AI infrastructure investment: the new center of Big Tech finance
AI infrastructure investment refers to the large, long-term spending by cloud hyperscalers on chips, data centers, power, and networking needed to run advanced artificial intelligence models at commercial scale. That line item has stopped being a side project and become the budget. Goldman Sachs, cited by MarketWatch, expects Amazon, Alphabet, Meta, Microsoft and Oracle to spend about USD 755 billion (approx. RM3.47 trillion) on capital expenditures in 2026, an 83% increase from 2025. This surge in Big Tech capex spending is no cosmetic upgrade; it absorbs cash that once flowed to share buyback programs and cushions earnings per share. The result is that public investors who treated these companies as both high-growth and reliable capital-return engines now own businesses whose balance sheets are tilted toward physical assets and future AI payoffs instead.
From buyback machines to capex-heavy hyperscalers
For most of the past decade, Alphabet, Microsoft, Meta and Amazon offered a simple proposition: durable growth plus steady buybacks. According to analysis reported by MarketWatch, hyperscaler buybacks fell by nearly two-thirds in the first quarter, as AI infrastructure investment took priority. Alphabet bought back no stock in its latest quarter after repurchasing about USD 15.1 billion (approx. RM69.46 billion) in the same period a year earlier, while Meta has raised its capital expenditure guidance and Amazon has not been a regular repurchaser for years. The Goldman Sachs forecast that hyperscaler capex will reach about USD 755 billion (approx. RM3.47 trillion) in 2026 shows why: every spare dollar is competing with chips, data centers and power contracts. The era when buybacks reliably shrank share counts and supported valuations is giving way to a spend-first, wait-for-AI-revenue mindset.
Cash flow compression and the new capital allocation math
The AI buildout is most visible in cash flow. MarketWatch noted that Amazon’s free cash flow for the 12 months through March 31, 2026 fell to USD 1.2 billion (approx. RM5.52 billion) from USD 25.9 billion (approx. RM119.14 billion) a year earlier, largely because purchases of property and equipment rose by USD 59.3 billion (approx. RM273.0 billion). The Wall Street Journal reported that free cash flow for the five hyperscalers is expected to drop 91% in 2026 to about USD 16 billion (approx. RM73.6 billion), even as net income is projected to rise 25% to USD 506 billion (approx. RM2.33 trillion). That gap defines the new hyperscaler capital allocation puzzle: net income looks strong, accounting spreads depreciation over years, but the cash is already gone into AI infrastructure. With less surplus cash, share buyback programs and special capital returns move down the priority list.
Microsoft’s AI capex surge and rising investor scrutiny
Microsoft is a case study in how opaque AI spending and slowing cloud momentum can collide with investor expectations. In fiscal Q2 2026, Microsoft posted USD 51.5 billion (approx. RM237.0 billion) in cloud revenue and still saw its stock fall about 10%, wiping out an estimated USD 357 billion (approx. RM1.64 trillion) in market value in a single day, according to reports cited by The Next Web and SAM Expert. Shareholders have since filed a securities class action, arguing that Microsoft oversold Azure’s AI growth story while downplaying a deceleration in Azure growth and the extent of its AI-related capacity diversion. Capital expenditures hit USD 37.5 billion (approx. RM172.5 billion) in one quarter, up 66% year on year, putting the annualized run rate near USD 100 billion (approx. RM460.0 billion). That level of AI infrastructure investment heightens scrutiny on how every dollar is allocated between growth, margins and shareholder returns.
What the end of the buyback era means for investors and partners
The AI spending race changes what it means to own Big Tech. Without steady repurchases, earnings per share will depend more on actual AI profitability than on shrinking share counts. Valuations now lean harder on the assumption that AI services from Azure, AWS and Google Cloud can scale revenues enough to cover hundreds of billions in depreciation. Founders and enterprise customers feel the shift too. As one analysis notes, when a hyperscaler becomes a strategic investor or preferred infrastructure provider, “every check now competes with data center leases, power contracts, GPUs, memory and debt service.” Hyperscaler capital allocation now favors long-lived AI assets and debt financing over cash-rich balance sheets and buyback cushions. Investors who once focused on headline growth and repurchase authorizations must now track capex guidance, free cash flow trends and the timing of AI payback more closely than ever.
