Fastly (FSLY) Q2 Recap

Fastly (FSLY) reported Q2 earnings on August 5, 2020. On the surface, the performance was impressive, beating estimates on both the top and bottom line and raising Q3 and full year guidance. While Q2 revenue accelerated 24% sequentially over Q1, the market was looking for greater outperformance. This was evidenced by the stock’s nearly 18% drop the following day. After the 5x run-up in FSLY stock since the beginning of 2020, there was likely some profit taking as well. As part of the earnings release, Fastly management provided updates on the Compute@Edge beta and discussed a number of customer wins. In this post, I review Fastly’s Q2 earnings and other business updates that occurred during the quarter. I also dig into their evolving product offerings and revisit the competitive landscape. For a refresher on my prior coverage of Fastly, please see my original analysis, Compute@Edge deep dive and past quarterly updates.

Headline Financial Results

• Q2 Revenue of $74.7M, up 61.7% year/year. This compares to the consensus estimate of $71.4M, which would have represented growth of about 55.3%. Q1 Revenue growth was 38%, so Q2 delivered a dramatic sequential acceleration. Q2 FY2019 Revenue growth was 34%.
• Q2 Non-GAAP EPS was $0.02 vs. ($0.01) expected, representing a beat of $0.03. This compares to ($0.06) in Q1 and ($0.16) in Q2 2019.
• Q2 Non-GAAP operating income was $1.8M, representing an operating margin of 2.4%. This compares to an operating loss of $9M in Q2 2019 and an operating margin of -20%. Q1 2020 operating margin was -9%. Positive operating margin in Q2 is the result of revenue outperformance and improvements in operational efficiency.
• Cash used in operations was $8.8M in Q2 and capital expenditures were $3.1M, generating FCF of ($11.9M) in the quarter. This yields a FCF margin of -15.9%. In Q1, cash used in operations was $7.2M and capital expenditures were $11.6M, yielding a FCF margin of -29.8%. In Q2 2019, FCF was ($10.0M) for a FCF margin of -21.6% Cash used in operations were higher in Q2 due to an increase in accounts receivable, as some customers requests payment extensions. The majority of those extensions have already been paid in July.

• Q3 Revenue estimate of $73.5 – 75.5M, representing growth of $49.6% year/year. This compares to the consensus revenue estimate of $72.2M, which would have represented growth of 45.0%. Q2 to Q3 revenue growth is basically flat, unless we assume some outperformance. For comparison, Q2 to Q3 FY2019 revenue increased by about 7.8% and Q3 FY2019 revenue growth was 35.3% year/year.
• Q3 Non-GAAP EPS estimate of ($0.01) to $0.01, versus ($0.03) expected, representing a raise of $0.03 at the midpoint. Note that both Q2 and Q3 should generate positive Non-GAAP net income. Q2 FY2019 EPS was ($0.09).
• Q3 estimated Non-GAAP operating income of $1M to ($1M). Assuming the high end of guidance, operating margins would be positive once again. Q3 2019 operating loss was $9M.
• FY 2020 Revenue estimate of $290-300M, representing growth of 47.1% year/year at the midpoint. This compares to the consensus revenue estimate of $287.3M for annualized growth of 43.3%. In Q1, the company guided to $280-290M for the full year, so the company raised their outlook by $10M. In February, the company estimated a revenue range of $255-265M for the year with their Q4 earnings release. Revenue growth in 2019 was 38.7%.
• FY 2020 Non-GAAP EPS estimate of ($0.06) to ($0.01), versus analyst consensus estimate of ($0.13). The company had guided to ($0.15) to ($0.08) in Q1. FY 2019 Non-GAAP EPS was ($0.52).
• FY 2020 estimated Non-GAAP operating loss of ($12M) to ($2M). This is up from ($20M) to ($10M) as the previous estimates. FY 2019 Non-GAAP operating loss was ($34M).
• Ended the quarter with cash, cash equivalents and marketable securities of $454M. This includes the proceeds of $275M from Fastly’s follow-on offering in May 2020.

Other Performance Indicators

• Adjusted EBITDA of $7M in Q2, up from ($6M) in Q2 2019. This represented the first positive quarter for EBITDA and further underscores Fastly’s transition to profitability.
• Significant improvement in gross margins. Non-GAAP gross margin was 61.7% in Q2, up from 55.6% in Q2 FY2019. This represents over 600 basis points of improvement. Fastly leadership attributes this to improved efficiency in the operating model, as a consequence of their software defined network and ability to leverage headcount through automation. Q1 Non-GAAP gross margin was 57.6%, so the improvement is recent.

• Continued improvement in operating expenses as a percentage of revenue by category. These numbers are Non-GAAP. GAAP measures exhibit a similar trend in operational leverage improvement.

Expense Category	Q2 2020	Q1 2020	Q2 2019
R&D	17%	20%	24%
S&M	24%	28%	35%
G&A	19%	18%	18%

• Dollar-Based Net Expansion Rate (DBNER) was 137% in Q2, which is up from 133% in Q1 and 132% in Q2 2019. Net Retention Rate (NRR) was 138% in Q2, up from 130% in Q1. DBNER does not include the impact of customer churn. NRR includes customer churn, but also includes the impact on billing overage. In the earnings call commentary, management mentioned that customer churn is less than 0.5%. A DBNER over 130% puts Fastly in the top segment of rapidly growing software providers.
• Fastly ended the quarter with 630 employees. With 2020 revenue estimated at $295M at the midpoint, this translates to a run rate of $468k in revenue per employee. Fastly is able to generate significant leverage from each employee through thoughtful automation and a focus on premium customers.

Customer Activity

Fastly’s customer base represents a Who’s Who of the most innovative, internet-first companies. This is a deliberate strategy.

We focus on partnering with large companies that have the capability and available budget to increase engagement on our platform. We highly value these customers because they tend to have the quickest ramp in usage and the highest demand for ancillary products and features, as well as the ability to provide us with a certain degree of visibility and stability relative to other market segments in uncertain times.

Fastly Q2 2020 Shareholder letter

This approach to customer curation allows Fastly to achieve best-in-class DBNER and average spend per customer. It also manifests on the bottom-line in the form of operational leverage, as fewer employees are needed in sales support and customer service. It is also worth considering that Fastly isn’t able to disclose the identity of some customers. For example, we know that TikTok is a customer based on verbal commentary and have anecdotal evidence that Amazon.com and IMDB use Fastly as a CDN. Yet, neither is listed in the most recent customer highlights view.

• Total customer count increased to 1,951 in Q2, up from 1,837 in Q1 and 1,627 in Q2 2019. This represents 20% annualized growth and 6.2% quarter/quarter. This compares to sequential growth in Q1 of 5.4%. Management noted that Q2’s increase in total customer count of 114 was the largest in company history.
• Total enterprise customer count of 304, up from 297 in Q1 and 262 in Q2 2019. Enterprise customers are defined as customers with revenue in excess of $100,000 over the previous 12-month period.
• Average enterprise customer spend of approximately $716k in Q2, up from $642k in Q1 and $556k in Q2 2019. This represents annual growth of 29% and sequential growth of 11.5%.
• Enterprise customers generated 88% of Fastly’s trailing 12-month total revenue. This was consistent with Q1, and up 2% over Q2 2019.

The increase in enterprise customer count in Q2 represented only 7 customers moving into this category and was the smallest sequential increase in the past 12 months. However, management noted on the earnings call that some customers in impacted verticals, like travel and hospitality, actually dropped below the $100k spending threshold in the quarter. They didn’t provide detailed counts for customers moving up versus moving down. Fastly has many long-time customers in travel and hospitality, like Kayak, Airbnb, TripAdvisor, Ticketmaster and Hotel Tonight. These companies have reported significant reductions in usage year/year, over 80% in some cases. It is understandable that they would be impacted. Presumably, as macro conditions improve, these customers will return to normal spending. Airbnb, for example, recently announced that they resumed over 1M bookings per day on July 8th, after falling below that count since March 3rd. Also, on the recent Oppenheimer analyst call, Fastly’s CEO mentioned starting to see “green shoots” of new activity with travel customers in June/July.

Fastly management also pointed to the record new customer adds in Q2 as an optimistic signal for future enterprise customer growth on a recent analyst call with KeyBanc. Most new customers start with a smaller usage allocation, as they validate performance and make operational adjustments. Then, from year 1 to 2, they typically scale up spend by 3x and add 40% year/year after that.

Fastly Total Customer Counts

Q4 2018 = 1581
Q1 2019 = 1621  (+39)
Q2 2019 = 1627  (+6)   ** 1 Year prior
Q3 2019 = 1684  (+57)
Q4 2019 = 1743  (+59)
Q1 2020 = 1837  (+94)
Q2 2020 = 1951  (+114)

Looking at historical new customer adds for 2019, we see a noticeable slowdown in Q2 2019, and then acceleration for the next 12 months. This should translate into increasingly large enterprise customer counts going forward, as these past customer adds hit their one year anniversary.

In terms of the total market for these enterprise customers, on the Q2 earnings call and in a subsequent analyst conference, Fastly’s CEO made the point that incumbent CDN providers (referring to AKAM primarily) have thousands of enterprise customers (6-7k for Akamai). Due to Fastly’s better performance, modern design and developer-friendly approach, the CEO is confident that they will continue to pull these large spenders from the incumbent providers. These types of technology migrations are often led by the innovative companies and then trickle down to mainstream enterprises, as the engineering teams at more traditional enterprises read about what the innovators are doing. On the KeyBanc analyst call, Fastly’s CEO emphasized this motion. Traditional enterprises (like the global 2000) are realizing they need to update their online presence. They examine how innovative companies (Etsy, Shopify, Spotify, Pinterest) are approaching their technology choices and then mirror them. This is driving the large increase in new customer adds, which are coming from companies outside of the Internet-first sphere. The case for a better-designed, user-friendly product offering displacing legacy technology solutions is not new. Investors just need to look at Datadog’s or Zoom’s success for an example.

Fastly generates the majority of its revenue from enterprise customers. That percentage has increased from 82% in Q2 2018 to 86% in Q2 2019 to 88% in this past Q2. The growth of spend by these customers has been impressive, reaching $716k on average in Q2, up 29% annually and almost 12% sequentially. Fastly management attributes the spending growth to not just increased usage of the core content delivery platform, but adoption of adjacent services in delivery optimization and security, including Media Shield, Cloud Optimizer, WAF and TLS.

Fastly’s consistently high DBNER (137% in Q2, versus 132% a year ago) further underscores this expansion motion. New customers will “test” the Fastly platform with an initial traffic allocation for content delivery. Once they experience improved performance over incumbent solutions, they will increase their traffic allocation and consider adjacent product offerings. As an example, a major global payments company on-boarded at the beginning of 2020. In Q2, they further increased their utilization and added security and site acceleration services. Going forward, they plan to migrate application logic to run on the network edge in Fastly’s POPs.

Fastly’s focus on fewer large customers generates a number of efficiencies for their operating model:

• A need for less sales and customer service resources, resulting in lower spend on Sales & Marketing as a percentage of revenue.
• Higher utilization of POP hardware. Fewer customers implies less individual content files to store in each POP. This would result in high cache hit rates, as large numbers of requests for the same files across fewer customers allows a greater percent of total content to be kept in cache. This is opposed to incumbent providers with 1,000’s of POPs and many customers.
• For edge compute, the same efficiencies would be realized, as less code needs to be maintained in local storage at each POP.

It is worth noting that Fastly services small and medium sized businesses through their partnerships with large web site aggregators, referred to as platform partners. These include Shopify, Wix and Adobe Magento for web sites. The Fastly platform is also utilized by a number of software service providers, like Stripe, GitHub, LaunchDarkly and New Relic. By referring smaller companies to these aggregators, Fastly is able to increase their business with these large partners (who also consume Fastly services) and obviate the need to maintain a sales/service function to support smaller customers.

In Q2, Fastly landed new business across multiple verticals including education services, online retail, wellness and streaming. Some examples mentioned in the Shareholder Letter and earnings call were:

• A top U.S. university chose Fastly to help transition parts of their curriculum online. Management thinks more universities will follow this trend.
• A leading U.K. furniture retailer migrated to Fastly, implementing WAF and Image Optimization services. This was executed on a short timeline to accelerate their transition to online retail. They are also leveraging Fastly’s edge application logic to perform A/B testing.
• A major mobile app for wellness rapidly responded to unanticipated popularity by implementing Fastly’s offerings to ensure high-quality delivery to its customers.
• A leading European broadcast solutions provider selected Fastly’s Media Shield to power a new cloud service for global broadcasters seeking to leverage Apple’s Low Latency HTTP Live Streaming for OTT delivery of live TV.
• A developer of massive multiplayer online (MMO) games chose Fastly to optimize and accelerate in-game assets and interactivity to over 75M of their browser and mobile casual games customers.

TikTok Exposure

During this quarter’s earnings call, Fastly management disclosed that TikTok generated 12% of Fastly’s revenue for the first 6 months of the year. About half of this revenue is in the U.S. Assuming a $300M revenue run rate for 2020, this would represent about $18M in total annual revenue. Given that the U.S. government is considering an outright ban of TikTok in the U.S. after September 15, this revenue may be at risk.

Some additional perspective on this situation:

• Depending on how the ban manifests, it could reduce usage of the TikTok app in the U.S. That would directly impact Fastly’s usage revenue tied to the app. Management addressed this on the earnings call, highlighting the revenue risk. On the KeyBanc analyst call, they supplied more detail. For TikTok, Fastly provides both API and video delivery. Fastly deliberately limits the amount of video delivery that they take from any customer, in order to balance video usage with higher-margin content delivery types on their network. Management is confident they could backfill the video delivery portion of TikTok’s usage with some advance notice, as other customers would increase their video delivery. API usage, on the other hand, would be lost.
• If TikTok is sold to a U.S. company, it is likely the relationship with Fastly would be preserved. Microsoft is a leading contender at this point. While Microsoft has a CDN solution, they do allow operating groups within the company to retain their own preferred content delivery solution. An example of this is GitHub, which has been a long-time Fastly customer. Their usage hasn’t changed following the completion of the acquisition in October 2018.
• More importantly to me as an investor, though, is consideration for how TikTok became such a large customer and what that says about Fastly’s solution relative to other CDN providers. Granted, TikTok is likely using more than one CDN, but an annual spend of $36M total on a single CDN provider is significant.
• Management stated on the KeyBanc call that the concentration of revenue by TikTok represents an outlier and they only reached that level of concentration due to their rapid growth.

Analyst Reactions

After Fastly’s Q1 2020 results, six analysts posted an update. Five out of six had a buy rating and the 12-month price targets ranged from $31 – $36. Following the subsequent run-up in price to over $100 over the next couple of months, several analysts significantly revised their price targets upwards in July.

Following Fastly’s Q2 earnings, only five sell-side analysts provided updated coverage ratings. Price targets were significantly higher than those issued after Q1. Two analysts tripled their price targets, which I don’t think I have ever seen for a software name in such a short time. Out of the five, three issued a buy rating. The average price target for these updates is $97.75, representing a 9% increase from the closing price of $89.64 on August 6th and a 22% increase from the current share price around $80.

Date	Analyst	Rating	Price Target
8/6	Piper Sandler	Neutral	Lowered from $89 to $88
8/6	Baird	Outperform	Raised from $85 to $105
8/6	Stifel Nicolaus	Buy	Raised from $30 to $98
8/6	Credit Suisse	Outperform	Raised from $32 to $100
8/6	Oppenheimer	Market Perform

Following the earnings results, Baird set the highest price target, raising from $85 to $105. Analyst William Power provided the following commentary.

Baird analyst William Power raised the firm’s price target on Fastly to $105 from $85 and keeps an Outperform rating on the shares. The analyst noted its strong Q2 results despite high expectations. He believes the company should be well positioned to reach net income and free cash flow positive on a quicker timetable.

TheFly.com, August 6, 2020

Credit Suisse tripled their price target. Analyst Brad Zelnick similarly had positive feedback on the quarter’s results and the longer term opportunity.

Credit Suisse analyst Brad Zelnick raised the firm’s price target on Fastly to $100 from $32 and keeps an Outperform rating on the shares. The analyst notes that Fastly posted a “strong” Q2 as elevated traffic levels continued to benefit the top-line and drive significant leverage. Guidance also calls for a “solid” second half of the year, but does suggest Q3 growth below normal seasonality, which he views as conservative, Zelnick contends.

TheFly.com, August 6, 2020

Product Development Activity

The majority of Fastly’s product development is focused on their Compute@Edge product, which is in private beta with a set of existing customers. The shareholder letter and earnings call both referenced enthusiastic customer uptake and some interesting use cases.

During Q2, Fastly announced enhanced support for observability as part of the Compute@Edge solution. In a customer survey, Fastly revealed that having adequate observability was a major concern for edge compute solutions. This makes sense, as DevOps teams need to trace a customer request through all compute environments in order to troubleshoot issues. Fastly’s observability support extends to logging, tracing and granular metrics, which represent the bulk of what DevOps teams expect. These can be viewed within Fastly’s management console, or more usefully, shipped to popular third-party observability solutions, like Datadog, Splunk or New Relic.

In May, Fastly launched the Developer Hub. This is meant to serve as a one-stop shop for developers to build solutions on Fastly’s platform, including API and framework documentation, open source sample code, change logs, learning materials and a lightweight testing tool. It is worth reviewing all the sample use cases with starter code. This should help investors appreciate the breadth of usage expansion opportunities for Fastly customers, as more logic is moving to the edge. As Compute@Edge moves out of beta, Fastly plans to publish sample code for edge applications in the Developer Hub. We can expect more functionality added to existing use cases, due to the increased programmability available in full-featured languages like Rust and WebAssembly versus VCL (Varnish Configuration Language, Fastly’s current mechanism to support programmability).

While Compute@Edge is still in private beta, the Q2 shareholder letter states that the product offering is “on track and expect to expand the availability of Compute@Edge.” Interestingly, the shareholder letter goes on to mention the expansion of Compute@Edge from beta in November. This is a little earlier than previous targets of 2021. I don’t want to read into this too much, but am happy to see the opening up of the beta program a little sooner than 2021. This would presumably set up Compute@Edge in a meaningful way for revenue contribution in early 2021. On the KeyBanc analyst call, the CFO did confirm that they expect a minor revenue contribution from Compute@Edge starting in 2021. He also highlighted the higher gross margins for Compute@Edge usage, due to less network traffic. They are putting together the pricing model now and will likely draft off of the usage based structure established by AWS Lambda@Edge.

A lot of investors have speculated about use cases for Compute@Edge and some have even posited that they will be incrementally limited, beyond what is possible today. Since the program has been in beta, we don’t have a lot of insight into the full opportunity. To be realistic, as with any new technology, it will take some time to generate meaningful traction. However, I assume that the Fastly team has been thoughtful about the opportunity and isn’t building a next-gen compute platform in a vacuum.

The first data point regarding planned usage we have is from the results from a survey that Fastly conducted with beta customers for Compute@Edge published in May. What struck me is that 64% of the beta customers stated that they already had a project in mind for Compute@Edge versus just trying it out.

In a recent blog post, Fastly’s Chief Product Architect quipped that the list of compute workloads that shouldn’t be moved to the edge is easier to rationalize than creating an exhaustive list of those that should. With that said, there are some use cases that are emerging as good candidates for the first pass. Based on various blog posts, analyst calls, videos and the earnings call, here is a short list of use cases Fastly has highlighted so far:

• Authentication and Personalization. This is a common theme, and is done to some extent currently (like with the NYT). User authentication can be performed at the edge and the application can determine what content the user is authorized to view. This content could then be cached at the edge and rapidly returned to the user. In the current incarnation, this logic is handled by Fastly’s Varnish Configuration Language. However, Varnish is fairly limited as a first-class language, so I suspect enhanced personalization logic could be performed on Compute@Edge with full-featured languages.
• API response stitching. All modern mobile and web apps communicate with the application back-end through APIs. Responses of different API types, cacheable and non-cacheable, can be coalesced into a single response, lowering load on the back-end servers (and compute cost). While some API response caching is done now with VCL, API stitching allows for multiple asynchronous network calls to be made for dynamic content and then combining responses with static content to form the full payload to return to the mobile device. An example might be a user’s shopping cart, where product selections and details are cacheable at the edge, but user profile data is stored at the origin.
• Localized Microservices. Modern software architectures are increasingly breaking up single, monolithic applications into individual microservices. This provides segmentation of large development teams, better performance, easier maintenance and faster deployment iterations. As microservices are independent of each other and usually have their own data store, a subset of these would be candidates to distribute to the edge of the network and run closer to the end users. In the earnings call, Fastly referenced an example of “an online retailer that wants to migrate availability, stock and pricing microservices to the edge so they can deliver live updates and faster updates to end users.”
• Entry Management for High Demand User Events. On the earnings call, Fastly’s Chief Architect (Artur Bergman) described an online retailer that wanted to make their “waiting rooms smarter” by improving the responsiveness of token management for user purchases. This sounds like a use case for a high demand event ticket retailer, like Ticketmaster.
• Dynamic manifest manipulation for OTT video delivery. This enables the list of video files in a manifest to be changed on the fly for each customer request. As an example, it would allow the CDN to be selected for every user request, based on some algorithm, like speed reduction or cost optimization.
• Server-side ad insertion for OTT video and header bidding for web content. These are common AdTech use cases.
• Machine Learning Inferencing. While machine learning model creation is computationally intensive and wouldn’t be a candidate for edge compute, applying an existing model to provide rapid decisions for incoming data queries (inferencing) would be feasible at the edge.
• IoT Data Summarization. As more data is created by sensors and devices distributed across the globe, there will be an increasing need to collect that data, summarize it and ship the results back to a permanent data store. Processing the raw data at the edge and throwing out redundant values obviates the cost of shipping the full data set back to central storage. Similarly, for cases (like a temperature spike) where an action needs to be triggered, housing that logic closer to the network edge would result in a faster response.

Investors have also wondered about the impact of the roll-out of 5G on edge compute. I posit that 5G will make distributed edge compute even more critical. This is because 5G networks only increase the bandwidth available from the user’s device to the entry point of the cellular network. The network pipes from the edge of the network to central data centers will not increase by the same magnitude. Therefore, moving compute capabilities closer to the edge of the network will allow the applications that service the higher bandwidth devices to be appropriately responsive.

Most of the use cases mentioned above are related to intelligent content delivery or moving web application logic closer to the user. On the KeyBanc analyst call just this week, Fastly’s CEO was asked about use cases for Compute@Edge. Beyond the expected use cases around web application functionality, he mentioned that he had been surprised by some new use cases brought by customers, which are security related or encompass internal enterprise functions. He also mentioned that the security related use cases could be packaged into new product offerings in the future. This reminds me of Cloudflare’s commentary that they used their Workers infrastructure to build their Teams product. I’m not implying Fastly would launch a competitive offering to Teams, but rather that a programmable distributed compute solution opens up opportunities to address new security oriented workloads.

Outside of Compute@Edge, one notable launch was Fastly Flow Control, which limits request rates from automated sources like bots or scrapers. This is important, as aggressive bots can generate substantial load on a site when scraping content, even if for benign purposes, like a search engine. The Flow Control feature can throttle these requests to a level in which they don’t impact site availability. This capability will become the foundation for formal bot mitigation and anti-automation products. Currently, Fastly partners with third party providers, like PerimeterX, for these capabilities. Having these capabilities in-house would generate a new revenue stream.

In Q2, Fastly also announced the expansion of their network capacity to 100Tbps. This represents 35% growth since the beginning of 2020 and almost 100% growth since Q2 2019. This rapid scaling of network capacity is in response to customer demand. On average, network utilization is under 20Tbps by a rough sampling of Fastly’s real-time network map, so there is plenty of headroom for growth. Fastly’s network has 55 POPs in 26 countries. As I have discussed previously, Fastly’s approach to POP distribution hinges on the notion that fewer, larger POPs can actually deliver content to the end user faster by maximizing the cache hit rate. The cache hit ratio for Fastly POPs is generally around 90%, versus 80% or lower for competitive solutions. This results in better overall performance, as a cache miss forces the end user’s request to be directed back to origin, which can increase the response time by 10x.

However, while fewer, larger POPs makes sense from a performance perspective, coverage by country might be a necessary consideration in the future. Cloudflare recently highlighted that their POP distribution passed 100 countries, which could provide advantage if data localization regulations expand beyond the EU and US. Currently, Fastly can address data localization with coverage in major countries, but if data requirements become more granular, this could become a constraint that Fastly will need to address in the future.

In June, Fastly announced the acquisition of Tesuto. Tesuto is a virtual network emulation platform, which allows users to clone any network and test planned configuration changes in a controlled sandbox. This will be applied to Fastly’s platform to facilitate automated testing and deployment of network configuration changes, which enables users to avoid unnecessary downtime. This adds another capability to the Fastly platform to facilitate fast and secure distributed compute, by providing more support for programmability and testing of the network configuration. Another aspect of the Tesuto acquisition that is noteworthy is the addition of network engineering talent to the Fastly team. Since their founding, Fastly has been attracting thought leaders in language design, networking and compute infrastructure, who often sit on the standards boards that are mapping out future protocols for internet technologies. With the Tesuto acquisition, Fastly picks up the three co-founders, with accomplishments between them of founding EdgeCast Networks and working on Google’s data center fabric and SD-WAN.

One distinction investors should consider as they think about the security opportunity is how Fastly divides the security market into two segments. This was recently reiterated by the CEO at the Oppenheimer Technology Conference. First, there is enterprise security which focuses on protecting the company’s assets, like employees, data centers, corporate network, etc. Examples of products in this segment are VPN, firewalls, endpoint protection (EPP) and identity management, with next gen providers like ZScaler, Crowdstrike and Okta. Fastly does not offer products in this space, nor plans to do so in the near term. This segment’s target customer audience is enterprise IT managers, like the CIO or CISO. The second segment of security is focused on application security. This has to do with protecting a company’s software-delivered product experiences, content, APIs and overall availability. Existing product offerings for this segment are services like DDOS mitigation, Web Application Firewall (WAF), bot management, content access, etc. This segment’s target customer is the CTO or VP of Software Engineering, as the leader of software development. This is the segment that Fastly is targeting, with both a current set of products and plans for future ones.

As it relates to the market opportunity, the CEO made an interesting comment on the Oppenheimer analyst call. He said that he had a discussion recently with a large customer’s CTO. That individual stated that for every $100 that they planned to spend on central cloud infrastructure, they would allocate $5-10 on application security and $5-10 to edge computing. This represents a powerful statement about the future TAM for Fastly products. While the application security market is fairly well understood, investors are still struggling to size the market for edge compute. According to Gartner, the combination of spend for IaaS, PaaS and Cloud Management services is projected to exceed $170B by 2022. This implies the market just for edge computing could range from $8-17B, and represent a similar opportunity for application security.

Competitive Activity

In terms of the competitive landscape, I think it is worth examining independent service providers and the cloud vendors. Of the independents, Cloudflare (NET) and Akamai (AKAM) are notable. I acknowledge there are other independent providers (like Limelight), but I don’t think they are candidates to dominate the category going forward. Of these three, Akamai is the incumbent and Fastly and Cloudflare are disrupters. For more detail on the product offerings from Akamai and Cloudflare, readers can refer to my original deep dive on Fastly and subsequent summaries of quarterly earnings, where these names were surfaced in the competitive sections. For a general background on Cloudflare, you can review a Hhhypergrowth article from February that provides an informative explanation of all of Cloudflare’s offerings. The Hhhypergrowth blog is written by a peer analyst with a strong technical background.

To kick off, let’s compare the three names on top-level performance metrics coming out of Q2 earnings. All earnings metrics are Non-GAAP.

Metric	FSLY	NET	AKAM
Q2 Rev Growth	62%	48%	13%
Q3 Rev Growth (Est)	50%	39%	8%
2020 Rev Growth (Est)	47%	41%	9%
Q2 Gross Margin	62%	77%	77%
Q2 Operating Margin	2%	-9%	32%
Q2 FCF Margin	-16%	-20%	22%
Q2 DBNER	137%	115%	N/A
2020 Total Rev (Est)	$295M	$406M	$3.150B
Employees	630	1,535	7,951
2020 Rev / Employee	$468k	$264k	$396k
2020 Gross Profit / Emp	$290k	$203k	$305k
Current P/S Ratio	31.2	28.2	5.9

Cloudflare

Cloudflare has been growing at a consistently rapid pace. For full year 2019, they increased revenue by 49%, which surpassed Fastly’s growth rate of 39%. As mentioned above, for 2020 they are currently projecting 41% annual growth, which we can expect will get raised into the high 40% range. It is encouraging to see a high sustained revenue growth rate year/year, as many software companies experience decelerating revenue growth as they scale.

In Q2, Cloudflare delivered particularly impressive metrics for customer growth. They increased their large customer count (spending more than $100k annually) by 65% year/year. Total paying customers grew by 18k over the prior 12 months and almost 7k sequentially. Their annualized total customer growth rate was 24%. They claim 16% of the Fortune 1000 as paying customers. For comparison, Fastly grew total customers by 20% year/year and enterprise customers by 16%. Fastly’s average enterprise customer spend increased by 29% year/year to $716k total.

Cloudflare also highlighted some impressive customer wins on the earnings call, several of which represented total contract value greater than $1M.

• One of the largest European financial services firms signed a three-year deal worth $450k per year for Cloudflare for Teams, their Zero Trust cloud-based replacement for legacy VPNs and firewalls. This was a competitive win and has further expansion potential into other product offerings.
• A large U.S. based industrial manufacturer signed a deal worth $350k per year to adopt Cloudflare’s Web Application Firewall. The company was moving to the cloud as part of their digital transformation, replacing a legacy on-premise solution. For this quarter, they are considering Cloudflare for Teams for their 50,000 employees.
• A leading identity and access management provider signed a two-year deal worth $500k per year. They moved workloads away from Amazon Web Services due to better speed and flexibility of Cloudflare’s platform.
• A cloud-first gaming platform signed an expansion of their existing Cloudflare contract for an additional $1.8 million annually. Part of their deployment makes use of Cloudflare Workers.

In terms of product development delivery, Cloudflare has been executing at a rapid pace. They announced the following product enhancements or new products over the course of Q2:

• Workers Unbound. Announced in late July as part of Serverless Week, this represented a major set of enhancements to the Cloudflare Workers product line. The improvements included removal of runtime constraints, lower costs, broader language support and better developer tooling. I provided full coverage of these releases in a separate blog post. I thought this was a very positive move for Cloudflare and re-sparked my interest in the name. Coming out of Q1, it seemed that Cloudflare was pivoting away from providing developer solutions towards an enterprise security focus with the release of Teams. As I mentioned in my coverage, I think Cloudflare’s serverless product is targeted at a different segment of the market than Fastly’s Compute@Edge. It is supporting longer-running workloads with a broader set of languages. This is enabled by their use of the V8 Engine, but comes at the expense of performance and resource utilization. Fastly’s decision to build their own runtime with Lucet has resulted in faster start-up times and a smaller server footprint, but with more limited language support. I think both approaches will find ample market opportunity for serverless distributed compute, as they will cater to different types of workloads and development organizations.
• Cloudflare Network Interconnect. Allows companies to connect their internal networks directly to the Cloudflare network. This provides the benefits of security, reliability, and performance versus using the public Internet to route traffic to Cloudflare’s network.
• Bring Your Own IPs. This capability allows Cloudflare to announce a customer’s IP space from their network edge and tie that IP prefix to the services being used. The customer’s IP space will be protected and accelerated as if they were Cloudflare’s own IPs.
• Regional Services. For customers that want to apply data locality rules, they now have control over which Cloudflare POP will process their traffic. This is useful on the border of two countries, for example, where the POP closest to a particular user may be located in an adjacent country. Regional Services allows the customer to tie geographic areas to certain POPs. Once in place, incoming user traffic is routed to the specified POP before it is processed.

There were other minor product announcements through the quarter as well. What struck me about Cloudflare’s product cadence is the pace and breadth of product releases. They are executing rapidly across a broad set of product categories. Additionally, I think the way they lead their product adoption motion with networking and security is interesting. With some of their newer product releases, Cloudflare is tying their network more directly into the network of their customers. This “trojan horse” approach then allows them to sell other application support product offerings, like Workers and CDN, into the organization as the relationship grows. The CEO highlighted this strategy in his opening remarks on the Q2 earnings call. This would appeal to more traditional enterprises where the development function falls under a CIO, who might be looking for a full suite of services from one vendor, starting with networking and enterprise security.

Akamai

On the Oppenheimer analyst call, Fastly’s CEO talked about the market for content delivery. He acknowledged that video delivery has become commoditized, where Fastly sees competition from Limelight, Level3, Akamai and others. For every other content delivery use case, though, he said that Fastly generally competes with Akamai for enterprise business. From the CEO’s perspective, Fastly’s advantages over comparable Akamai solutions revolve around programmability, ease-of-use, developer tooling/controls and documentation. He also talked about how Fastly’s implementation for a new customer is usually very fast, requiring little outside help. This is opposed to Akamai, where the sales cycle is prolonged and involves engaging a professional services organization that is largely offshore. Based on my past experience using Akamai’s services, I can appreciate this perspective.

In May, Akamai announced a new product offering called Page Integrity Manager. This represents an in-browser monitoring capability that identifies malicious third-party Javascript code that might be used to skim personal user information or disrupt the user experience. The monitor examines user sessions in real-time for script behavior, including source, data accessed and any export across the network. Based on a scoring algorithm, it rates the threat level and surfaces alerts for security personnel to review. This is offered as part of Akamai’s Web Application security products.

As it relates to edge computing, Akamai published a blog post in early August detailing their history of providing edge compute solutions. While they highlight a number of their product features that do indeed perform computations on the edge of their network (like DDOS or WAF), I am most interested in a programmable serverless compute environment with complete developer tooling that can be synchronized across all POPs. For this, Akamai launched their EdgeWorkers product in October 2019. This solution is based on the V8 Engine and currently only supports JavaScript for development. The product is in private beta for existing customers, requiring a sign-up form and review to utilize.

I have detailed the limitations of using the V8 Engine in a prior blog post. Additionally, while Cloudflare utilizes the V8 Engine, they have substantially enhanced the capabilities wrapping it as part of their Workers product. As mentioned above, these include broader language support, developer tooling, longer runtimes, competitive pricing and substantial security monitoring. Also, Cloudflare’s product is live, not in beta.

Cloud Vendors

From the cloud vendors (AWS, Azure, GCP), there were no major developments during the quarter in CDN or serverless offerings. I previously wrote about how independent software providers would compete favorably with look-alike offerings from the cloud vendors. I still maintain that in certain software categories, independent providers have built superior products and will continue to extend their lead. I believe that CDN, distributed serverless compute and security are three areas where independents can effectively compete with the cloud vendors. The primary tailwind generating leverage for the independents is the multi-cloud trend. As enterprises evolve their cloud footprint, they are increasingly considering a multi-cloud strategy. This allows them to reduce lock-in, improve their negotiation posture and manage risk. Multi-cloud deployments bias towards neutral, independent software providers, as they provide a single API interface for service access and benefit from the focus of their technology development on a subset of use cases. I think these factors will favor leading independent providers like Fastly and Cloudflare.

Fastly Take-aways

Putting aside elevated investor expectations for Q2, I think Fastly had an exceptional quarter. Revenue growth accelerated by 24% year/year. Gross margins improved by 600 basis points. They delivered positive EBITDA for the first time and had record new customer adds. The Compute@Edge product beta appears to be progressing nicely with real customer use cases and the potential to generate revenue contribution in 2021. Fastly leadership has also referenced new security offerings that could further pile on revenue streams. Revenue growth is approaching 50% for the full year of 2020. Looking out to 2021, if DBNER stays above 130%, customer adds continue to accelerate and new product offerings begin to contribute revenue, this high revenue growth rate could be sustained.

Transitioning to profitability with this sustained high revenue growth is a significant milestone as well. Investors are used to high growth software stack companies postponing profitability for several years after IPO, while they focus on revenue growth. The fact that FSLY ended FY2019 with a -17% Non-GAAP operating margin and is now delivering positive operating margin and EBITDA is a strong signal about leverage in their operating model.

These profitability improvements were further underscored by the increase in gross margin by over 600 basis points. Management attributed this to “increased leverage” of their software defined network. As I mentioned previously, Fastly invested significant engineering effort in the design of their POPs infrastructure and the control of the network routing layer between POPs. By moving internet routing logic to individual POP servers (rather than a one-size-fits-all approach at the border router), Fastly is able to dynamically optimize traffic routing for every user request. Beyond intelligently routing traffic to maximize response time, they could also include cost as a variable to optimize. This would allow them to reduce network traffic costs over traditional CDN architecture. Additionally, Fastly’s continuous performance optimization of their software-driven solutions boosts POP capacity and ultimately “increases the amount of revenue per server.” As an example looking forward, Compute@Edge cold start times of 35 microseconds with a runtime footprint of several kilobytes of memory represents the smallest resource utilization of any comparable serverless solution on the market. This will drive lower CapEx costs for POP expansion and less operational overhead.

Fastly is demonstrating strong leverage in their go-to-market model, by focusing on enterprise customers. They have a high revenue / employee ratio, resulting in a low Non-GAAP S&M spend as a percentage of revenue of 24%. This puts Fastly on par with other low sales cost companies, like TWLO, which also spends 24% of revenue on S&M. For comparison, NET spent 48% on S&M in Q2 and DDOG allocated 33% of revenue (granted DDOG has gross margins of 80%).

Q3 revenue growth is estimated at $74.5M or 49.6% annualized growth at the midpoint. If Fastly delivers a similar sized beat of about 7%, we could expect actual revenue growth in the mid 50% range. If Fastly delivered at the estimate for Q3, total quarterly revenue would essentially be flat from Q2 to Q3 this year. In 2019, revenue grew about 7.8% sequentially from Q2 to Q3. I think we can expect a similar level of outperformance this quarter.

Additionally, we can get a sense for Q4 estimates at this point. For the first two quarters of 2020, Fastly delivered $137.6M of revenue. If we add in the Q3 estimate of $74.5M, we get to $212.1M. At the high end of estimated 2020 revenue, Fastly is targeting $300M. This means that Fastly management is currently expecting Q4 revenue of $87.9M. This would imply Q4 revenue growth of about 50%. Q4 is traditionally Fastly’s strongest quarter. This year, e-commerce spending will likely be higher than in the past, which might provide an additional tailwind. As an example, many iconic bricks-and-mortar companies, like Dressbarn and Pier 1 Imports, are being acquired out of bankruptcy and re-launched as pure e-commerce plays. For Dressbarn at least, online operations were migrated to Shopify Plus, which is a Fastly customer. On the Oppenheimer analyst call, Fastly management also expressed optimism around this year’s Q4 delivering record e-commerce activity.

Fastly does still pose execution risks that investors will need to monitor. A cloud vendor could double-down on serverless compute and launch an innovative new offering to compete with Compute@Edge. The adoption of new edge compute use cases may take longer than expected, similar to the initial excitement and then more gradual ramp investors are witnessing with Twilio’s Flex product. The anticipated acceleration of digital transformation might slow once the COVID-19 situation clears and enterprises feel they have more time to execute. However, for investors with a long-term perspective, I think that Fastly is well-positioned to capitalize on several evolving software infrastructure trends with large growth potential.

Investment Plan

In light of Fastly’s strong Q2 performance and momentum going into 2021, I continue to be bullish on the name. I will maintain an outsized allocation to FSLY in my personal portfolio. Additionally, I had established a 5 year price target of $155 in May, when FSLY was trading at $38.50. I think that target is easily achievable before 2024 and will revisit the number at the end of 2020.

In parallel, I have initiated a position in Cloudflare (NET), with proceeds from other sales. After my deep dive on their enhanced serverless product offering, I was sufficiently convinced that Cloudflare should also gain a significant share of future edge compute, advanced networking and security spend. I would like exposure to both names. Due to Fastly’s current momentum, I am maintaining a higher ratio of allocation between FSLY and NET. I will continue to track progress against the investment thesis for both names going forward and will make adjustments to my allocations as the companies evolve.