Cloudflare Q2 2026: The Numbers
Cloudflare reported Q2 2026 earnings after market close on August 6:
- Revenue: $696.1M, up 36% YoY
- Gross margin: 73.1% — first quarter-over-quarter improvement in 8 quarters
- Non-GAAP operating income: $96.1M, up 33% YoY
- Free cash flow: $56.4M, up 69% YoY
- RPO: $2.732B, up 38% YoY
- Large customers (>$100k annual spend): 4,698, up 27% YoY — net adds of 986 over the past 12 months, with record net adds at every tier ($100k, $500k, $1M, $5M+)
- Dollar-based net retention: 120% (+2pp QoQ, +6pp YoY)
- New paying customers: 80,000+, up 74% YoY
- Net loss: $205.7M, more than tripled YoY
- CapEx: 14–15% of revenue (~$430M), low for a cloud provider
- Agent gateway layer — ADLC + Agents + Cloudflare.pay make agent-era SDLC and payments platform defaults. The August 5 coverage (topicId 178594086) covered the ADLC trio separately; this earnings call framed it in financial terms.
- Training data layer — non-human traffic is the main future channel for training data collection. Controlling distribution and billing at this scale means pricing power over AI companies' next data costs.
- Chinese counterparts — Alibaba Qwen's integration into Apple Intelligence (topicId 178597110 follow-up), Ant's Ling-3.0-flash end-to-end open-source pipeline (Aug 9, topicId 178603081), and ByteDance SeedRealtime full-duplex (Aug 6, topicId 178597113) optimize the "model + inference + cache" chain; Cloudflare is doing the same at the network layer.
- Cloudflare Q2 2026 earnings call transcript (Yahoo Finance / Quartr): https://finance.yahoo.com/quote/NET/earnings/NET-Q2-2026-earnings_call-662709.html
- Zhitong Caijing earnings analysis (Chinese): https://www.zhitongcaijing.com/content/detail/1477439.html
- IT Home report, Aug 8 (revenue / loss / Prince quotes): https://www.ithome.com/0/987/438.htm
- Biggo Finance earnings data: https://finance.biggo.com/news/US_NET_2026-08-06
- Aug 5 ADLC + Agents trio coverage (topicId 178594086)
Full-year revenue guidance was raised to $2.864–2.870B (32% YoY growth).
'Humans Will Be a Rounding Error'
The real headline came from CEO Matthew Prince in the Q&A. Asked in November 2025 when non-human traffic would exceed human traffic, his team's data pointed to H2 2027. Re-run in March 2026, the answer moved to H1 2027. By May 2026, the team told him it had already happened.
Prince's quote:
> "In other words, humans will be a rounding error on the Internet, not because human traffic goes down, but that's just how fast we're seeing non-human traffic grow."
His bolder prediction: if the trend continues, non-human traffic will be 1,000x human traffic within five years.
What makes this plausible is a fundamental difference in agent behavior: a human shopping for a camera checks five retailers; an AI agent might check 5,000 sites. Millions of users simultaneously running comparison shopping, research, and summarization compounds traffic exponentially. Supporting datapoint: Cloudflare added nearly 2 million developers in Q2 alone — more than all of 2025.
Cloudflare's Repositioning as the Agent-Era Internet Backbone
Three pieces of evidence:
1. Agents Week: a full stack in one week
Cloudflare launched Cloudflare.pay, ADLC (Agent Development Life Cycle), and Agents together. Cloudflare.pay is a payment layer designed for agent traffic — each request through Cloudflare's network is billed at fractions of a cent, replacing the ad-defined business model. Prince was explicit:> "I think that the business model of the Internet for the last 27 years has been largely defined by advertising and really defined by Google. I think the business model of the next 27 years of the Internet is going to be very different."
2. 80%+ of major AI companies are already Cloudflare customers
Prince repeated this figure throughout the Q&A. The implication: every major model company and agent platform routes external access through Cloudflare, making it the gateway + billing layer of the agent era — analogous to Google's position in search advertising in 2003.3. CapEx discipline as a moat
At 14–15% of revenue, CapEx is low among cloud providers. Prince countered the "why not build GPU datacenters" question:> "We know what the typical kind of inference loads are at the hyperscalers, and their GPU utilization. It's super low." > "If we can, and in some cases, get 10 times as much utilization out of every CapEx dollar, then we can play these games and continue to grow revenue."
Cloudflare's model isn't selling GPU boxes; it's scheduling workloads onto existing infrastructure to extract 10x efficiency per CapEx dollar.
Points That Need Unpacking
AI traffic is not neutral. Prince split AI traffic in two: "For part of it, we actively block. We don't charge customers extra, because we think that's what a security company should do." The typical malicious case: AI companies scraping ad-funded media content. This draws a clear boundary — Cloudflare is both agent gateway and copyright moat.
Losses are widening — but by design. The $205.7M net loss reflects aggressive agent-era infrastructure investment, yet free cash flow grew 69% and guidance was raised. This is "CapEx investment loss," not "operating loss."
Geographic balance. US: 51% of revenue (41% growth); EMEA: 27% (30%); APAC: 14% (32%). No single-region dependency.
Context: The Broader AI Coding × Embodied AI Landscape
Three Judgments
1. The agent internet's business model is being forcibly defined. The last 27 years ran on ads; Cloudflare is betting the next 27 on pay-per-request. The real risk: agent traffic is far more complex than human traffic (5,000 sites vs 5), and Cloudflare must prove it can handle that volume without quality loss. The 10x CapEx utilization story is the answer — the moat is scheduling, caching, and protocol handling, not bare-metal GPUs.
2. "Humans as a rounding error" is a roadmap, not rhetoric. A 1000:1 non-human-to-human ratio means product strategy tilts toward agents: CAPTCHA replaced by agent credentials, Bot Management evolving from "blocking malicious bots" to "managing agents," CDN routing evolving from "page acceleration" to "tool-call acceleration." This quarter is the first time that pivot moved from narrative to numbers.
3. What it means for Chinese To B / agent startups. If you build agent products, "your agent's cost of calling third-party sites/APIs" becomes a core operating metric within three years. If Cloudflare.pay's per-request billing works, expect equivalent copies from Chinese clouds (Alibaba Cloud and Tencent Cloud edge networks already have the components). Calculating your product's "request bill" is more urgent than optimizing single-request latency.