Alphabet and peers’ hyperscaler spending, data center buildout

Alphabet (GOOGL) and other hyperscalers are at the center of this cycle, directing spend to accelerators, interconnects, and large-scale data center campuses that can accommodate dense power and cooling. As reported by AOL, big tech companies are committing hundreds of billions to AI infrastructure, and Alphabet is viewed as having structural advantages versus rivals.

Industry leaders have described this as a durable transition in computing rather than a short-lived spike. “Once businesses commit to AI, they will need ever more compute capacity,” said Jensen Huang, CEO of Nvidia.

At the time of this writing, public market context reflects sustained focus on AI supply chains; based on Nasdaq data, NVIDIA Corporation (NVDA) closed at 177.19 on 27 February, moved to 177.81 after hours (+0.35%), and traded within a 52-week range of 86.62 to 212.19, with an intra-day market capitalization near $4.31 trillion. These figures are descriptive and do not imply any outlook.

Economics: ROI, free cash flow strain, utilization and power constraints

The economics hinge on converting high capex into revenue and cash flow via rising utilization and customer demand for inference and training. According to The Motley Fool, the spending wave is a significant tailwind for infrastructure suppliers, while also raising questions about whether hyperscalers’ free cash flow can keep pace with elevated outlays.

As reported by Fortune, research warns that profits may lag these investment levels, implying longer payback periods if monetization or utilization build more slowly than expected. Framing the range this way underscores the need for disciplined deployment, efficient models, and workload mix that can absorb capacity at acceptable margins.

Operationally, utilization rates, energy availability, and the timing of grid connections and permitting are likely to be decisive constraints on rollout speed. These factors, together with unit economics for accelerators, networking, and power, will shape whether today’s AI infrastructure capex ultimately translates into durable returns and steadier free cash flow over time.