AI infrastructure spending: from cash machine to capex supercycle
AI infrastructure spending refers to the massive capital required for chips, data centers, networks, and related hardware that support large-scale artificial intelligence services and cloud workloads. For the largest hyperscalers, this spending has turned into an arms race. Goldman Sachs estimates that the leading AI hyperscalers will spend about USD 755 billion (approx. RM3.48 trillion) on capital expenditures in 2026, an 83% increase from the previous year. This sum spans Amazon, Alphabet, Meta, Microsoft, and Oracle, and it is no longer a side story to growth; it is the growth story. The cash that once flowed easily into stock support is now fused into concrete, silicon, and long-lived assets. In accounting statements, the assets are depreciated slowly, but on the cash ledger the outflow is abrupt, recasting how investors should think about tech capex trends and AI capital allocation.
Stock buyback decline and the end of a shareholder era
The surge in AI infrastructure spending is rewriting the bargain that defined the last decade of Big Tech. Companies like Alphabet, Microsoft, Meta, and Amazon once delivered strong growth while shrinking share counts through generous buybacks. That pattern is fading. According to a Goldman Sachs analysis cited by MarketWatch, hyperscaler buybacks fell by nearly two-thirds in the first quarter, with Microsoft the major exception. Alphabet halted repurchases in its latest quarter after buying back about USD 15.1 billion (approx. RM69.9 billion) a year earlier, while Meta and Amazon have redirected more cash toward infrastructure. For shareholders, buybacks were more than a bonus; they helped support earnings per share and provided a technical floor when sentiment soured. Removing that cushion means valuations now rest more heavily on confidence that AI revenue arrives at scale and on time.
Hyperscaler investment priorities and the free cash flow squeeze
The current wave of hyperscaler investment is a cash flow story as much as a technology story. Amazon has outlined roughly USD 200 billion (approx. RM925 billion) in capital spending for 2026, with CEO Andy Jassy linking the outlay to AI infrastructure, custom chips, data centers, and robotics. Alphabet is guided by analysts into the USD 175 billion to USD 185 billion (approx. RM809 billion to RM856 billion) range, while Meta has raised its capital expenditure outlook to USD 125 billion to USD 145 billion (approx. RM578 billion to RM670 billion), citing infrastructure and memory costs. MarketWatch reports that Amazon’s free cash flow for the 12 months through March 31, 2026 fell to USD 1.2 billion (approx. RM5.6 billion) from USD 25.9 billion (approx. RM120 billion) a year earlier as property and equipment purchases rose by USD 59.3 billion (approx. RM274 billion). The Wall Street Journal projects free cash flow for the five major hyperscalers dropping 91% in 2026 to about USD 16 billion (approx. RM74 billion), even as net income climbs.
Competitive barriers and the squeeze on smaller players
The consolidation of AI capital allocation among a handful of hyperscalers is reshaping competition. When a small group commits USD 755 billion (approx. RM3.48 trillion) to AI infrastructure spending, they raise the bar for anyone trying to compete on raw compute, storage, and network capacity. Data centers, custom accelerators, and global networks are capital-heavy assets that reward scale and balance sheet strength. Smaller cloud providers, startups, and independent AI companies cannot easily match this pace of hyperscaler investment, especially while these giants lock in supply from key chipmakers and build proprietary hardware. At the same time, broader tech news shows AI taking center stage across the stack—from Microsoft leaning on external cloud capacity for GitHub to Databricks, OpenAI, and others racing to deliver enterprise AI tools. The result is an ecosystem where foundational infrastructure tilts toward the largest players, while smaller firms are pushed toward niches, partnerships, or asset-light models.
Repricing tech: how investors must rethink AI capital allocation
For investors, the stock buyback decline among hyperscalers means a different risk-reward profile for tech valuations. Where buybacks once offered consistent support, capital is now trapped in assets that pay off only if AI services scale profitably. Accounting smooths this shift, since hardware is depreciated over years, but as Morgan Stanley analyst Todd Castagno notes, this is a period where “everyone looks good” while the cash bills arrive far earlier. Valuations must now reflect heavier exposure to long-term AI infrastructure spending and a smaller buffer from repurchases. At the same time, AI-driven changes across the industry—from automation-related layoffs to heightened security risks—underline how dependent future earnings are on reliable, secure AI platforms. Investors need to recalibrate their models for tech capex trends, paying closer attention to free cash flow trajectories, infrastructure utilization, and how each hyperscaler plans to turn massive AI capital allocation into durable revenue.
