Cloudflare
Thesis
- The business:
- Aggregator of customers' networks, monetized through delivery, security, and compute.
- Growth from selling more software over one network it already owns.
- Moat:
- Traffic scale that earns network terms and lowers data transit cost.
- 26.1% of all websites and 85.2% of reverse-proxy sites; broadest interconnection and peering arrangements with ISPs.
- The data loop that links traffic scale to better security (plausible but unverified).
- Two-sided network effect between agents and publishers (still early days).
- Fast growing developer ecosystem (mostly AI builders).
- Win on high utilization and lower cost (but at lower margin, unverified), and neutrality in competition with hyperscalers.
- One software stack on every server -> high operating leverage.
- Traffic scale that earns network terms and lowers data transit cost.
- Business model:
- Free tier for data (a "vast sensor network" and "virtual quality assurance"), funnel (for "graduating" into paid accounts), and bargaining power (for improving the breadth and economic terms of interconnections, bandwidth costs, and co-location expenses).
- Large customers and value added services (security and compute) for economics.
- AI edge-network-inference that wins on utilization and cost advantage.
- Financing: Growth is funded by zero coupon convertible notes with capped calls instead of buybacks to hedge dilution. Stock compensation is the dilution channel.
- Cyclicality: Most revenue is contracted subscription, which smooths out both up and down cycles.
- Growth drivers:
- Application security and delivery rides agent traffic. Cloudflare fronts 85% of reverse proxied websites, and agentic traffic is >50%.
- Developer and AI compute (Workers) is the fastest growing act.
- SASE sold through channel partners.
- Platform expansion inside large customers - more customers and more spending per customer.
- Improved sales capacity and productivity.
- Margins and cash:
- High gross margin and falling for two years due to more paid traffic and more developer products with below average gross margins. Operating margin held because much of the gross-margin loss was network cost that previously sat in sales and marketing.
- Cash conversion exceeded reported profit due to customers billing ahead of service, non-cash stock pay, and interest on cash.
- Capex "behind the demand, not ahead". Step-up in FY2026 reflects GPU deployment as demand shifted from AI training to inference.
- Competition, challenges and risks:
- Risk of margin squeeze as a result of hyperscaler bundling caps pricing for SMEs and memory and server inflation.
- Risk of correlated failure given the network scale.
- Leadership turnover at the growth engine. The President of Revenue who rebuilt sales leaves at the end of 2026, the President of Product and Engineering left in 2025, and the Chief Legal Officer resigned in Q1 2026
- What to monitor:
- net retention, large-customer adds, non-GAAP gross margin, stock compensation as a share of revenue and network capex intensity, every quarter.
Cloudflare versus Palo Alto Networks
- Cloudflare has faster organic growth with structural drivers, and higher average gross margin leveraging an owned network. PANW has better operating margin attributed to its larger scale.
- Cloudflare is the stronger business for the next 5 to 10 years, narrowly: Palo Alto Networks wins today on margins and contract lock-in, but Cloudflare's growth is organic and architectural while PANW's is increasingly bought.
- The load-bearing fact: at Cloudflare's current size, PANW had the same cost structure (S&M 52% of revenue, operating margin minus 10% in FY2017), so the 20-point margin gap is mostly scale, while the growth gap (about 35% organic vs about 14%) comes from things PANW cannot buy: a global network carrying a quarter of all websites and the agent traffic now running over it.
Cloudflare versus Akamai
- Growth: Cloudflare wins by selling more software over one network it already owns, while Akamai's new growth is contracted AI capacity that it must build with borrowed money for largely one customer.
- Power: Cloudflare holds the stronger power position: cost and developer moats that take years to copy, and price-setting across a diffuse base. Akamai's power is trust with conservative enterprises, which is real but defends share rather than expanding it.
- Power over customers: Akamai's future margin will be negotiated with one counterparty; Cloudflare's with thousands.
- Power over suppliers: Akamai is better protected against a supply shock for the capacity it has sold and the supply it has locked. The edge goes to Akamai only while demand holds; locked supply is a liability if contracts slip.
- Cost: Cloudflare runs the leaner network (COGS) and the heavier company (SG&A); Akamai runs the heavier network and the leaner company.
- Risk: Akamai breaks first in AI spending downturn. Akamai's exposure is balance sheet and counterparty risk. Cloudflare's exposure is correlated failure.
