The New Definition of AI Infrastructure Spending
AI infrastructure spending is the large‑scale capital outlay by hyperscalers on chips, data centers, networks and power to run and commercialize artificial intelligence workloads across cloud platforms, consumer services and enterprise software. For most of the past decade, those same companies treated rising cash flow as a twin engine: fuel growth and fund shareholder buybacks. That model is now breaking. Goldman Sachs estimates the major hyperscalers will pour about USD 755 billion (approx. RM3.48 trillion) into capital expenditures in 2026, an 83% jump from 2025. According to MarketWatch, this surge is concentrated in Amazon, Alphabet, Meta, Microsoft and Oracle, whose AI buildouts define the current spending cycle. The result is a structural shift: cash that once retired stock now pays for GPUs, land, power contracts and, increasingly, debt. Shareholders no longer own a growth-plus-buyback hybrid; they own pure growth vehicles tied to AI.
Big Tech Capex Explosion and the Vanishing Buyback Cushion
The scale of big tech capex is staggering. Amazon outlined a roughly USD 200 billion (approx. RM922 billion) capital plan for 2026 tied to AI infrastructure, custom chips, data centers and robotics. Analysts have guided Alphabet into the USD 175–185 billion (approx. RM807–853 billion) range, while Meta raised its 2026 capex outlook to USD 125–145 billion (approx. RM576–667 billion), blaming higher infrastructure costs and memory pricing. Microsoft is spending at levels above the USD 88.2 billion (approx. RM406.7 billion) it recorded in fiscal 2025. The impact on shareholder buybacks is immediate: Goldman’s analysis, cited by MarketWatch, shows hyperscaler buybacks falling by nearly two‑thirds in the first quarter, with Alphabet repurchasing no stock after buying about USD 15.1 billion (approx. RM69.6 billion) a year earlier. That technical cushion under valuations has thinned, leaving prices more exposed to sentiment and AI execution risk.
Microsoft’s Lawsuit Shows the Risks of Opaque AI Capex
Microsoft illustrates how AI capex can collide with market expectations. The company reported USD 51.5 billion (approx. RM237.4 billion) in cloud revenue in fiscal Q2 2026, yet its stock fell about 10%, erasing an estimated USD 357 billion (approx. RM1.65 trillion) in market value in a single day. A Michigan pension fund has filed a securities class action, arguing Microsoft sold a cleaner AI growth story than its numbers supported. The complaint highlights Azure growth slowing to 39% from 40%, capital expenditures hitting USD 37.5 billion (approx. RM173 billion) in one quarter—up 66% year on year—and gross margin slipping to just over 68%. Plaintiffs say Microsoft framed capacity issues as supply constraints while redirecting GPUs and data‑center capacity toward AI and OpenAI‑linked workloads. The case underlines a core concern: investors worry less about spending than about unclear AI ROI and uneven disclosure.
Free Cash Flow Squeeze and the New Capital Allocation Math
The AI hyperscaler investment boom is a cash flow story as much as a growth story. MarketWatch noted that Amazon’s free cash flow for the 12 months through March 31, 2026 fell to USD 1.2 billion (approx. RM5.5 billion) from USD 25.9 billion (approx. RM119.4 billion) a year earlier, driven by a USD 59.3 billion (approx. RM273.3 billion) jump in property and equipment purchases. The Wall Street Journal, citing analyst estimates, reported free cash flow for the five big hyperscalers could drop 91% in 2026 to about USD 16 billion (approx. RM73.8 billion), even as net income rises 25% to USD 506 billion (approx. RM2.33 trillion). Accounting smooths the optics—hardware is capitalized and depreciated over years—yet the cash leaves now. Debt markets are stepping in as another funding source, meaning capex, buybacks and leverage are now competing claims on the same dollar.
Shareholders’ New Choice: AI Bet or Near‑Term Cash
With buybacks shrinking, shareholders face a sharper choice. One option is to stay for the long‑term AI bet, accepting thinner near‑term capital returns in exchange for exposure to future AI revenue and platform dominance. The other is to demand capital discipline and clearer AI ROI before endorsing hyperscaler investment plans on this scale. For many, buybacks had been more than a bonus; they reduced share counts, supported earnings per share and provided a floor when sentiment soured. Remove that support and valuations depend more heavily on proof that AI infrastructure can be run efficiently and monetized at attractive margins. Founders and enterprise customers should also take note: when hyperscalers ration cash among data centers, GPUs and debt, every strategic investment, partnership or cloud discount must clear a higher hurdle. The buyback era may not be over forever, but its dominance has clearly ended.
