Alphabet Raises 2026 CapEx to $205 Billion and Admits to Using Third-Party Capacity
The search giant has raised its 2026 CapEx target to up to $205 billion and has acknowledged that it will rely on third-party capacity until its own construction meets cloud demand, which has grown by 82% in the quarter.
Alphabet has increased its 2026 CapEx to a range of $195 to $205 billion, up from the previous estimate of $180 to $190 billion made three months ago. The company spent $44.9 billion just in the second quarter, approximately 60% on servers and 40% on data centers and networking equipment. The announcement, made on the evening of July 22 during the earnings conference call, caused shares to drop by about 5% in after-hours trading, despite Google Cloud's revenue growing 82% year over year, reaching $24.8 billion in quarterly receipts.
The contracted Cloud backlog surged by over $50 billion in a single quarter to $514 billion, and the operational margin of the unit more than tripled from 20.7% to 35.6%. Today, each new dollar of cloud revenue contributes more profit than it did a year ago, and this metric supports Sundar Pichai's argument for increased spending. According to the CEO, during the call with analysts, "we feel we are in the very early stages of what seems like a secular shift across multiple areas."
CFO Anat Ashkenazi made a more concerning revelation: the company will resort to "third-party capacity" in the third quarter as a "bridge strategy" while it completes the construction of its own infrastructure. In other words, even one of the world's top three companies with the most data centers cannot build quickly enough to meet its own demand. Ashkenazi also indicated that 2027 should see a "significant increase" in CapEx.
Where the Cost Burden Lies Outside the United States
For corporate customers in Germany, this solidifies Google Cloud's position as a provider with investment power that surpasses any sovereign European alternative. Regions such as Frankfurt and Berlin, now pillars for banks and insurers that need to keep data within the bloc due to GDPR, rely on the influx of servers derived from this $205 billion investment. Gaia-X and sovereign cloud initiatives remain rhetorical figures when comparing $200 billion against a few billion in Europe annually.
In Japan, every gigawatt of capacity that Alphabet, Microsoft, and Amazon subtract from the market raises the cost of energy, GPUs, and industrial real estate that SoftBank also needs to replicate the formula in Osaka and Chiba. The Japanese group is already operating as a strategic partner of OpenAI, and the race for watts and silicon is turning into a local issue: Nomura analysts have been pointing out margin pressures in the group’s data center division since last fourth quarter.
For consulting firms and integrators in India, particularly Infosys, TCS, and Wipro, Google Cloud is becoming the counter where most of the business agents' pipeline will be delivered over the next 18 months. The $514 billion contracted backlog doesn’t execute itself, and the market for certified labor in Vertex AI and BigQuery has heated up again in Bengaluru and Hyderabad after a slow start to the year.
What the Stock Decline Actually Indicates
The herd mentality read was "Alphabet is spending too much." A more useful interpretation is different: the return on CapEx depends on maintaining Google Cloud's growth at 80% per year on an increasingly larger base. If that rate drops to 40%, this year’s $205 billion and the promised increase for 2027 become a heavy burden. If it remains at 80%, Alphabet may be the only hyperscaler capable of digesting its own investment independently.
The counter-argument comes from Mike Mayo at Wells Fargo, who views the current cycle as a "turning point" with hyperscalers still delivering better returns on invested capital than any comparable sector. On the skeptical side, Bernstein has warned that delays in Gemini 3 and reliance on "third-party capacity" indicate that Alphabet itself acknowledges the physical limits of current expansion, a fact that undermines its own return argument.
What changes for the CIO negotiating multi-year contracts in 2026 is not the nominal price per token, which is still falling, but rather the vendor's bargaining power. A supplier that grows its backlog by 12% in the quarter above its own construction pipeline speaks from a different position than it did two years ago. Those relying on aggressive discounts to justify the TCO of migration should start accepting that the next renewal will come under worse conditions.