Big tech reining in its hyper-aggressive artificial intelligence spending may be a tantalizing prospect for investors worried about margins. It would also be an incredibly costly admission.
While the markets watch closely for signs of a slowdown, a cooling-off period doesn’t seem to be in the cards just yet. Second-quarter earnings reports are highly likely to reveal yet another period of blowout AI capital expenditures. Wall Street analysts estimate that combined capital spending by the four dominant “hyperscalers”—Google, Microsoft, Amazon, and Meta—surged an astonishing 74% year-over-year to hit $168 billion in the June-ending quarter alone.
Signs of Cracks in the “Build at All Costs” Mentality
Despite the outward push to spend, underlying market movements indicate that tech giants are actively searching for ways to rationalize, offset, or de-risk these massive computing budgets:
-
Meta’s Cloud Consideration: Whispers that Meta Platforms might begin renting out its excess computing capacity to external businesses have raised eyebrows across Wall Street. As a company without a legacy commercial cloud business (unlike Microsoft Azure or Amazon Web Services), Meta entering the infrastructure-leasing market strongly hints that internal infrastructure capacity may temporarily be outstripping actual usage.
-
Capacity Sharing Agreements: The race is so expensive that even new titans are partnering up to share the burden. For instance, Elon Musk’s xAI division signed a massive infrastructure agreement to effectively share its computing power with Anthropic for a staggering $1.25 billion per month.
-
The “Free Sample” Startup War: Major AI model makers like OpenAI, Anthropic, and Google Cloud are locked in a fierce battle to secure enterprise market share. They are aggressively handing out massive computing credits—in some cases totaling more than $3 million per startup—effectively funding seed-stage operational costs just to lock developers into their respective ecosystems.
The Ripple Effect on Wall Street
The reality of these massive expenditures is hitting tech companies and chip manufacturers from two different sides:
The Margin Squeeze: This relentless infrastructure build-out is significantly compressing free cash flow. In fact, among the core hyperscalers, only Google’s parent company, Alphabet, has managed to outperform the broader S&P 500 index over the first half of the year.
Hardware Volatility: Investors are highly reactive to any hint of oversupply. News of potential compute overcapacity sparked sharp pullbacks across the semiconductor space, causing the PHLX Semiconductor Index to drop 11% over a two-day period as major chip manufacturers like Nvidia, Broadcom, and AMD faced heavy selling pressure.
The Bottom Line for the Next Cycle
Based on full-year projections, the combined capital expenditures of Google, Amazon, Meta, and Microsoft are expected to climb to an astronomical $710 billion.
Whether Meta’s exploration into renting compute is a temporary strategic offset or the first real warning sign that tech infrastructure has drastically overbuilt depends on one metric: enterprise adoption. If real-world generative AI workloads fail to rapidly scale and absorb this staggering amount of computational supply, tech’s high-stakes game of chicken may finally force someone to blink.
