Alphabet Q3 2026: $51.4B Capex, Cloud Up 103%, Free Cash Flow at -$9.1B

Nov 10, 2026 · 7 min · Ajay Kumar

Alphabet spent $51.4 billion on property and equipment in the three months to 30 September 2026, more than double the $24.0 billion of a year earlier, and for the second quarter running that spending was larger than the cash its business generated. The Q3 2026 earnings release, published on 28 October 2026, shows operating cash flow of $42.3 billion against that $51.4 billion, leaving free cash flow at negative $9.1 billion. On the same page, Google Cloud revenue rose 103% to $30.7 billion. The market chose to look at the cash: the shares fell more than 12% the next day, according to SiliconANGLE.

What happened

Alphabet reported revenue of $127.6 billion for the quarter, up 25%. Google Services grew 11% to $96.3 billion. Google Cloud more than doubled, and its operating income went from $3.6 billion to $11.1 billion. The release says TPU system sales were "a meaningful contributor" to cloud growth, which matters because Alphabet now sells its accelerators as hardware, not only as rented capacity.

Net income still fell, from $35.0 billion to $26.2 billion, because other income swung from a $12.8 billion gain to an $8.2 billion loss. AP reported that an $8.7 billion loss on equity investments cut 55 cents off earnings per share, and that revenue came in just under the FactSet consensus of $127.97 billion. Shares dropped more than 10% after hours.

On capex guidance, nothing moved. Alphabet kept its 2026 range of $195 billion to $205 billion (SiliconANGLE), and CFO Anat Ashkenazi said spending will "increase significantly" in 2027, with detail promised on the fourth-quarter call (AP). Some outlets have printed a specific 2027 figure. I could not find one in Alphabet's own documents, so I am not repeating it.

The numbers

Metric Q3 2025 Q3 2026 Source
Revenue $102.3B $127.6B Q3 release
Google Cloud revenue $15.2B $30.7B Q3 release
Google Cloud operating income $3.6B $11.1B Q3 release
Purchases of property and equipment $24.0B $51.4B Q3 release
Operating cash flow n/a $42.3B Q3 release
Free cash flow n/a -$9.1B Q3 release
Revenue backlog (Cloud share) n/a $567.7B ($560.8B) 10-Q
Contractual commitments n/a $802.1B 10-Q
Data center leases signed, not yet started n/a $130.5B 10-Q
Long-term debt $46.5B (Dec 2025) $125.1B Q3 release

The quarterly trend is the clearest part of the release. Capex went from $27.9 billion in Q4 2025 to $35.7 billion, $44.9 billion and then $51.4 billion. Over the same four quarters free cash flow went from $24.6 billion to $10.1 billion, then -$5.9 billion, then -$9.1 billion.

Alphabet capex per quarter ($B) and free cash flow Q4 2025 27.9 FCF +24.6 Q1 2026 35.7 FCF +10.1 Q2 2026 44.9 FCF -5.9 Q3 2026 51.4 -9.1 Shaded terracotta: quarters where capex exceeded operating cash flow. Q4 needs about $63B to $73B to land in guidance.
Purchases of property and equipment and non-GAAP free cash flow by quarter, from the free cash flow reconciliation in Alphabet's Q3 2026 earnings release.

There is also an arithmetic point in that chart. Alphabet has spent $132.0 billion in the first nine months of 2026. Hitting the $195 billion to $205 billion range means spending roughly $63 billion to $73 billion in Q4 alone, another big step up.

How Alphabet is paying for it

This is the part I think deserves more attention than the share price. Alphabet's balance sheet is no longer funding AI from cash flow alone. According to the release and 10-Q:

  • In June 2026 it sold 6.25% mandatory convertible preferred stock for net proceeds of $19.0 billion, which the filing says will fund "capital expenditures to scale AI infrastructure and global compute".
  • In Q3 it issued senior notes for $28.7 billion of net proceeds and sold 36 million shares through an at-the-market program for $12.4 billion, with $27.6 billion of that program still available.
  • On 5 November 2026 it closed a €6 billion euro bond in seven tranches running to 2045, with coupons up to 5.25%.

Long-term debt has gone from $46.5 billion at the end of 2025 to $125.1 billion. Cash and marketable securities also rose, to $265.1 billion, so this is a company raising money ahead of commitments, not one running out of it. The commitments are large, though: $802.1 billion of contractual obligations plus $130.5 billion of data center and energy leases that are signed but have not started, with start dates as late as 2031.

Why it matters

The demand side looks real on paper. A $560.8 billion cloud backlog, just over 55% of which Alphabet expects to recognise within 24 months, is a lot of signed revenue. Cloud margins of 36% show the business can be profitable at scale. And the 10-Q describes TPU systems sold with options for customers to make Google repurchase the hardware at pre-set prices, which tells you how hard Google is pushing to place its own silicon.

The supply side is where the risk sits. Alphabet has committed to almost a trillion dollars of purchases and leases on the assumption that this demand persists for years. Amazon reported a similar pattern the following day, with AWS growing 39% while trailing free cash flow went negative.

What sceptics say

The bear case is simple and it is in the filings. Free cash flow is negative for two quarters, and guidance implies a bigger capex quarter next. Alphabet now pays preferred dividends of $15.63 a share each quarter on the convertible preferred, and it is selling common stock, which dilutes existing holders. SiliconANGLE noted that free cash flow came in far below what analysts expected, and that the drop wiped more than $500 billion off the company's market value.

The bull response is also in the filings. Capex came in below the Street's estimate ($51.4 billion against about $55.1 billion, per SiliconANGLE), the backlog is contracted rather than hoped for, and Cloud profit tripled. A company with $265 billion of cash and securities is not in distress. The real question is whether a backlog this concentrated in AI infrastructure turns into cash on schedule.

What it means for engineering and DevOps teams

My take, as someone who buys cloud capacity rather than sells it: when a provider's backlog is this large, capacity goes first to the customers who signed multi-year commitments. If you rely on Google Cloud for GPUs or TPUs on demand, expect quota negotiations, reservation requirements and less flexibility on regions. I would budget for committed-use contracts and keep a second provider qualified for inference, the same way I argued in the neocloud economics piece.

Second, watch pricing. A company running negative free cash flow while signing $130 billion of leases has every incentive to protect margins. Discounts on AI services may get thinner, and your FinOps practice for AI spend should assume that unit prices will not keep falling as fast as they did in 2024 and 2025.

Third, power and siting. Many of those unstarted leases are for energy infrastructure, which is the constraint I described in the data centre power post. Region availability follows power, not customer preference.

FAQ

How much did Alphabet spend on capex in Q3 2026?

Alphabet's purchases of property and equipment were $51.4 billion in the quarter ended 30 September 2026, up from $24.0 billion a year earlier. Its 2026 guidance stays at $195 billion to $205 billion.

Why was Alphabet's free cash flow negative?

Operating cash flow was $42.3 billion and capex was $51.4 billion, leaving free cash flow at negative $9.1 billion. It was also negative in Q2 2026, at $5.9 billion.

How big is the Google Cloud backlog?

Alphabet's 10-Q reports $567.7 billion of remaining performance obligations, $560.8 billion of it for Google Cloud. It expects to recognise just over 55% within 24 months.

Related: AI Power in 2026: 485 TWh, a 0.3 GW Stargate, a 116 GW Turbine Backlog, and 1,800 MW That Dropped Off the Grid in Seconds, GPU Clouds in 2026: H100s at $2.59 an Hour, $104B of CoreWeave Backlog on $35B of Debt, and a Six-Year Bet on Silicon and FinOps for AI in 2026.

Spot an error or something out of date? Email me and I will fix it and log it on the corrections page.

I'm Ajay Kumar — I build and operate PandaStack, an open-source Firecracker microVM cloud for AI agents, and write about the infrastructure and tooling around it.

Need this kind of infrastructure work? See what I do or email hello@ajayk.sh.


Related