Big Tech AI Capex Spending Alarm

For the first time in the 22 years since Alphabet went public, the company's free cash flow turned negative in the second quarter of 2026 ^[1]. That is not a rounding error - it reflects Alphabet roughly doubling its own quarterly capital spending, to about $44.9 billion year over year, most of it poured into AI data centers, custom chips and cloud infrastructure ^[3]. Management didn't treat this as a one-off: it raised full-year 2026 capex guidance for the third time in 2026 alone, from a $180-190 billion range to $195-205 billion, and told investors 2027 spending will rise significantly and keep pressuring margins ^[1]. The market's reaction looked disproportionate to otherwise strong underlying growth - the stock still tumbled 6-7.7 percent on earnings day, because the $15 billion swing in guidance, not the quarter that just closed, is what investors are pricing ^[2].

The capex story isn't just a stock-market story, it's becoming a credit story. Moody's flagged AI spending as unprecedented and said it now threatens the credit quality of six companies it tracks - Amazon, Meta, Microsoft, Alphabet, Oracle and CoreWeave - because AI hardware demands enormous upfront cash while the revenue it's meant to generate arrives on a much longer timeline ^[4]. Bond investors are already voting with their orders: coverage on hyperscaler debt sales fell from nearly 5x in February 2026 to below 2x by July, forcing issuers like Amazon to offer 18-21 basis points of extra yield just to move a $25 billion bond issue ^[5]. JPMorgan estimates hyperscalers will need roughly $375 billion in bond proceeds between 2026 and 2030 to keep funding the buildout ^[5], even as its own analyst trimmed Alphabet's price target from $460 to $420 on the spending outlook ^[6]. CNBC described the shift bluntly back in February: Big Tech's AI bond binge has shattered the old unspoken deal with investors that hyperscalers would self-fund growth rather than lever up ^[7]. Collectively, hyperscaler USD debt has more than doubled to over $360 billion since last September ^[8].

Capex anxiety didn't stay contained to search-engine economics. Tesla, a company in an entirely different sector, posted its own first negative free cash flow in two years - about $1.1 billion - after burning roughly $5.8 billion on AI and robotics initiatives in the same quarter, and its stock fell 12-14.5 percent in its worst single-day drop in over a year ^[9]. That the two selloffs landed together, even though Tesla and Alphabet share no direct business relationship, shows investors are now pricing AI capex risk broadly rather than company by company ^[10]. The timing raises the stakes for what comes next: Meta and Microsoft report July 29, and Amazon reports July 30, with combined 2026 AI capex across the four projected near $725 billion, up roughly 77 percent from about $410 billion in 2025 ^[11][12]. Each now has to explain comparable or larger spending increases without repeating Alphabet's cash-flow scare, and Citi Research has warned that a pullback in AI investment, if it came instead, could generate a mild recession given how much current growth already depends on it ^[5].

Not everyone reads the selloff the same way. JPMorgan's Jamie Dimon frames the scale plainly: AI capital spending went from $400 billion industrywide in 2025 to $700 billion in 2026, and could exceed a little over a trillion in 2027, while also cautioning markets are getting close to as good as it gets ^[13]. Guggenheim's Michael Morris thinks Alphabet alone could hit nearly $350 billion in 2027 capex, well above the roughly $265 billion Street consensus ^[6], and Baird's Colin Sebastian expects that spending to translate directly into margin pressure ^[14]. Alphabet's own CFO, Anat Ashkenazi, frames the acceleration differently - as a response to a supply-constrained environment where the company is racing to add capacity to meet demand it can't otherwise serve, saying it will keep investing as long as long-term ROI stays compelling ^[14]. Skeptics reach for a specific historical parallel: the 1999-2000 telecom buildout deployed roughly $500 billion in fiber over four years, and 85-95 percent of that fiber sat dark even four years after the bubble burst - a pace of overbuild today's AI infrastructure spending has already exceeded on an annual basis ^[15]. Whether this cycle ends the same way is exactly the open question investors are now pricing into both stocks and bonds at once.