The Invisible Hand of Product Agility and TAM Expansion

The agile software development process is based on four time-honored, yet simple, tenets codified in the Agile Manifesto. They emphasize responding to change, actively soliciting customer feedback and most importantly, shipping software as early as possible. The expected outcome is to speed up software development cycles and get relevant product into the hands of customers. Then, the cycle repeats. Test, observe, iterate and release again, always striving to increase the output of each subsequent “sprint”. Leading software infrastructure companies have taken this approach to heart, exhibiting a strong bias towards rapid product release cycles. This allows them to continuously expand their product footprint, enter new markets and stay far ahead of competitors. Agility is the beating heart of their product cycle.

The concept of a total addressable market (TAM) is well-understood by investors. It provides an estimate of the revenue opportunity available for a product or service. Companies often include this estimate in their S-1 filing, investor decks and quarterly reports. As investors, we will usually take a summary view, sometimes skeptical, of management’s sizing of the market and how they calculate that. As long as the TAM is big, we usually check the box and move on to historical financial indicators like revenue growth and profitability to evaluate our potential holdings.

Yet, product agility and addressable market expansion can have an impact on stock valuation, primarily for the forward view. These influences can manifest when the market perceives that a company’s TAM has expanded. This may be the result of a product announcement or strategic partnership. Even if the underlying stock’s financial metrics haven’t changed in the interim, the perceived potential associated with an expanded market opportunity can drive up valuation.

Similarly, when a company is rapidly launching new products and extending the functionality of existing ones, investors gain confidence that the company will successfully fill the addressable market. Agile companies always seem to have a new product announcement (or clusters of them), with beta customers lined up to use them. Even if the initial version isn’t complete, agile companies work with users to quickly flesh out the features that are important and discard those that aren’t. The team repeats the process, incorporating learnings and software components from the prior cycle into the next build. Because of this re-use, more customer value can be crammed into each cycle.

While some software companies embrace product agility, Cloudflare provides a textbook example. They are delivering the fastest product release cadence in software infrastructure, dazzling investors with each subsequent “Week” of releases. Not only are they rapidly iterating on existing platform capabilities, but they are entering new billion dollar markets with alarming frequency. Birthday Week delivered several major releases, which in aggregate likely doubled their TAM. And leadership claims the product development pace is accelerating from here.

In this post, I will explore the signals that investors can watch that indicate a bias towards shipping product and TAM expansion. I will conduct this investigation using Cloudflare as the example, providing updates on their recent product announcements and what this implies for future growth. In the abstract, these principles can be applied to any software infrastructure company, generating insight into why some maintain hypergrowth for years and others just seem to be plodding along.

An Example with Cloudflare

The signals of product agility and TAM expansion are best exhibited by using Cloudflare as an example. This analysis may also help investors rationalize the jaw-dropping valuation expansion coming out of Birthday Week. I won’t try to justify the extreme valuation multiple, which surpassed 100 P/S recently. Rather, I will try to elucidate what has changed over the last month and what that implies for the future durability of Cloudflare’s growth.

Many analysts, myself included, put forth models as to why or why not the stock should be assigned the relative valuation the market has conveyed to it. Valuations are usually based on some multiple of a financial metric, like revenue (P/S), EPS (P/E) or cash flow. The relative size of the valuation often correlates to the rate of growth in those financial metrics, with revenue growth being a strong indicator. Companies with high revenue growth rates (and gradually improving profitability) are usually rewarded with the highest valuation multiples.

Valuation multiples can also be compared between peers in an industry, with the assumption that two companies exhibiting similar levels of revenue growth and profitability should be assigned comparable valuations. This makes sense on many levels and fits nicely into consistent models that allow analysts to speculate on future price movement.

However, these metrics are backwards looking for the most part. They don’t capture what could be in the future. Given that the future is hard to predict, it is understandable that we analysts gravitate towards metrics that have been reported. And, we are usually right, as past performance often correlates to the future. Winners keep on winning.

In software, I think we can gain some additional insight into the future potential for a company’s revenue growth by looking at their current product offerings and estimating the size of the addressable market for those products. This gives a sense for how big revenue could eventually be. Industry analysts, like Gartner, will even predict the size of a market several years into the future for us.

However, current product set and size of market are just single data points, like the absolute size of revenue. As financial analysts, we like to examine the rate of change. We can quickly compare this quarter’s revenue to that of a year ago, and determine a growth rate. That growth rate then provides a sanity check for current valuation. Growth of other metrics, like customers, net expansion, margins, earnings, etc. all round out the picture.

I think we can overlook signals that are harder to measure, but likely contribute more to the future potential for a company than past revenue growth. Those signals are growth in product offerings over time and the implied expansion of total addressable market. While we are comfortable scrutinizing revenue growth rates, we pay less attention to expansion of the addressable market.

A company’s product offerings change with new product releases. Each release brings incremental features to existing products and can represent entry into a new market segment. If the company is listening to their customers, these releases should bring more sales, whether through net expansion to existing ones or initial deals with new ones. The pace of product releases, then, can provide a forward-looking indicator of future sales growth.

We investors get particularly excited when a company’s annualized revenue growth rate is accelerating. This occurs when the growth rate keeps increasing each quarter. Accelerating revenue growth justifies an even higher valuation, in anticipation that compounding will deliver much greater revenue in the future.

But, what about acceleration of product release velocity? If a company is releasing more product to customers each quarter and entering new markets at a faster rate this year as compared to last, shouldn’t that be factored into valuation? Further, if that acceleration is driving a product release cadence that far exceeds peers, does that help rationalize a valuation that doesn’t compare along revenue growth rates?

And with that question, let’s look at what Cloudflare has been up to.

Cloudflare Birthday Week

To provide a reference example of the effect of product agility and TAM expansion, we can examine what Cloudflare did at the end of September and the subsequent movement of the stock over the following weeks. This activity was outside of any financial announcements, so we can observe the cause and effect in relative isolation.

Birthday Week kicked off on Sunday evening, September 26th, with the Founder’s Letter. This provided an aspirational view of the future, highlighting Cloudflare’s commitment to team, culture building and their underlying mission, “to help build a better Internet.” The reference to helping acknowledges that Cloudflare doesn’t intend to, nor can they, do this alone. At the same time, they make it clear they aspire to reach their vision through innovation. For investors, the founders provided a signpost that their vision and scope goes far beyond the product perimeters we may have established.

What’s exciting is that the pace at which the Internet is getting better is accelerating. And, in turn, the pace at which we are able to launch innovative new products is accelerating along with it. As the Internet grows and acquires more capabilities, we believe we will continue to grow with it. An investment in Cloudflare is, fundamentally, we feel an investment in the Internet itself.

Cloudflare Founder’s letter, september 2021

Along with commentary that Cloudflare has set their sights on becoming the fourth major public cloud, investors can appreciate their broad vision. Suddenly, Cloudflare’s TAM isn’t just an amalgamation of several product lines in CDN, network connectivity, hosting services and security. Rather, it encompasses supporting software infrastructure for all activities that we perform on the Internet.

The other poignant aspect of this vision statement is the tip to acceleration in product releases. Cloudflare’s pace of product development over the past 16 months has been nothing short of remarkable. I take this back to Serverless Week in late July 2020, during which Cloudflare announced a number of improvements to their Workers product. These included a cold start time workaround, extended runtime allowances, lower costs, expanded language support, improved security, developer tooling and data localization. This was a lot for one release week. Surely, at this point, they were done with Workers for a while. Legacy software peers would wait a year for the next batch of announcements.

But, they kept iterating and releasing over the following months, adding Durable Objects (another data storage method), Cloudflare Pages, a partnership with Nvidia for AI on the edge, data management integration with FaunaDB/Macrometa, support for cron, and much more. And the platform is getting real-world usage, with management reporting 50k or more developers deploying code on Workers for the past few quarters, along with a number of reference enterprise customers building production applications with it.

I am focusing on development in Workers because this represents not just a TAM expander, but an example of strong product agility. In the Summer of 2020, it appeared that Fastly had fielded a comparable serverless edge compute platform with Compute@Edge. I even felt that Fastly had some advantages in architectural design and performance. From that point forward, however, Cloudflare just kept innovating and releasing, while Fastly seems to have stalled behind the curtain of a beta program. Granted, Fastly may have a body of work to announce at their annual user conference later this year. But, based on what we can see in public product releases and customer usage trends, Cloudflare appears to be far ahead.

This isn’t meant to be a comparison of Cloudflare to Fastly, but rather, recognition of the future-facing value of rapid product development. At that time, Fastly’s reported revenue growth for Q2 2020 was higher than Cloudflare’s, and they enjoyed a more favorable P/S multiple. However, the subsequent pace of product releases by Cloudflare provided signals for future growth. Cloudflare’s Serverless Week, Durable Objects, Cron Triggers all demonstrated a cycle of continuous improvement. Meanwhile, Fastly released little. While I remained focused on a single product capabilities data point in early 2020, Cloudflare demonstrated the value of an accelerating product release trajectory. If not better today, it will be soon enough.

If product coverage determines TAM, then rapid product releases lead to faster TAM expansion. Layering on new product vectors monthly, we can see how Cloudflare’s vision to make its addressable market the set of all Internet services is feasible. And management asserts the pace at which they launch innovative new products is accelerating.

With that picture, let’s get back to Birthday Week, because it is illustrative of how valuation can follow perceived TAM expansion. This example is useful because it occurred outside of any changes in financial performance or other announcements. In fact, NET stock was slumping the first few days of the week due to an analyst downgrade the Friday before.

Monday, Sept 27th

Email Security Suite. This represents Cloudflare’s entry into email security, offering users the ability to create custom email addresses, manage incoming email routing and prevent email spoofing and phishing. It also makes it easy for email administrators to configure SPF and DKIM for their domains, which is often overlooked, but creates issues with email deliverability. In addition to the linked blog post, Cloudflare highlighted their entry into email security with a press release on the Investor Relations page.

The basic email security service is being offered for free to existing customers. Cloudflare also announced an Advanced Email Security Suite, which will be part of the Cloudflare One commercial bundle. This service will provide full scanning of emails for security threats and work in front of any email hosting provider to supplement the anti-spam and security protections in place. 

As part of the release, Cloudflare’s CEO stated they “ultimately hope to become the leader in email security.” The email security market is already occupied by several independent players, most notably Proofpoint (PFPT), which was acquired by Thoma Bravo for $12B in August 2021. Statista estimates the market for cloud-based email security solutions to be about $1B in 2021.

Tuesday, Sept 28th 

R2 Storage.  This was a major announcement for Cloudflare, with a press release in addition to the blog post. It also prompted a well-publicized Protocol interview in which CEO Matthew Prince quipped “We’re aiming to be the fourth major public cloud”. This product release by Cloudflare is clearly directed at AWS and their unexpectedly high data egress fees. AWS’ S3 service is a foundational building block for many cloud computing workloads and arguably drove much of AWS’ adoption in its early days. High egress fees insinuate that AWS purposely makes it difficult for customers to move off of AWS. This is because their data footprint grows over time, making migration incrementally more costly with each year of use. Stratechery dug into this pain point and Cloudflare’s new position in a post on Sept 30th. The name R2 is not a Star Wars reference, by the way, but rather means “one less than S3”.

Specifically, R2 allows developers to store large amounts of unstructured data at the network edge. Object storage of this type is suitable for handling a wide variety of raw data, like media or log files, or even application-specific data sets. This isn’t a relational database, but could provide a permanent store for user or product metadata formatted in an object file, like JSON. This is all retrievable with consistent latency, high durability and extensive capacity.

To make migration from other hyperscalers easy, R2 is compatible with S3 API’s. R2 will be priced at $0.015 per GB for storage (less than incumbent providers) and egress will be free. Object access fees will be low as well, with no charge for infrequent object retrieval. Storage is multi-region, with automatic replication of objects to the location where they are frequently requested.  The system is designed for high resilience with eleven 9’s of durability.

Cloudflare is also planning to introduce jurisdictional restrictions to R2 storage to keep data within desired geographic boundaries.  This allows application developers to meet data sovereignty requirements and represents an advantage over incumbent solutions. With more and more countries considering data privacy regulations that require their citizen’s data to remain in country, geographic data storage controls become a necessary capability. Durable Objects offers this data sovereignty capability as well.

R2 will be integrated with the Workers platform, Cloudflare’s distributed edge compute runtime. This allows objects to be written and manipulated easily from the serverless environment. One of Cloudflare’s goals with R2 is to further appeal to developers considering building applications on the Cloudflare platform.  Beyond use for web apps, the capability would support data pipelines from the edge to the center of the network. That could be applied to use cases like IoT data processing and caching of smart device content. R2 is still under development, with a waitlist for access.  Cloudflare’s intent is to conduct early user testing and announce an open beta soon.  

To me, this further underscores Cloudflare’s commitment to build distributed state at the network edge.  Prior to R2, Cloudflare already had a few data storage tools, with Workers KV, Durable Objects and the Fauna/Macrometa integrations. R2 moves data storage capabilities even further ahead. To underscore the structured thinking around Cloudflare’s product agility, on a CloudflareTV segment, the product manager mentioned that R2 utilizes Durable Objects to store mapping meta-data. He shared that R2 development would have taken much more time if they couldn’t leverage Durable Objects to store the look-up tables for R2 object locations.

This notion of building “primitives” for developers is very powerful. It further highlights how Cloudflare is dogfooding their own platform and helps understand the rapid product development pace. If Cloudflare engineers can re-use their own software infrastructure components to build new features, then delivery cycles will be dramatically sped up. This idea of software component re-use falls under the design pattern of composability. This concept has a parallel in open source software, where developers can download open source packages to jumpstart a development project. More recently, the term has been co-opted by blockchain enthusiasts as they recognize the power of smart contracts. New blockchain applications can leverage the prior work embodied in composable smart contracts to shorten their development cycles.

With R2’s re-use of Durable Objects, Cloudflare is doing the same thing. Their product and engineering teams harness the same benefits of composability. They can pull existing software and service building blocks off the shelf to address new product opportunities and expand into adjacent markets. Each cycle delivers more building blocks, often larger than before.

Cloudflare leadership added that they have more big ideas for the future of data storage at Cloudflare’s edge. I think data pipelines represent another big opportunity. On the same CloudflareTV segment referenced previously, a product manager hinted at a future product offering called Queues. Message queues represent an adjacent data processing segment that allows developers to asynchronously send events from one system to another. As we think about enabling distributed systems to coordinate activities between them, message queues will be a useful ingredient.

I think R2 is an important step for Cloudflare, as it offers an object store, but on a globally distributed edge network. Cloudflare is quickly creating all the supporting infrastructure that a developer would expect in order to build modern software applications on Cloudflare’s edge platform. While we can debate the benefits of edge versus centralized compute in the near term for the current slate of Web 2.0 applications, if a developer can build the same application on Cloudflare’s distributed platform, then they may as well take advantage of the inherent speed benefits.

While the hyperscalers are pushing their compute and storage offerings out of regional data centers towards the “edge”, these largely resemble the architecture of their centralized data centers, just on a minified basis. For the most part, they are still single tenant, container-driven and cluster data storage. These mini data centers deployed locally generally lack some of the capabilities of their larger centralized peers, creating hierarchies of access rules. Developer service X is available at the edge, but requests for service Y and Z must route back to the central Region. Even Google’s newly announced Distributed Cloud offering reflects this architectural approach.

Cloudflare, on the other hand, offers an application development toolset that is identical across all 250+ PoPs. These are located within close geographic proximity to 95% of the global population, all across one network fabric with traffic dynamically routable between PoPs. The application processing runtime is serverless and multi-tenant, able to run the same code in parallel in every location. Data storage is similarly distributed with seamless access across all locations, with multiple options for data type and access pattern.

These capabilities don’t just distinguish Cloudflare’s cloud offering from the hyperscalers, but they provide an appealing path for hosting the next generation of distributed applications. Serverless is foundational to this and recent optimizations to the runtime allow it to address synchronous workloads. Serverless processing represents the most efficient allocation of compute, as the runtime only spins up for each request. This is contrasted with the typical container based deployment where the runtime typically idles until a new request comes in.

While R2 in itself didn’t represent the final lynchpin of capabilities needed to capture the breadth of cloud computing workloads, it might be considered a tipping point. Given that perspective, it may have allowed the market to add the roughly $200B in 2021 spend on cloud services infrastructure to Cloudflare’s TAM.

Wednesday, September 29th

Cloudflare for Offices. This garnered a full press release in addition to the linked blog post. Cloudflare will extend its network directly to 1,000 of the world’s busiest office and residential buildings. This brings network connectivity within microseconds of millions of workers and home users. Whether from home or offices, we can generally assume that large groups of people will be clustered in buildings in population centers going forward, in spite of COVID-19.

Besides providing a network onramp for organizations and individuals, this brings direct access to Cloudflare’s security services and enhanced routing packaged in Cloudflare One. If you have tried setting up Internet connectivity for your organization in an office building, it can be a pain. You have to work with one or more providers, purchase and configure network equipment and manage your own security settings. Connectivity can be spotty and vary based on location, particularly when routing to your company’s central data center or shared SaaS applications.

For the enterprise with multiple locations, data centers and some remote employees, Cloudflare One provides real savings. Provisioning the network connectivity and equipment to connect all entities can be very cumbersome, costly and inefficient. Remote employees are generally backhauled to the nearest office location to make a VPN connection, which could be on another continent. Larger office and data center clusters are connected through MPLS links, which are leased from telecomm companies. Each of these network connections is sized for maximum anticipated usage and paid monthly.

With Cloudflare One, all of these encumbrances are eliminated. What is strategically insightful about the positioning of this offering is that it goes beyond a “better security” argument and addresses the real-world cost of network connectivity. By themselves, security improvements through Zero Trust involve what can be a longer sales cycle as they require an incrementally better ROI case than what existing hardware-based solutions can provide. However, coupled with the more practical matter of setting up and paying for network connectivity through MPLS circuits, Cloudflare One’s network as a service creates a much more compelling value proposition.

Extending Cloudflare One a step closer to physical office buildings and residences makes Cloudflare’s $40B – $50B projected TAM for this service easier to penetrate. This release isn’t necessarily a TAM expander, like the others from Birthday Week, but does grease the wheels for Cloudflare One adoption.

Tucked into the end of the blog post was a very interesting note, which I found exciting. As investors may know, Cloudflare designs and builds the hardware boxes that they cluster at edge locations in their PoPs. They intend to deploy special purpose hardware to these 1,000 office and residential buildings. The hardware design will account for being located outside of a data center with considerations for environmental durability and physical security. While these boxes will start with network switching capabilities, they “plan to add compute and storage capabilities in short order.” The words compute and storage are linked to Workers and the new R2 data storage option.

This is a significant hint on a number of levels. First, it implies that Cloudflare will be able to deploy their own hyper-local network gear into smaller physical footprints and outside of data centers. This is similar to efforts by some hyperscalers, like the Wavelength, Outposts or Snow Family products from AWS. With custom hardware that can be plugged into any location with power and network connectivity, Cloudflare could extend the reach of their network to locations outside of buildings including cell towers, municipal infrastructure or transportation hubs.

What is different is that Cloudflare’s hyper-local deployments would presumably bring the same edge compute and storage functionality that is available within their PoPs. This would equally be multi-tenant and distributed, unlike the single tenant, clustered architectures that the hyperscalers appear to be deploying towards the edge. That implies that Cloudflare’s edge fabric will be more far-reaching than before and capable of extending compute and local storage as close to end devices as possible. As industrial IoT and autonomous device fleets proliferate, Cloudflare’s edge network will be well-positioned to offer a distributed data and processing backplane to facilitate communication and coordination across them.

Thursday, September 30th

Real Time Communications Platform, Live Video Streaming and Cloudflare TVaaS. And the hits keep coming. On Thursday, Cloudflare announced a series of products to enter the market for real-time communications (RTC) and video steaming at scale. They are even repurposing their popular Cloudflare TV broadcast system as a product that customers could adopt and set up for themselves (more composability).

First, Cloudflare’s Real Time Communications platform will provide a new suite of products designed to help customers build the next wave of real-time, interactive applications. Use cases include one-to-one video calling, group audio or video-conferencing. Cloudflare’s first step in tackling this market is through the launch of WebRTC Components. WebRTC is the most popular way to deliver a real-time communications experience to end users as much of the underlying infrastructure is addressed through open standards and browser APIs. These define how to make connections, transfer data and maintain security. This can all be delivered through a browser, whether on a desktop or mobile device.

Existing WebRTC-based services, whether self-managed or deployed on the cloud, typically rely on a TURN server to collect and relay traffic between peers, because managing full peer-to-peer connectivity at the client level would be nearly impossible to configure and scale. Cloudflare is addressing this by deploying a TURN server equivalent into every one of their 250+ globally distributed PoPs. This will bring RTC connectivity within 50ms of 95% of the world’s population, and even faster for those in the new Cloudflare for Offices locations.

One of Cloudflare’s advantages in this offering is that it leverages their ever-expanding global network of onramps. RTC communications in particular are latency sensitive, as anyone trying a Zoom or Teams call on a bad connection can attest. However, as Cloudflare pushes their edge network fabric closer and closer to end users, they can control how RTC traffic is routed once it enters their network. This routing can be performed dynamically and optimized for the audience on that particular call or event. This would provide an advantage over other RTC providers with fewer network entry points or those that rely on third-party cloud infrastructure.

As expected (more composability), Cloudflare is taking an interesting approach to this offering, by making the other capabilities of their platform available to WebRTC Components customers. First, it will be integrated with Cloudflare for Teams, which would enforce organizational policies for controlled access. Second, compute and data storage capabilities will be accessible for developers through Workers, Durable Objects and R2, enabling the creation of rich applications, running in a distributed fashion on the edge. Developers could trigger Workers on participant events (joining, leaving, etc.) or build stateful chat and collaboration tools.

Of course, there are many other providers in the WebRTC market. These include services that could be utilized by developers to build their own applications, generally through programmable APIs. Examples include Avaya, Cisco, Oracle, Vonage, Agora and Twilio. There are also popular implementations of WebRTC in consumer apps, like Facebook Messenger and Google Hangout.

WebRTC Components is being offered as a closed beta, with a sign-up available on the Cloudflare web site. At the time of this writing, the closed beta is already full, signifying customer interest for this service. Grand View Research estimates the global market for WebRTC services to have been $3.2B in 2020, growing to $40.6B by 2027, representing a phenomenal 43% CAGR.

If that’s not enough, the new Live Video Streaming service takes Cloudflare’s existing video streaming offering to the next level, by enabling live video recording, encoding and packaging on the Cloudflare network. These can then be streamed to any user or simulcast on another platform like YouTube or Twitch. The delay is about 15 seconds at this point, but Cloudflare intends to reduce this over time. Stream Live is available for customers now and billed on a pay-as-you-go basis, determined by the duration of videos recorded and amount of video viewed by consumers.

Finally, Cloudflare is taking their popular Cloudflare TV offering and exposing the software services for customers to leverage to create their own in-house video publishing network. Cloudflare has successfully utilized Cloudflare TV as a product marketing communications channel, broadcasting both live shows around company announcements and pre-recorded sessions of interviews, industry trends and product usage discussions. Their TV-as-a-Service offering will package all the tools and technologies that Cloudflare’s own internal team has used to run their broadcast network. Potential customers can sign-up to express interest in early use of the new offering.

I won’t try to size the markets for live video streaming or corporate broadcast networks, but assume these have potential. To demonstrate further composability and re-use of Cloudflare’s platform, Cloudflare TV as a service will utilize the newly announced RTC communications service and Stream Live as well as their existing Stream Connect product.

Friday, October 1st

Cloudflare Web3 Gateways. Saving the best for last, perhaps, Cloudflare ended the string of product announcements in Birthday Week with the most future facing offering. Recognizing that new paradigms for application development are emerging as encapsulated in the Web 3.0 movement and decentralized apps (dapps), Cloudflare wants to leverage their existing platform and maintain their relevancy for developers going forward. They announced the launch of a new Web3 product suite, which starts with web gateways for Ethereum and IPFS.

For Web 3.0 applications, Ethereum provides a processing layer, with a development environment and a distributed runtime. It can access a blockchain to permanently store an immutable set of records that are cryptographically signed. IPFS provides a convenient storage layer, organized like a distributed file system with content globally referenced by its hash. These two systems provide developers with building blocks for their distributed applications.

However, to bring distributed applications to web scale, supporting infrastructure is needed just like with the current generation of Web 2.0 apps. This infrastructure includes conveniences like network routing, content caching and application security. Cloudflare gateways will provide an API layer in front of IPFS and Ethereum to make it easier for developers to scale their dapps built using these technologies. To be clear, Cloudflare is not trying to replace Ethereum or IPFS, rather they are providing services to help developers use them. As with Web 2.0, Cloudflare just wants to provide the picks and shovels for developers to use.

The IPFS gateway will provide data caching, so that content fetched from IPFS can be delivered to the user from a location close to them (like a CDN) versus fetching from the IPFS origin server on every request. Also, Cloudflare will enforce a security layer over the content, protecting it from manipulation in transit. For the Ethereum gateway, Cloudflare enables secure and reliable access to the Ethereum network. This prevents developers from requiring users to run their own nodes or redirecting them to distant nodes through a JSON RPC API service. As with IPFS, the Cloudflare Ethereum Gateway leverages their global network to provide users with local access to resources on the Ethereum network.

Cloudflare announced that they already have an alpha version of these services being utilized by a set of early customers. The adoption metrics cited by Cloudflare are impressive, signaling both demand for the service and real-world usage.

Our current alpha includes companies that have raised billions of dollars in venture capital, companies that power the decentralised finance ecosystem on Ethereum, and emerging metaverses that make use of NFT technology.

In fact, we have over 2,000 customers leveraging our IPFS gateway lending to over 275TB of traffic per month. For Ethereum, we have over 200 customers transacting over 13TB, including 1.6 billion requests per month. We’ve seen extremely stable results from these customers and fully expect to see these metrics continue to ramp up as we add more customers to use this new product.

Cloudflare Blog Post, October 2021

With this announcement, Cloudflare is opening up access to the new Web3 gateway products as a private beta. Interested customers can sign up on the Cloudflare web site. As this technology is new, I won’t try to estimate the incremental TAM added by these offerings. The data cited above provides some measure of usage. At the very least, it is reassuring to see Cloudflare evolve to support developers migrating to the newer distributed application model. As Cloudflare is entering new markets as the disrupter, they are also watching their flank, lest they be disrupted by Web 3.0.

Market Reaction

Following these releases, let’s look at the subsequent price action in Cloudflare’s stock. I pulled the one month stock chart for NET, encompassing the period before and after Birthday Week (Sept 17 – Oct 15). Since these announcements occurred during the trading week, we can see the cumulative impact on the chart. Prior to this period, NET had been trading in a range of $120 to $135 since early August. As I was writing this, the stock stepped up again to a high of $180 on news of a partnership with Microsoft.

NET stock entered Birthday Week following a downgrade to a Perform rating by Oppenheimer on September 24th. Further, the analyst removed their $130 price target and shared a DCF model value of $110. The analyst cited competitive and valuation concerns. This downgrade created a catalyst for strong downward momentum for Cloudflare’s stock, as Birthday Week announcements began.

As the week progressed, the stock continued to drop even as exciting product announcement proceeded. By Thursday, it seemed that analysts were starting to take notice, as Seeking Alpha published an update. They called out the recent 18% price drop and the Oppenheimer downgrade, but also mentioned a positive comment from Needham about the email security market entry, calling that “just one more example of why Cloudflare will become a major company.”

At this point, we start to see NET’s price bottoming out. Incidentally, I added to my position here, after digesting the product announcements from the prior days, particularly the R2 release. By the week following Birthday Week, we can see that upward price momentum begins to take shape. The market had likely read the positive coverage on Stratchery by this point and considered the CEO’s comment about aspiring to be the “fourth” hyperscaler.

During this period, there were no financial releases or other information relative to the Cloudflare’s fundamentals. To be fair, other software stocks began an ascent around this time, but the change in Cloudflare’s sentiment due to Birthday Week had a magnified effect. Since the low of around $112 on September 30th, the stock has appreciated about 60%. This is significantly greater than the price increase in other hypergrowth peers like DDOG, CRWD and SNOW.

Therefore, the only conclusion we can draw is that the price action was an outcome of the announcements made during Birthday Week. A segment of the market recognized the significance of what Birthday Week represented. I boil it down to two factors. First, an optimistic view would conclude that Cloudflare added somewhere around $100B to their long-term TAM. Recall the slide from Investor Day where they projected a TAM of $86B by 2022. The incremental TAM contribution of email security, WebRTC and broader cloud compute penetration doubles this over several years.

Second, they demonstrated that their already blistering product release pace is accelerating. While Cloudflare had been gradually increasing the pace of product development over the past 12 months, Birthday Week represented a watershed moment. The amount of innovation packaged into 5 days of release announcements eclipsed anything in the past. And management telegraphed even more to come.

The combination of these two factors is what I think is driving up the valuation for NET following Birthday Week. I can’t say whether the current valuation around 100 P/S is justified or whether it will go up or down from here. What is clear, though, is that the accelerating rate of product development and TAM expansion should provide the fuel for sustained hypergrowth for longer than expected. It will allow Cloudflare to outmaneuver competitors and hold off encroachment from the hyperscalers.

We can see evidence of the product development acceleration by comparing the magnitude of product releases in this year’s Birthday Week, versus last year. In 2020, Cloudflare held Birthday Week from September 28th to October 2nd. It also started with a Founder’s Letter on Sunday, followed by a bevy of product announcements from Monday through Friday. These included:

• Durable Objects – Data storage for Workers with strong consistency
• Cron Triggers – Ability to trigger code on Workers based on a schedule
• Web Vitals in Browser Insights – Provide page load time data to site owners
• Free, Privacy First Analytics – Basic web site analytics without tracking users
• Cloudflare Radar – Internet-wide reporting on security, usage and performance
• API Shield – Control client access and proper usage patterns for external APIs
• Automatic Platform Optimization – Intelligent caching for WordPress sites

Durable Objects added a significant enhancement to distributed data storage for Workers, bringing a solution with strong consistency to supplement the existing key-value store. API Shield provided an addition to Cloudflare’s application security suite. APO generates a little revenue and Cron Triggers provided an often requested feature for Workers. The Analytics products made it easier for web site owners to understand activity on their sites and tune performance.

These were all noteworthy releases in themselves and increased Cloudflare’s usability and visibility. They were somewhat incremental to TAM and Cloudflare’s perception as a leading infrastructure provider for the modern Internet. Several of the offerings were free to existing customers as well, exhibiting Cloudflare’s position as benefactor of the Internet.

Yet, Birthday Week 2021 eclipsed 2020’s impact by an order of magnitude. Product releases in 2021 represented whole new markets or significant additions to existing large product offerings. As I mentioned, some of Birthday Week 2020’s capabilities provided building blocks for 2021, like Durable Objects for R2. This composability and re-use is driving the acceleration of product development going forward. The development time for each new innovation is reduced as developers can incrementally pull more production-ready components to assemble into the next product offering. Workers is often the cornerstone of many new application offerings for Cloudflare, and now R2 provides another significant building block.

It’s also worth mentioning the value of the analytics services launched in 2020. While not direct revenue generators at the time, they provide a baseline for future analytics products. I wouldn’t be surprised to see Cloudflare move into some form of digital operations management, whether business intelligence or observability (or both). Data collected through Radar and these analytics services is also informing Cloudflare’s security suite, enabling more effective threat identification, monitoring and protection. As Cloudflare processes more of the Internet’s traffic across its pipes, it will have a bird’s-eye view into all threat activity.

Given the increase in product release heft, it is easy to understand why Cloudflare leadership can assert that the pace of product development is accelerating. The velocity thus far in 2021 is exceeding that of 2020. Leadership expects that velocity to increase further looking forward. Given this, it’s hard to imagine what new innovations will come in 2022 and the years beyond. It’s safe to assume that big announcements will occur with increasing frequency. I am already hard-pressed to finish reading each day’s blog posts.

Software Product Cycles are Speeding Up

Investors and analysts alike are flummoxed by Cloudflare’s rising valuation. I think a lot of the difficulty has to do with the challenge of modeling future financial projections. Typical valuation models are based on some future estimate of cash flow or revenue, an assumed exit multiple and then extrapolation of the market cap growth rate that would occur from current to future time. These generally hinge on revenue growth rates.

These models work neatly for companies that stick to one market, or perhaps add another in a predictable way. In those cases, revenue estimates can expand into the market size with decreasing growth rates as the market becomes saturated. However, how does that work if the company keeps expanding their total addressable market through rapid product development? While revenue growth rates might decay over time in early markets, they would be robust in new ones. This sets up a situation in which high revenue growth can be sustained for much longer.

Such is likely the case for Cloudflare. With tentacles continually expanding into new markets and a reliable (management even says accelerating) product release cadence that ensures they quickly move beyond a basic feature set, it becomes easier to envision a company that could grow revenue by 50% annually forever. Granted, nothing is forever, but the compounding effect of growth at that rate over many years can make a 10-year DCF model look appealing at the current high valuation.

It also helps when industry prognosticators are highlighting shifts to your favored technology approach. Having a large TAM is one thing. It’s another when it appears that your company has bet on a forward-thinking architectural transformation and the rest of the industry starts to pivot your way. Such is the case with serverless, which is slowly gaining steam as the preferred method of provisioning cloud-based workloads.

Some industry pundits are even calling the migration to serverless “inevitable”, as long time technology innovator and futurist Simon Wardley did on a recent episode of Serverless Chats. If all centralized compute moves to a serverless model, then distributed solutions based purely on a serverless architecture stand to grab more share. Granted, this transition will take time and single-tenant centralized serverless solutions (hyperscalers) will continue to retain some large enterprise customers. Nonetheless, this trend provides a significant tailwind for Cloudflare, particularly with newer, cloud-native companies. We see not just a large addressable market for Cloudflare, but tailwinds in technology evolution pushing demand their way.

As investors, I don’t think we have adequate tools or even data to model the impact of product release velocity on TAM expansion and eventually revenue. Yet, TAM expansion will have material impact on future growth. If anything, we should look for signals that indicate a TAM expansion and anticipate contribution to future revenue streams. This consideration for future growth can make valuation metrics like P/S ratios inflate in the near term, even while peer comparisons on historical revenue growth rates don’t justify the valuation increase.

This exercise was easier in the past, pre-Information Age. We could look at a company like General Electric for an example. At its height, it was the world’s most valuable company and generated over $180B of revenue. GE is a multi-national conglomerate with products in far-ranging markets spanning power systems, aircraft engines, appliances, health care, turbines, materials science, the industrial Internet, finance and media. Their leadership tactics were celebrated and codified in Jack Welch’s 1998 book about “the GE Way”.

While GE entered and dominated multiple industries, the product development pace was relatively gradual. On the History section of the GE web site, they provide an interesting graphic that lists all the market segments that GE entered over their entire history, with small circles marking major product announcements, acquisitions and corporate accomplishments. While the timeline is filled with about one hundred events, these are spread across 150 years.

Looking at the graphic, we notice the clustering of events and more activity in recent years. While the frequency of events appears higher, major product or market events are still spread out by a year or two due to the extended time scale. As we would expect, many of these markets involve physical products that require factories, laborers and supply chains to build them. In the world of jet engines or MRI scanners, entering one market every few years is fast.

In the digital world, however, these timelines can be compressed. With the speed of software development and composability of core building blocks, the timeframe to bring a new product to market can be significantly accelerated. Much of the supporting infrastructure is digitally fungible and many steps can be automated. Packaging and distribution of finished product is handled by other software systems, with whole assembly workflows initiated by a button click. Thousands of customers can receive a product update at once, and provide valuable feedback in hours.

Coupled with these speed advantages, the addressable market for software infrastructure products in the digital world can be as large as GE’s markets. For example, the global market for aircraft engines is projected to reach $85B by 2027. This is about twice the size of WebRTC’s projection for the same time frame. And WebRTC was just one of several product announcements during Birthday Week.

With that in mind, how then do we model a company that adds new multi-billion dollar markets monthly, or even enters several new large markets over the course of a week? That, I think is the key difference for hypergrowth software companies that deliver new digital products on top of a solidified platform of re-usable software components and services. They can quickly launch new products by assembling building blocks into unique offerings that target a market segment. This creation process isn’t captured in a DCF model.

Further, the resources and capital investment necessary to enter new product categories is diminishing. While GE might require a factory and a couple hundred workers to build a prototype for a new home appliance or jet engine, software companies can deliver a beta product with a small team of engineers. Amazon famously talked about the ideal team size being what could be fed with two pizzas.

In a recent blog post about how Cloudflare can launch products so quickly, the Chief Product Officer commented on their approach. She referred to Amazon’s analogy and reduced it further to “split a Twinkie”. So, that’s maybe 2-3 people to build a prototype for a billion dollar market? Mind-boggling. She also tips her hat to the Workers product and the composability of Cloudflare’s platform.

While others in the industry talk about building “pizza box” teams — teams small enough that they can all share a pizza together — many of our innovation efforts start with teams even smaller. At Cloudflare, it is not unusual for an initial product idea to start with a team small enough to split a pack of Twinkies and for the initial proof of concept to go from whiteboard to rolled out in days.

Jen Taylor, Cloudflare Chief Product Officer

We could point out that competitors or start-ups could also do this. However, I think Cloudflare is uniquely positioned because their current product development velocity is driven by the aggregate of their platform and prior software services built to date. These are purposely designed for composability, meaning they can be re-used to form new products (which are then composable). As most software engineers can appreciate, building a new software application from scratch takes much longer than assembling composable building blocks pulled from the company’s software shelf.

What’s more interesting and provides a significant advantage for Cloudflare is their millions of free users. On the surface, a free tier might appear to be a drag on profitability. Cloudflare doesn’t make any money from these millions of web sites and users on their platform. However, there is a much larger, unappreciated benefit to their product development velocity. Those users provide a ready set of product testers. They pressure test and signal interest for new product experiments. While a product is in prototype or alpha mode, they can rapidly provide actionable feedback. That feedback can be incorporated into the necessary polishing to get a minimum viable product out the door.

The idea of a MVP in software development has been long understood, but Cloudflare has used this concept to their advantage by leveraging their free user tier. Competitors without an army of free users are at a severe disadvantage. Rather than focusing on the boon for product velocity that these users provide, those companies focus on the cost of “free”.

A minimum viable product (MVP) is a version of a product with just enough features to be usable by early customers who can then provide feedback for future product development. A focus on releasing an MVP means that developers potentially avoid lengthy and (ultimately) unnecessary work. Instead, they iterate on working versions and respond to feedback, challenging and validating assumptions about a product’s requirements.

It is utilized so that prospective entrepreneurs would know whether a given business idea would actually be viable and profitable by testing the assumptions behind a product or business idea. The concept can be used to validate a market need for a product and for incremental developments of an existing product. As it tests a potential business model to customers to see how the market would react, it is especially useful for new/startup companies who are more concerned with finding out where potential business opportunities exist rather than executing a prefabricated, isolated business model.

Wikipedia, Minimum Viable Product (MVP)

The definition of an MVP from Wikipedia provides insight into an important aspect of Cloudflare’s product development secret sauce and their ability to push products through the idea, prototype, beta, GA cycle so quickly. Their free tier users provide the fuel for Cloudflare’s product development velocity. So, it’s not truly free, rather a very under-appreciated value exchange. This is particularly evident if we shift our company valuation focus from backwards looking financial metrics to forward looking product development velocity and new market entry.

Because the time to market for new products is so compressed, we can’t model the revenue impact of future market pursuits in 12 months simply because we don’t know what they are. We can only judge the company’s growth potential based on the arc of product development and the direction they are pointed. In Cloudflare’s case, major product releases every month directed at “helping to make a better Internet” leaves a lot of room for TAM expansion. As revenue follows TAM for well-run companies, we can start to understand why forward valuation multiples can inflate beyond what would be justified by the last earnings report.

Signals

I have discussed the influence of a rapid product development cycle feeding TAM expansion with Cloudflare as the example. I think investors can recognize the take-aways here, but let’s briefly summarize and flesh out a few of the signals investors can monitor.

Product Development Cadence

Product agility is based on the tenets codified in the Agile Manifesto. They emphasize responding to change and meeting customer demands. The outcome of an agile development process is the ability to ship products much faster than before. This allows the product team to gather real-world user feedback as quickly as possible. Then, refine, iterate and release again. Many leading software infrastructure companies have taken this approach to heart, exhibiting a bias towards rapid product releases. As investors, we watch leading companies expand their product footprint and enter new markets. Agile development principles are at heart of their product cycle.

As discussed, Cloudflare has taken the agile approach to the next level. They not only maintain a blistering product release cadence, but that pace has been accelerating. They continue to push out new releases, expanding into ever new product directions and markets. This is facilitated by their strong platform foundation and composable software components that provide the building blocks to rapidly assemble new product offerings. With their millions of free users, they have a ready audience to test fledging products with low risk, providing valuable feedback on market interest and usability.

Product development pace also provides a perspective on competitive position. The independent software companies that I cover are all competing with the hyperscalers in one fashion or another. The hyperscalers certainly represent a formidable presence, based on their financial resources and customer penetration. Investors sometimes overestimate the threat from hyperscalers or panic if one enters an adjacent market.

I think product agility and release cadence provides smaller independent companies with a clear advantage in individual market segments. This likely explains why the hyperscalers are increasingly ceding major segments of cloud infrastructure to independents – examples like observability (Datadog), data cloud (Snowflake), transactional databases (MongoDB), CPaaS (Twilio), Identity (Okta) all come to mind. Hyperscalers generally release new products on a much slower cadence, often coinciding with an annual user conference. Given their size, inertia, culture and decision-making bureaucracy, this is understandable. Additionally, their dominance creates an Innovator’s Dilemma, which serves as a counterweight to their further market disruption and hampers creativity.

While periodically they make a major product announcement, we have to wonder when the next one will be? In another year? Agile independents are launching quarterly or sooner. Over the next year, Cloudflare will likely iterate 5-10 times on their edge network, compute and storage solutions. Going back to the comparison with Fastly, a view of the current product snapshot isn’t revealing. Rather, the pace of product release cycles is the proper signal.

As a way to supplement fundamental analysis for software companies, investors may want to consider tracking a company’s pace of product development. If they are frequently making new product announcements, that is a positive signal. If you find yourself reacting to a new product announcement with the thought that the company just launched something major last week and wondering how could there be another, that is a signal. If this release appears to represent a TAM expansion, even better.

We could come up with quarterly metrics like releases/quarter or TAM added in the quarter. This is a little subjective and would require some investor work, as most companies don’t report the number of product releases that matter. Some companies talk about the number of production or feature releases. That is pointed in the right direction, but doesn’t tell the whole story. A production release could be a single line of code to fix a bug. A feature might be something mundane that doesn’t add customers or TAM.

As much attention as we pay to annualized revenue growth, a view of product release velocity would be useful. This would need to include a measure of the size of the release, lest we count bug fixes. A concept like story points in agile development might be useful. It would be great if companies began reporting the relative size of releases – again agile provides a method for this in T-Shirt Sizing. Using that model, Cloudflare’s R2 and CF for Offices releases might be Larges, with Email Security and WebRTC as Mediums. It’s a little subjective, but investors can make the intuitive mental leap.

I think we can generally assess the magnitude of a release, particularly if we think about what is new and its orientation towards a target market. The key is not to focus on the exact size of each one, but the frequency. If a company keeps delivering new product innovations at an increasing rate, that is a strong signal that they will iterate their way to a larger TAM, more revenue and a larger valuation.

As investors, we can observe a few channels for signals of product agility and release frequency. Additionally, evidence of product adoption, like references to customer usage, provide confidence that the company is not just releasing quickly but delivering products that the market values.

Company Blog

I think a company’s blog is the best source of information. An active and curated blog with frequent, inspiring posts is a strong signal. The blog should not just speak to new product releases, but also provide background information on the industry, customer usage patterns and future directions. An easy metric to collect is the number of blog posts over a period. As an example, Cloudflare might have published one post a day a year ago. Now, there are several.

Formal Press Releases

To supplement product announcements on the company blog, investors should pay attention to formal press releases on the company’s investor relations site. When a company expects a new product announcement or market entry to have a more tangible impact on future financials, they often duplicate the announcement on both the blog and the investor relations site. The two posts will have different content, catered for their respective audiences. Investors should read both, but give more weight to announcements that make it onto the IR site.

For example, with Cloudflare’s Birthday Week announcements, the IR team chose to issue a press release for R2, Cloudflare for Offices and Email Security. This likely reflects that they felt those products represent the most measurable potential impact on future financial results. Investors could pay attention to the frequency of these types of press releases.

Partnerships and Peers

Another signal of product development progress and industry perception is to watch for announcements or commentary from other companies relative to your company of interest. If you see your target company getting mentions in another company’s press releases, quarterly results or partnership engagements, that provides a mechanism to triangulate to the innovators.

We can look to some examples of this concept. Cloudflare recently announced that Microsoft and several other search engines would be collaborating with Cloudflare on a mechanism to reduce the amount of bot traffic on the Internet by indicating to crawlers when content had actually changed. Crowdstrike often refers to their partnership with Zscaler on earnings calls, citing go-to-market wins enabled by one-another. Snowflake has been announcing industry specific Data Clouds, which include major players in media and financial services. The key for investors is not to look at each announcement in isolation, but the aggregate of inclusion of their target company within broader industry initiatives.

A Note on Leadership

The key to pursuing a large TAM and executing through it is grounded in strong leadership. The executive team sets the standards, shapes the culture and encourages all employees to bring their best effort each day to win. The CEO, of course, leads the leaders. Their most important function, I think, is to provide a vision for the company and chart a path to get there. One that is ambitious, exciting and bold. This inspires employees, and as importantly, attracts new talent who share the vision.

For software companies, my preference is generally to have a technical founder who is still leading the company. Ideally, this is as CEO, but at least somewhere in the C Suite (often CTO). This is important because the founder/CEO retains the original context for starting the company. They often “scratched their own itch” and intimately understand the value being created for customers. Additionally, having a hand in building the first few iterations of the product provides a true sense of what is involved in creating a great product, including delivery timelines and feature completeness. This insight empowers the founder/CEO as the organization grows to push back on the product and engineering teams, questioning elongated delivery cycles or feature sets that miss the mark.

For this role, I also like to see founders of software companies who were coders themselves. As the building blocks of a software company are lines of code, developers understand the process of gathering requirements from users, designing systems, testing and following up on a release. They have lived the agile development process and gotten their hands dirty. They appreciate the benefit of re-use through composability. The normal path to coding is through a university program that conveys a CS degree. Of course, some great coders are self-taught as well. Tobi Lutke at Shopify is a great example.

Some examples of great technical founder/CEO’s of software companies include Jeff Lawson, Olivier Pomel, Eric Yuan and Todd McKinnon. There are others. These individuals all have CS degrees, coded for a living at one point and even held engineering leadership roles at other companies, before venturing out on their own. Their backgrounds prepared them for the analytical rigor of designing software products, building complex systems and applying technology to solve problems. They also gained an appreciation for customer collaboration, responding to change and rapidly delivering working software, as a consequence of exposure to newer software development methodologies, like agile.

From my perspective, this grooming is necessary to found and scale a successful software company. CEO’s without this skillset will have to depend heavily on their leadership team, particularly in the areas of product and engineering. Further, their ability to set a vision will be limited by their comprehension of the underlying technology. Fundamentally, a technology visionary determines how to apply available technologies to improve an industry. It combines the art of the possible with an acute sense of what is needed.

It also requires a dose of reality and the ability to distill the technology into digestible bites for customers to consume. As a for-profit business (not a university research project), the solutions delivered must cater to the market demand. They need to be comprehensible and practical. Products need a feature set that allows customers to address their problems, but they also don’t require over-engineering or perfection. New products can beta tested, refined and iterated. Finding this balance between over-polishing and getting real-world customer feedback is critical. Many software projects have failed by building for months or years in isolation before letting users play with it. Agile methodologies taught us this.

With this necessary balance of technology comprehension and judicious application, Cloudflare’s Matthew Prince may be the best practitioner in the room. While he has a more diverse background than other software company founder/CEOs, he is a builder, assembling the Unspam email protection service before Cloudflare and even contributing some early PHP code to the Cloudflare platform. Equally important, he understands product marketing and strategic thinking. He was a pupil of Clayton Christensen at Harvard Business School and embraces many of the principles of disruptive innovation.

This background enables Prince to appreciate both worlds. He can discuss the mechanics of Internet scale with the software engineers, but also orchestrate the product strategy to meet real customer demand. The grounding in market disruption theory allows him to appreciate competitor strengths and their vulnerabilities. Going up against AWS is possible if you recognize the inherent Innovator’s Dilemma.

As an example of his insight, during Serverless Week in July 2020, Prince shared his theory of the “Hierarchy of Developer Needs”. In his view, when considering the requirements for an edge-based serverless platform, like Cloudflare Workers or Fastly’s Compute@Edge, the order of importance for product capabilities is grounded in compliance and works up a pyramid in decreasing consideration to ease of use, cost and consistency, with speed at the top. Optimizing for processing speed can be useful, but only after the other layers have been addressed.

At the time, I was puzzled by this perspective. We are talking about software systems, after all. Performance optimization is kinda the point. Personally, I would have inverted the pyramid, with performance as the most important and compliance as the after-thought. That’s the software engineer in me.

However, Prince recognized that market adoption is the point and many companies have failed by iterating too much on performance and not enough on what customers think is important. This is the foundation of product agility. It allows that some product features can achieve a threshold of being “good enough” and further optimization is a waste of effort. A 35 microsecond cold start time is outstanding, but 3-5 milliseconds is more than sufficient.

Ease of use and compliance considerations are larger drivers of customer adoption. By leveraging the V8 Engine for the Workers runtime, customers could write edge applications in the language they prefer, most notably Javascript. Cloudflare’s Data Localization Suite allows customers to meet data locality, privacy, and compliance needs. This is particularly important as governments increasingly regulate the use of their citizens’ data. Compliance capabilities built into Cloudflare’s platform future-proofs their product for customers. Looping back to speed, Cloudflare even added a creative workaround that minimized the impact of cold start time. This was done by a clever hack, rather than a complete rebuild.

Of course, Prince has been aided by a talented team of executives and the other two co-founders of Cloudflare who rounded out the operational and technology roles. The insight for software company investors is to pay attention to the leadership team both for indications that they understand the technology underpinnings, but also possess keen insight into how to penetrate their market. The hypergrowth software infrastructure companies that we track are natural disruptors of the status quo. A leadership team that understands disruptive innovation enables them to navigate the competitive landscape and make thoughtful trade-offs in bringing new products to market.

Investor Take-aways

As I write this, Cloudflare’s P/S ratio hovers around 100. Using traditional valuation metrics, it’s hard to reconcile the stock price for NET, particularly as it relates to peers. Revenue growth, margins, profitability, customer activity all correlate to lower valued software companies. Yet, the market sees something to justify a much higher valuation for NET, particularly coming out Birthday Week. It may be that future facing signals, like product agility and TAM expansion, are signaling much more growth to come. If not an acceleration of revenue growth near term, at least durability of high growth for longer than expected.

I think the change for Cloudflare that has occurred is a significant expansion of its addressable market. While this has been happening for the past year or so, the releases and commentary coming out of Birthday Week put a fine point on it and convinced the market that a broader growth opportunity is tangible. Aspiring to become the fourth public cloud helps.

Looking forward, a larger TAM implies a larger revenue opportunity in the future. Cloudflare’s rapid product development cadence represents their secret sauce and provides confidence that the addressable market expansion will result in real penetration. Not only has Cloudflare’s release pace over the last 16 months been blistering, leadership asserts that this pace is accelerating. A comparison of Cloudflare today to the Cloudflare of July 2020 looks like two different companies. What, then, does that imply for 1-2 years from now?

Given our understanding of industries that build physical products, this kind of change is hard to imagine. As investors, we are anchored in prior models of market expansion, where new market entry required building factories, hiring an army of workers and launching stand-alone divisions. GE provides a good example. At their peak, they covered a broad landscape of product categories, but these were built up over 100 years. Multi-billion dollar market segments could be launched every couple of years at best. Even more recent software companies, like IBM, Microsoft, Oracle and Cisco, executed faster than capital-intensive industries, but still organize large product launches around annual events.

But what happens if a company can enter a new multi-billion dollar market every month? I realize this is another “this time is different” thesis. However, with intangible assets like software, the building blocks are virtual and composable. Small teams of software engineers can assemble product components into new offerings, leveraging the work done in the prior iteration. Coupled with a broad vision, ingrained and repeatable processes, a release-focused culture and millions of “free” beta testers, it is conceivable how this could happen. Maybe the software equivalent of GE can be built in a day. In a world of Capitalism without Capital, that can be an investment thesis by itself.

That is the opportunity as I see it for Cloudflare. I am not looking for revenue growth to suddenly accelerate over the next few quarters. That would be great, but it’s just as likely growth continues in the 50% range. However, with addressable market expansion and ever-shortening product development cycles, that high growth rate is sustainable for a much longer period. Maybe forever. That compounding over time will provide the justification for a high valuation.

A focus on product development agility and addressable market expansion can be applied to evaluating other hypergrowth software companies. The ingredients are evident – an extensible platform, a bias towards release, strong leadership and a broad vision. Leaders in other software categories exhibit these traits. Examples include Datadog, Snowflake, Elastic, Crowdstrike, MongoDB, Twilio and others, to some extent or another. Investors can watch the pace of product releases and new market entries for signals to gauge their trajectory.

Yet, Cloudflare may be the granddaddy of them all. If we take a trader mindset and focus on just one chart, then the product release trajectory would represent the underlying story for NET. The frequency and magnitude of product releases is both staggering and accelerating. Putting aside financial performance, competition with hyperscalers, enterprise sales ramp and all the other factors, accelerating product development makes Cloudflare a very interesting story. On this factor, Cloudflare is out-performing all peers.

NOTE: This article does not represent investment advice and is solely the author’s opinion for managing his own investment portfolio. Readers are expected to perform their own due diligence before making investment decisions. Please see the Disclaimer for more detail.

Additional Reading / Research

• Cloudflare Birthday Week – Summary Page, Blog Roll-up, CloudflareTV
• Protocol Article on R2 Launch
• Cloudflare Serverless Week blog post
• The Theory of Disruptive Innovation from the Christensen Institute
• Hhhypergrowth’s extensive coverage of Cloudflare – See Flare Ups, What are Edge Networks, A Cloudflare Deep Dive. Subscribe to the Premium Service for more great stuff.