Alphabet (NASDAQ:GOOGL | GOOGL Price Prediction) is sitting on a Google Cloud backlog that nearly doubled quarter-on-quarter to $462 billion, and Sundar Pichai went on the most recent earnings call and said, in plain English, “We are compute constrained in the near term.” When the supplier is rationing for everyone, the interesting question is what the rationing tells you about the next two years of AI economics.
What the Financial Times reported
The FT’s June 28, 2026 story said Google limited Meta’s (NASDAQ:META) access to Gemini models because it could not supply the full computing capacity Meta wanted to buy. Google told Meta around March that it couldn’t meet the requested capacity, which disrupted and delayed some of Meta’s internal AI projects and pushed Meta to economize on tokens.
A capacity shortfall is a very different fact pattern from a deliberate competitive cutoff, though this does lean into territory where AI companies are forced to make some choices. Competition is obviously going to be a factor when companies choose who to sell compute to.
Compute is the binding constraint
“We are seeing strong demand and we are working hard to bring more capacity online,” Pichai has told investors, framing the supply problem as structural rather than transient. “Our Cloud revenue would have been higher if we were able to meet the demand,” he said. Google Cloud reported $20.03 billion in Q1 revenue, up 63% year over year, with operating margin expanding to 32.9% from 17.8% in the year-ago quarter. To chase the backlog, Alphabet raised 2026 capex guidance to a range of $180 billion to $190 billion, with 2027 capex expected to significantly increase.
The hyperscalers collectively have earmarked more than $725 billion in 2026 AI capex. Demand still outruns it. Q1 capex alone hit $35.67 billion, more than doubling year over year, which compressed free cash flow to $10.12 billion, down 46.6%. The trajectory matters: full-year 2025 capex landed at $91.45B, and the 2026 guide nearly doubles that again, a signal management is willing to absorb near-term free cash flow pain to chase the backlog.
The vertical integration subtext
Owning the models, the TPUs, and the data centers means Google has a coherent allocation framework. “The compute needed for model development is a foundation for everything we do,” he said, with Search, YouTube, and Cloud customers slotted in around it under a “robust ROIC framework.” In a shortage, that integration is an advantage. Internal workloads and paying Cloud customers will reasonably sit ahead of a competitor buying capacity on the side.
That is structural, and the reporting establishes no intent beyond it. It also explains why Meta, which runs its own Llama family but still wanted Gemini for specific workloads, found itself behind Google’s first-party demand in the queue. The same logic applies to every external buyer competing with the host’s own roadmap.
What this signals from here
Markets are reading the moment with unease. GOOGL is down 7% over the past month, though GOOG stock is rebounding swiftly as of this writing. The stock is on a 97% one-year run. GOOG stock trades around $350.
Gemini is now processing more than 16 billion tokens per minute via direct API use, up 60% quarter over quarter, and Gemini Enterprise paid monthly active users grew 40% sequentially. This is the demand signal that is forcing the rationing in the first place.
The throttle headline will fade. The constraint behind it will not. If you want one number to watch, it is the backlog. As long as it keeps growing faster than capacity comes online, “we couldn’t supply what they wanted to buy” is going to be a recurring sentence across the industry.
Top Gaining Stocks
Top Losing Stocks