Fastly: Help Me Edge Compute, You're My Only Hope
Summary
- Fastly’s financial performance continues to disappoint investors who believed the Q2 2020 jump in revenue growth was sustainable.
- Fastly’s performance remains reasonable, although should be viewed as disappointing in light of the favorable macro environment and the stock’s recent high valuation.
- Management is guiding for stronger growth later in the year, and the stock is now more reasonably priced.
- For those that invested in Fastly for the company's technology, the general availability of Compute@Edge in 2021 and how customers respond is critical.
Fastly’s (NYSE:FSLY) stock has struggled in recent months as investors have grappled with the high valuation, customer concentration and modest growth. The acquisition of Signal Sciences gives Fastly a more well-rounded product portfolio and the general availability of Compute@Edge could allow the company to begin fulfilling its potential. Management is guiding for relatively weak organic growth though, and it is not clear whether this is due to caution, tough comparable periods from 2020 or more fundamental issues. Some investors have pointed to the loss of TikTok as a customer as the driver of Fastly’s recent struggles, but it is not clear that this is the case. TikTok are still a customer, with Fastly referring to them as an important customer.
Fastly continues to achieve relatively rapid usage expansion within existing customers and new business wins across multiple verticals including gaming, education, financial services, e-commerce and media. They are promoting cross-sell and upsell opportunities from the Signal Sciences acquisition and have indicated these opportunities are larger than expected. Fastly have also indicated there is momentum from the Google Marketplace Partnership and strong developer interest in their edge compute service. None of this has shown up in their financial performance or guidance so far though.
Expectations for Fastly’s edge compute service are high, but I would not expect it to be a meaningful contributor to growth in the short-term and this is what Fastly’s management are guiding. Cloudflare (NET), Amazon (AMZN), Akamai (AKAM) and Limelight (LLNW) have all launched edge compute solutions with no obvious financial impact. Unless the speed and cost of Fastly’s solution is so superior to competitors that it opens up a range of new use cases, investors may be disappointed.
Figure 1: Fastly Revenue Growth
(source: Created by author using data from Fastly)
Fastly’s net retention rate has declined significantly over the past 2 quarters and was only 107% in Q1, compared to Cloudflare’s 123%. Given Fastly’s focus on large customers and usage-based business model this number should be significantly higher. The fact that it is not may be the result of a lack of product innovation (unable to onboard customers to multiple products) or internet traffic normalizing after COVID. Fastly’s calculation method makes use of a year-over-year comparison of the final month in a quarter, which also introduces volatility. Fastly’s management believe that viewed over the trailing 12 months this metric remains relatively robust.
Figure 2: Fastly Dollar-Based Retention and Net Expansion Rates
(source: Created by author using data from Fastly)
Fastly’s gross margins have edged higher in recent periods with the acquisition of Signal Sciences, although were lower in the most recent quarter. Management are guiding that this is a temporary result of an expansion of network capacity and that margins will improve later in the year as growth accelerates. With the launch of the higher value add Compute@Edge service it would be reasonable to expect margins to continue improving. Fastly has stated that they are targeting non-GAAP gross margins of up to 70% in the long-run, which is essential if Fastly is to justify its stock price. In the short-term, operating margins are likely to remain depressed until Signal Sciences has been fully integrated and the business has further scaled.
Figure 3: Fastly Profit Margins
(source: Created by author using data from Fastly)
Investors have latched onto the idea of edge compute companies as complementary / competitors to the hyperscaler cloud companies. As a result, the expectation is that the market leaders could eventually be worth hundreds of billions of dollars. This is a possible, although optimistic scenario, and as such edge compute companies should be investing extremely aggressively to build out a portfolio of solutions which establish a competitive advantage. Fastly is failing to capitalize on the market opportunity being presented though, as evidenced by a number of indicators. Cloudflare has been investing more heavily in capital, sales and marketing and R&D in recent years and the impact of this is becoming increasingly evident in the performance of the two companies.
Figure 4: Fastly’s Growth Investments
(source: Created by author using data from company reports)
Much of Fastly’s capital expenditure is going towards increasing the number of Points of Presence (POPs) and expanding their networks capacity. Fastly increased the capacity of their network by 48% over the past 12 months which indicates confidence that they will be handling larger volumes of traffic in the future.
Figure 5: Fastly Network Capacity
(source: Fastly)
As a relatively small business that is still in the process of developing its core technology, Fastly should be aggressively investing in R&D and sales and marketing. Technology markets often exhibit increasing returns to scale, leading to winner take most markets and a sustainable competitive advantage for the market leader. Fastly has an opportunity to establish itself as the market leader, or one of the market leaders, but it must be willing to take risks to do so. Cloudflare is hiring more rapidly, resulting in a much larger headcount than Fastly. This may be contributing to Cloudflare’s ability to rapidly innovate and expand their customer base.
Figure 6: Fastly Headcount
(source: Created by author using data from company reports)
Fastly’s lack of investment and hiring could be interpreted as meaning that Fastly is more efficient, or that it is trying to grow in a more prudent manner. This does not appear to be the case though when Fastly’s growth is compared to Cloudflare’s. Cloudflare are rapidly innovating, causing their addressable market to expand. They are also beginning to move upmarket, leading to rapid growth in enterprise customers. Fastly’s addressable market has been stagnant since listing and they are struggling to grow their enterprise customer base.
Table 1: Fastly Total Addressable Market Estimate
(source: Created by author using data from company reports)
It is worthwhile noting that at least some of Fastly’s slow customer growth can be attributed to a conscious choice by the company to avoid commoditized parts of the delivery business. Fastly is focused on customers who value performance, security and reliability and are willing to pay a premium for it. As a result, Fastly counts technology leaders like Microsoft, Google, Amazon, Shopify, Stripe, Spotify, Slack, Twitter and Pinterest as customers. Fastly’s strategy is to target these types of large and rapidly growing customers and then grow with them, as a result of their usage-based business model. I believe this is a questionable approach though, as it creates an over reliance on large customers like TikTok and I believe Fastly will find it difficult to move down market once they have saturated their niche.
Fastly currently has 2,207 customers, of which 336 are enterprise customers. In Q4 2020 Signal Sciences had 280 customers with 78 enterprise customers (25% overlap with existing Fastly customers). Fastly’s enterprise customer growth rate had fallen alarmingly over the past few years but this trend appears to have reversed. Approximately 90% of Fastly’s revenue comes from enterprise customers and most of this likely comes from a handful of customers, indicating the company’s current customer concentration. This is a situation that other software companies, like Twilio (TWLO), have found themselves in the past. While Fastly’s strategy may be to target a relatively small number of customers who are performance focused, they must also broaden their customer base if they are to create sustainable long-term growth.
Figure 7: Fastly Enterprise Customers
(source: Created by author using data from company reports)
Figure 8: Fastly Enterprise Customer Growth
(source: Created by author using data from Fastly)
While there are legitimate concerns about Fastly’s business, investors should keep in mind that it is the company’s differentiated technology that generated initial excitement. Legacy vendors are at a disadvantage as they were faced with higher latency when they built out their networks, and as a result, they utilize a larger number of smaller POPs. Since that time telecommunications infrastructure has improved significantly, which has allowed Fastly to build out a network with a smaller number of larger POPs. This increases latency somewhat, but as a result Fastly achieves much higher cache hit rates, improving the performance of their network. Fastly’s servers generally have 384 GB of RAM, 6 TB of SSD storage and each CPU has 25 MB of L3 cache. A typical POP has 32 machines with these specifications. SSDs have no moving parts, creating constant lookup times for content no matter how big the drives are. This setup is more capital intensive than typical edge servers but supports Fastly’s performance focus.
Fastly custom built their own network architecture with a focus on software rather than hardware. They have customized almost all aspects of their content delivery solution to be programmable, performant and cost efficient. They use customized switches which function similarly to more expensive routers and push external routing logic down to the host level, reducing load on the switches and providing more fine-grained control over traffic routing.
Fastly also took the time to build their edge compute product from the ground up, giving them advantages that will be difficult to replicate. Fastly is utilizing WebAssembly (WASM) for their edge compute service, similar to competitors like Cloudflare. WASM translates several popular languages into an assembly language that runs quickly on the target computer architecture. Unlike competitors who rely on the Chromium V8 engine for compiling and running WASM, Fastly built their own Lucet compiler and runtime, which was optimized for performance, security and compactness. This has allowed Fastly to achieve cold start times of approximately 35 microseconds in comparison to 3-5 milliseconds for Chromium V8.
Table 2: Approximate Cold Start Times
(source: Created by author using data from Fastly)
In addition to being faster, the Lucet runtime also occupies only a few kilobytes of memory, versus at least 3MB for the V8 Engine. This may be the most important feature of Fastly’s edge compute solution. It requires significantly less overhead, making it more scalable with lower capital expenditure. This could eventually translate to lower costs for customers or higher margins for Fastly.
While Cloudflare’s edge compute solution may be technically inferior in terms of speed and overhead, it has been on the market for a number of years and Cloudflare have already leveraged it to build innovative solutions, and are adding features (higher compute and memory limits, durable objects) to enhance functionality. In comparison, Fastly’s edge compute solution has only just hit the market and they are still working on features like state at the edge. Fastly’s state at the edge solution allows reads and writes that are local to a given POP in an eventually consistent system. To do this Fastly utilizes a state primitive called the Conflict-Free Replicated Data Type (CRDT) structure. Operations on this type of data structure are always associative, commutative, and idempotent, which may limit use cases.
Fastly is unlikely to dominate the market based just on superior technology. They need to shift focus from engineering to strategy and sales if they are to be successful. This is a process they have started in the past 12 months with founder Artur Bergman relinquishing his role as CEO so he could focus on product development. Fastly have also hired an experienced chief revenue officer to develop and drive the company’s global sales and client services strategy.
Fastly’s Vision
It is not particularly clear from Fastly’s public material what they are trying to achieve. They emphasize a focus on providing developers with a secure and programmable edge cloud platform and promote the virtues of edge computing, yet remain overly focused on content delivery for performance sensitive applications like e-commerce, media and gaming.
Figure 9: Fastly's Portfolio
(source: Fastly)
Figure 10: Edge Cloud Architecture
(source: Fastly)
Fastly’s Secure@Edge service is trying to create a comprehensive, unified solution for web and API protection. To aid this Fastly acquired Signal Sciences to broaden their security portfolio. The deal was for 575 million USD in Fastly stock and 200 million USD in cash. If nothing else, it appears to have been an attractively priced acquisition given Signal Sciences had an ARR of 28 million USD at the time and Fastly’s stock was significantly overvalued. Fastly believed that Signal Sciences was a good fit for their developer focus and that the acquisition had the potential to help accelerate adoption of Compute@Edge. Fastly plans to create an authentication solution at the edge, which presumably could be used for both public and private applications, increasing their addressable market and the attractiveness of their platform.
Table 3: Secure@Edge Solutions
(source: Created by author using data from Fastly)
Fastly continues to add capabilities and use cases to Compute@Edge which will help to leverage their high performance serverless offering.
Figure 11: Fastly Compute@Edge Roadmap
(source: Fastly)
Fastly recently announced full support for granular metrics, logging and tracing of their serverless compute runtimes. Logs can be streamed to a number of log analysis solutions, like Datadog (DDOG), Splunk (SPLK) and Elasticsearch (ESTC). The ability of edge cloud providers to expand into adjacent verticals like observability and access management is part of the reason I believe it is an attractive market for investors.
Figure 12: Fastly Observe@Edge Roadmap
(source: Fastly)
Competition
Fastly generally draws comparisons to Cloudflare, but I do not believe the two are close competitors as Fastly is more focused on high performance content delivery, whereas Cloudflare is focused on enterprise networking. Fastly’s investor material dismisses Cloudflare as a small business focused CDN, which should concern investors as it indicates they fundamentally do not understand the strongest company in the edge compute market. Cloudflare built a leading network by acting as a CDN for small businesses, but they are now leveraging this to become the future of the enterprise network. If Fastly are going to compete with Cloudflare, they need to develop a portfolio of complementary services that appeal to the broader market, similar to what Cloudflare is doing. Fastly’s only current advantage is the performance of their network and if Cloudflare believe Lucet is a competitive advantage, they are likely already developing a comparable solution.
Figure 13: Fastly's Competitive Advantages
(source: Fastly)
In the commoditized bulk video delivery market Fastly faces competition from Limelight, Level3 and Akamai amongst others. For other content delivery use cases, Fastly generally competes with Akamai for enterprise business. I believe Fastly is well positioned to take market share from Akamai due to the inherent advantages of their network.
Most CDNs now have an edge compute solution on the market with varying degrees of sophistication and maturity. 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. AWS offers their Lambda@Edge product for edge computing. This is an extension of Amazon Cloudfront, which is their CDN service, and AWS Lambda, which provides Amazon’s serverless solution. In a similar approach to Amazon, Microsoft Azure offers edge computing through the use of Azure Functions, which can be deployed on Azure, Azure Stack or Azure IoT Edge. Google supports serverless runtimes through their Cloud Functions.
Valuation
Fastly is guiding for weak organic revenue growth in 2021 and it is not clear whether they are being conservative or if there are underlying problems with the business. With the Signal Science acquisition (cross-sell and upsell opportunities) and the launch of Compute@Edge, growth should be accelerating in spite of tough comparable periods from 2020.
Even after the recent pullback, Fastly’s valuation implies a relatively high level of growth, but it is not clear if this is warranted. Fastly have been slow bringing their edge compute service to market and instead of developing their own security solutions, acquired Signal Sciences. This raises the prospect of execution risk, in addition to a lack of clarity in the company’s strategy and a failure to capitalize on the large market opportunity being presented to them. Based on a discounted cash flow analysis I estimate Fastly’s stock is worth approximately 50 USD per share. If investors continue to buy into the narrative of Fastly’s technology advantage, downside risk may be relatively low, although evidence of improving customer acquisition and revenue growth is likely needed for the stock to move higher.
Figure 14: Fastly Relative Valuation
(source: Created by author using data from Yahoo Finance)
This article was written by
Disclosure: I am/we are long NET. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.