From Cash Machines to AI Infrastructure Companies
AI infrastructure spending is the rapid, large‑scale deployment of cash into chips, data centers, networking gear, and related systems that power artificial intelligence workloads, replacing the prior focus on returning excess capital to shareholders through stock buybacks and signaling a durable shift in how major technology firms treat growth, risk, and balance sheet capacity. For most of the past decade, Alphabet, Microsoft, Meta and Amazon offered investors a comfortable trade: high growth plus generous buybacks. That trade is over. Goldman Sachs now expects the largest AI hyperscalers to spend about USD 755 billion (approx. RM3.47 trillion) on capital expenditures in 2026, an 83% jump from 2025. This is not a marginal adjustment; it is a wholesale reallocation of cash flows away from shareholder returns and into AI infrastructure. Anyone still valuing these names as if they remain predictable “cash machines” is clinging to an outdated model.
The Buyback Era Is Ending, and That’s the Point
The clearest signal of Big Tech capital allocation shifting is the steep tech buyback decline. Hyperscaler buybacks fell by nearly two‑thirds in the first quarter, with Microsoft a notable exception. Alphabet bought back no stock in its latest quarter after repurchasing about USD 15.1 billion (approx. RM69.4 billion) in the same period a year earlier. Meta is pushing more cash into infrastructure, and Amazon has long stopped being a regular buyer of its own shares. Axios reports that analysts now see hyperscaler capital expenditures near USD 770 billion (approx. RM3.54 trillion) in 2026, roughly equal to cash flows from operations, funded increasingly with debt and equity issuance while buybacks are pulled back. In other words, management teams have chosen AI capex over financial engineering. The shareholder contract has changed from “we shrink the share count” to “we build the rails of the AI economy and you wait for the payoff.”
Cash Flow Squeeze: AI Capex Meets Corporate Finance Reality
This AI infrastructure spending arms race is now a cash flow story, not a hype story. Amazon has laid out around USD 200 billion (approx. RM920 billion) of capital spending for 2026 tied to AI infrastructure, custom chips, data centers and robotics. Meta has lifted its 2026 guidance to USD 125 billion–145 billion (approx. RM575–667 billion), blaming higher infrastructure and memory costs. The Wall Street Journal reports that free cash flow for the five big 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 is the AI buildout in one sentence: earnings up, cash down. J.P. Morgan estimates AI capital spending could reach USD 5.5 trillion (approx. RM25.3 trillion) by 2030, with USD 4.1 trillion (approx. RM18.9 trillion) financed through debt. These companies increasingly resemble infrastructure providers, with heavy capex, long payback periods, and rising financing needs.
Capacity Strains and Consolidation: A Sector Restructuring in Real Time
The spending spree is already exposing limits and triggering consolidation. Microsoft has turned to Amazon Web Services to support GitHub as AI‑driven coding demand overwhelms Azure’s capacity, underscoring how even a top hyperscaler can run out of room when AI usage spikes. At the same time, mega mergers show capital chasing scale and data wherever it can. Fox plans to acquire Roku in a USD 22 billion (approx. RM101.2 billion) cash‑and‑stock deal, giving it control of about 73% of the combined company and access to vast viewer data. SpaceX has announced a USD 60 billion (approx. RM276 billion) all‑stock acquisition of AI coding startup Cursor (Anysphere) to deepen its grip on developer data and strengthen its xAI models like Grok. Salesforce will buy AI customer service startup Fin for USD 3.6 billion (approx. RM16.6 billion), adding AI agent technology to its platform as it battles Microsoft and Oracle in enterprise automation. This is all one story: scale, data, and infrastructure are now worth more than near‑term cash returns.
What This New Capital Regime Signals for Investors
Investors cannot treat hyperscaler capex trends as a temporary spike; they are the new baseline. Goldman’s USD 755 billion (approx. RM3.47 trillion) capex forecast for 2026 is not a one‑off, and analysts already see 2026 capital spending roughly matching operating cash flows. Frankly, investors need to stop treating this as the old Big Tech model with a temporary AI surcharge. If enterprise customers keep moving workloads into AI systems and cloud pricing holds, analysts expect a cash flow rebound in 2028 and 2029. But between now and then, the market must price these companies more like railroads, utilities, or telecom networks: long‑duration projects, financing cycles, and less excess cash to feed buybacks. Anyone still pricing Alphabet, Microsoft, Meta, Amazon or Oracle as if the 2017–2022 buyback machine is humming is working from an outdated model. The $755 billion question isn’t whether they can afford this AI infrastructure spending. It’s whether shareholders are ready to own infrastructure businesses instead of cash‑rich tech darlings.
