[Note: this was published originally on December 27, 2018 in one of our letters to investors.]
The API Economy
We made a new investment during the quarter in a company called Twilio. This company fits squarely within our framework: it is led by a mission-driven founder/CEO, operates a business with a large addressable market, has strong competitive advantages, and is riding several technological adoption curves (S-curves). I believe Twilio has the potential to be worth multiples of our cost over the next several years.
You’ve probably used Twilio many times without ever knowing it. When a website sends you a text message with a code for two-factor authentication, that’s functionality powered by Twilio. Phone calls with an Uber driver have the number masked, for privacy; that’s also Twilio. And when you unlock a Lime Bike with your smartphone, that’s done through Twilio’s programmable wireless SIM cards.
Like Amazon Web Services (AWS), Twilio makes building blocks for the compute economy; but while AWS is focused on the underlying infrastructure layer (storage, compute, databases, machine learning), Twilio is building a communications network on top of AWS.
It’s useful to think of APIs—application programming interfaces—as standard interfaces through which information or resources can flow. The electrical socket on your wall can be thought of as a standard interface. Every electric appliance maker—from vacuum cleaners, to microwave ovens, TVs and iPhones—can simply assume there will be a standard socket on the wall to plug into and get electricity. That socket is an API for electricity.
Once electricity is assumed to exist for everyone, society can move one layer up and begin making other assumptions. Parking a car in American cities today requires a smartphone, a credit card and a parking app. City planners have begun with the assumption that of course everyone has a credit card and a smartphone. The smartphone and the app can be thought of as APIs through which you tell the city where your car is parked and how you’d like to pay. To recharge your smartphone, you rely on the electricity API in the previous layer.
Twilio benefits from a trend we wrote about last quarter: “The big picture is that software is eating the world—that is, many of the products and services developed over the past 150 years are transforming into, or being disrupted by software (full credit to venture capitalist Marc Andreessen for articulating this idea years ago).”
In Twilio’s case, it’s software eating communications: Twilio is unbundling the century-old copper-wire telecommunications infrastructure.
The notion of unbundling came from Jim Barksdale, a co-founder of Netscape alongside Marc Andreessen. One day, at an investor meeting, Jim came up with the idea:
Barksdale: It actually was a quote at the end of a very long road show when we were taking Netscape public, a trip through Europe that ended in London. It was our last show and we were in a big hurry to get the plane. We were at the Savoy Hotel, and the ballroom was full of all these British investment bankers, and we give our normal pitch. Peter Currie, our CFO, and I were doing it, and I said, “All right, one last question.” And this fellow, he says something about, “How do you know that Microsoft isn’t just going to bundle a browser into their product?”
I said, really just to end the conversation, “Gentlemen, there’s only two ways I know of to make money: bundling and unbundling.” And said, “We’ve got an airplane to catch.” And we left, and Peter Currie, walking out the door, said, “Those people are looking at you, Barksdale, like you’re crazy. What did you just say?” And I said, “Well, best I can tell, most people spend half their time adding and other people spend half their time subtracting, so that’s what works out.”
Andreessen: It also demonstrates a huge advantage to having to leave for a plane.
Barksdale: That’s right. It had the added advantage of being true, as a friend of mine once said.
Once you’re aware of this idea, you see it everywhere. The newspaper was a bundle of news and advertising; the internet unbundled it, and a new bundle was formed (Facebook). Music albums were bundles of songs; the internet unbundled the album, first through MP3 downloads, and later through streaming services like Spotify. In the world of computers, the first bundles were mainframe computers, massive machines that integrated hardware and software into one appliance. Over time, these functions were unbundled and different companies were created for each: Intel made the chip, Dell assembled the computer, Microsoft made the operating system and Oracle made the database. Previously, one company (IBM) made it all.
This concept dovetails nicely with that of “jobs to be done,” an idea from Clay Christensen, the father of disruption theory. Customers “hire” a product or service to get a job done. When a customer hires an Uber, getting the Uber isn’t the job; the job is getting from point A to point B. This job can be unbundled from car ownership, and from the car altogether. Uber can help you accomplish the job by offering up a scooter, bike or public transportation, depending on your preferences. (Economist Theodore Levitt, on this point, once wrote that “people don’t want to buy a quarter-inch drill; they want a quarter-inch hole.”)
This focus on jobs to be done has led to the further unbundling of compute functionality into “primitives,” or basic building blocks that can be accessed over the internet and paid for on a usage basis. The most well-known example of this is Amazon Web Services (AWS), which is Amazon’s cloud computing division. Intel’s chip is abstracted into a virtual machine, and customers can rent different types of VMs for different jobs. The operating system is abstracted away as well, and customers can run whatever operating system they want. Oracle’s database functionality is unbundled into different types of databases for various jobs, all at a fraction of the cost. All of AWS’s functionality can be accessed over the internet through APIs, and developers around the world can build any application on top of these building blocks.
Before centralized electric utilities in the early 20th century, manufacturing companies owned their own power plants and had teams of technicians to keep the machinery supplying electricity. Samuel Insull, who had worked under Thomas Edison, was one of the first to understand the value of economies of scale in building large, centralized power plants, obviating the need for individual manufacturers to maintain their own. High fixed costs and heavy capital expenditures could instead be converted into variable costs. The scale afforded by a centralized operation meant lower costs for every customer than any one customer could achieve on its own.
Thomas Edison's 1882 Pearl Street Central Station
Source: Edison Papers at Rutgers University
AWS does the same for the compute economy, but on steroids. The division was created within Amazon almost by accident. The company wanted to get the business and engineering teams within Amazon to stop fighting over resources, and instead to talk to each other through hardened APIs. At the time, Amazon’s CEO Jeff Bezos had become enamored with the idea of “primitives,” or basic building blocks, and wanted Amazon’s engineering team to “break its infrastructure down to the smallest, simplest atomic components and allow developers to freely access them with as much flexibility as possible. As Bezos proclaimed at the time, according to numerous employees: ‘Developers are alchemists and our job is to do everything we can to get them to do their alchemy.’ Bezos directed groups of engineers in brainstorming possible primitives. Storage, bandwidth, messaging, payments and processing all made the list. In an informal way—as if the company didn’t quite know the insight around primitives was an extraordinary one—Amazon then started building teams to develop the services described on that list.” 
This was happening around 2002-2003. By 2012, when AWS held its first Re:Invent developer conference, revenues for the division were estimated at around $2.2 billion. Only six years later, they’re at $27 billion and growing 45 percent year over year. AWS is the fastest-growing enterprise software business in history, and based on the size of the addressable market and AWS’s commanding leadership, I believe this division will within a decade be worth more than all of Amazon is worth today.
AWS was officially launched in its current form in March 2006, with the announcement of S3, AWS’s simple storage service. The press release expressed the ethos that has guided AWS’s services:
Amazon S3 is storage for the Internet. It's designed to make web-scale computing easier for developers. Amazon S3 provides a simple web services interface that can be used to store and retrieve any amount of data, at any time, from anywhere on the web. It gives any developer access to the same highly scalable, reliable, fast, inexpensive data storage infrastructure that Amazon uses to run its own global network of web sites. The service aims to maximize benefits of scale and to pass those benefits on to developers.
"Amazon S3 is based on the idea that quality Internet-based storage should be taken for granted," said Andy Jassy, vice president of Amazon Web Services. "It helps free developers from worrying about where they are going to store data, whether it will be safe and secure, if it will be available when they need it, the costs associated with server maintenance, or whether they have enough storage available. Amazon S3 enables developers to focus on innovating with data, rather than figuring out how to store it."
At the time of the announcement, engineer James Hamilton was an employee at a Microsoft subsidiary. Ten years later, then an engineer at AWS, he wrote the following about S3’s announcement:
I felt like I understood the customer value of cloud hosted services. I knew customers loved the speed of provisioning and low cost so, in many ways, I was already a convert. I already felt pretty sure that cloud hosting was the future.
But still the Amazon Simple Storage Service announcement was an eye opener for me. The technology industry has 100s of announcements each day and I don’t look at many. For the most part, they are uninteresting. But the S3 announcement was game changing. Most startling was the cost of the service. It was nearly 2 orders of magnitude less expensive than we were currently paying for multi-data center redundant storage. But what was even more disruptive was a credit card was all that was needed to provision storage. There was no required proposal for financial approval, there was no RFP, no vendor selection process, no vendor negotiation, and no data center space need be found. I could just sign up and start working.
Around the time of AWS’s creation, Jeff Lawson became one of the company’s earliest product managers. A serial entrepreneur, Lawson had always encountered the same problem at his startups: setting up enterprise-quality phone systems and call centers was always a cumbersome process involving lots of copper wires, racks of equipment, dozens of expensive engineers, millions of dollars and months if not years of deployment time.
To a software developer like Lawson, this was anathema. Developers like to prototype and iterate, in periods of weeks and for a very low cost, if not for free, until they have a beta version of a product that can be tested and scaled.
So in 2008, Lawson co-founded Twilio to solve these problems. In Lawson’s words, while using software to create applications online at that point was well understood, making a phone ring seemed “like magic.” This is how venture capitalist Fred Wilson, whose Union Square Ventures led Twilio’s first A round of funding for $3.7 million, described Lawson’s pitch:
It was 2008 in our old offices on the 14th floor of the building we still work in. My partner Albert, who led our investment in Twilio, had met Jeff and was impressed with him and his vision for Twilio. He asked me if I would meet with him and so I did.
Jeff came into the conference room, sat down, and said “we have taken the entire messy and complex world of telephony and reduced it to five API calls”.
I said “get out of here, that’s impossible.”
Jeff proceeded to reel them off and I said “wow”.
He then pulled out his laptop, fired up an editor, and started live coding an app. He asked me for my cell phone number and within 30 seconds my phone was ringing.
I said “you can stop there. that’s amazing”.
It was, and remains, the best seed pitch I’ve ever gotten. I’ve told him that many times and have told this story many times.
Jeff Lawson brought to Twilio many of the principles he learned from the other Jeff (Bezos) at Amazon, such as a love of invention, being frugal, and having a relentless focus on the customer. And because he could assume that AWS’s infrastructure was there—compute, storage, databases, etc., on-demand and pay-per-use—there was no need to invest in owned data centers or heavy infrastructure. Similar to the way factories outsourced their electricity needs to Sam Insull’s centralized power stations over 100 years ago, it’s more cost effective and convenient to outsource compute to AWS. So Twilio built its entire business on top of AWS. This allowed Twilio to quickly scale its business atop AWS’s global infrastructure.
Twilio’s master plan is as follows: first, work to migrate existing communications workloads onto Twilio, and then, to make sure that future workloads of communications are invented on top of Twilio.
To accomplish this, three steps are required. The first one is to build a broad platform that is widely applicable, priced aggressively and designed to enable developers’ creativity to flourish across the widest set of use cases imaginable. To do this, Twilio built high-quality offerings across a variety of lower-level building blocks, like voice and messaging, and then priced these offerings to maximize their utility across a broad set of use cases, some of which haven’t even been invented yet.
Then comes the second step: based on the most common use cases that emerge from that broad platform, Twilio executes a strategy to drive deep penetration in each. They listen to their customers, watch what use cases they’re building for on the platform, and bring those learnings to a wider audience.
The third step is to take those use cases and turn them into higher-level APIs. For example, Twilio noticed that many customers were using its lower-level building blocks to build call centers. Why not do more of the heavy lifting on their behalf? This year, Twilio launched Flex, its call center API.
Previously, call centers were complex affairs. After millions of dollars and years of paying consultants to customize software and hardware, it was possible to get a working call center. It could be fully customized, but any change would require the same army of consultants and a lot of money and time. It also sat on premises and required a lot of maintenance.
Then, the software-as-a-service (SaaS) model emerged; customers could subscribe to a cloud-based call center software. There was no army of consultants to hire and no hardware to maintain. Deployment was quick. This solution was perfect, except for the fact that customizing SaaS solutions isn’t easy.
Twilio Flex solves both problems: it offers all the flexibility and integration of an on-premises, custom-designed system, with the low-cost, scalability and availability of a cloud-based SaaS solution. It also excels in a third dimension: customers can iterate and experiment in real-time just as they can with any other web-based software.
I saw a live demo of Flex and it was mind-blowing. In just a few minutes, anyone can log into their Twilio account and spin up a complete call center with call recording, sentiment analysis, interactive voice responses, all with full flexibility. Lawson’s vision for the future of the contact center is one in which you call and the agent knows exactly who you are and what problem you’re calling about (“I see you’re scheduled to be on flight 237, and it’s been cancelled. Can I put you on the next flight?”). It’s a future in which you don’t have to repeat the same information to multiple agents as you make your way across departments. The agent can push a card to your smartphone with the new flight information, and you can approve and authenticate using your fingerprint or facial recognition. All of this is fully customizable, fast, low risk (allowing endless experimentation), resilient and scalable.
I should note that Flex is an offering that just went live last month; it hasn’t yet contributed to Twilio’s revenues, which look like this thus far:
Source: Twilio's SEC filings
Flex is shaping up to be a strong revenue contributor. The first indicator was the level of interest at Twilio’s developer conference in October, even before Flex became generally available; I attended the conference and witnessed the enthusiasm for Flex. Then, in early November, Twilio’s COO George Hu mentioned that “90 percent of the world’s contact center infrastructure is still on-premises and I think there’s a lot of developers that want to move to the cloud… we’re still in the early innings of that… I’m very impressed with the turnout in the partner summit at SIGNAL [the developer conference]… I’m very convinced the reason it was overflowing was because of Flex.”
Flex is part of the top layer of APIs, which Twilio calls its Engagement Cloud. It includes APIs like Authy, for tokenized user authentication and other functionality. One layer below that is Twilio’s Communications Cloud, which has primitives for programmable SMS, voice, chat, fax, video, and integrations with popular messaging apps like WhatsApp and Facebook Messenger. And finally, at the bottom is Twilio’s Super Network. This is the layer of software that integrates directly with telecom operators all around the world. If a multi-national needs toll-free numbers in dozens of countries, previously it would have to contact telco providers in each location and go through a cumbersome process; with Twilio, it’s just a few API calls.
Twilio has built its Super Network for high availability. Downtime at the level of local carriers gets automatically re-routed, making Twilio’s network more reliable than its underlying carriers. As customers grow their use of Twilio’s APIs, Twilio gets better data; this data allows Twilio to further improve the Super Network, resulting in a flywheel effect of increasing returns to scale and an improved competitive advantage over time.
Similar to AWS, Twilio’s actual customers are developers. The company’s goal is to be in every developer’s toolkit. When a problem involving communications arises, Twilio is top-of-mind and the developer can quickly begin prototyping a product. The sales cycle is inverted: instead of a company’s Chief Technology Officer soliciting proposals from several vendors, holding a bake-off, then selecting a winner based on price and feature set, Twilio wins from the bottom up, having developers inside large companies act as ambassadors on its behalf.
Source: Twilio; chart as of May 31, 2016.
Developer accounts as of early 2020 are around 6 million.
As Jeff Lawson noted at Twilio’s developer conference in October, no company ever differentiated itself because of its PBX system; but engagement, that’s how you differentiate yourself. Twilio is building up a powerful business around customer engagement; that month, Twilio announced the acquisition of SendGrid, a publicly traded company that looks similar to Twilio with a go-to-market strategy focused solely on developers. SendGrid focuses on developing the world’s best APIs for email delivery.
Email delivery is far from a “solved problem.” SendGrid has spent nearly a decade developing the plumbing to make sure that transactional emails arrive in your inbox quickly (just a couple of seconds) and don’t end up in the spam folder. This might sound trivial, but SendGrid has built this in to a business with around $130 million in yearly revenues growing consistently over 30 percent; the company has over 74,000 customers, delivers emails to over 50 percent of the world’s email addresses, and sends out over 1.5 billion emails every day.
Using the framework of jobs to be done, Twilio is addressing the job of customer engagement. One of the biggest requests by Twilio developers was email functionality; meanwhile, SendGrid customers wanted voice, SMS and other channels like WhatsApp added to email. With the merger of Twilio and SendGrid, that job becomes easier to do.
Like the AWS developer conference, Twilio’s was filled with interesting case studies of customers solving real pain points and saving money in the process. Spirit Airlines, for instance, had a call center with an average wait time of 45 minutes. Nearly two thirds of the calls were for password resets, which impacted its pilots (they couldn’t log into their iPads to access flight plans), leading to delayed flights. Twilio’s solution reduced the 45-minute wait to a few seconds through SMS-based authentication and password resets. Spirit says this created a 95 percent cost savings for that workflow.
While there are some competitors for aspects of Twilio’s business—for instance, AWS has Connect, which is a call center API—none of them are as reliable, high quality and easy to use as Twilio’s solutions (Connect might be as reliable, but it certainly isn’t as easy to deploy). Indeed, Twilio’s offerings are so good, AWS itself uses Twilio to power some of the APIs AWS offers (Amazon was also an early investor in Twilio).
It is hard to overstate the energy and developer enthusiasm at Twilio’s developer conference this year; there is no other company doing anything close to what Twilio is doing in communications APIs. As Twilio grows and innovates into higher level APIs, and as its Super Network becomes more reliable and cost-effective, it’s possible that Twilio reaches a tipping point and captures so much developer mindshare and brand awareness, it gets to the point where “you can’t get fired for using Twilio.”
I am also very impressed by the caliber of talent Twilio has been able to attract. Above I mentioned COO George Hu, who came to Twilio recently from cloud pioneer Salesforce. There, he spent 13 years growing the company from $20 million in revenues to over $5 billion. Hu noted that “one of the reasons I’m here at Twilio is because I see the same type of opportunity, a massive market ready to be disrupted, an amazing set of technologies […], great visionary leadership team, and I’m here to help accelerate this company’s growth in the future.”
On Twilio’s board sits Rich Dalzell, a much-beloved, now retired, Amazon executive. Dalzell is featured prominently in Brad Stone’s book The Everything Store. Dalzell had a career at Walmart and then spent ten years at Amazon, where he rose to Chief Information Officer.
Given its disruptive business model and large market, Twilio has been growing briskly. Revenues are running at around $560 million per year, and growth has been exceptional, north of 40 percent year-over-year. Because of its self-serve business model, where any developer anywhere in the world can start prototyping for very little money and build larger applications over time, Twilio’s revenues per customer cohort tends to grow over time. The image below is an example of this revenue expansion:
This year’s third quarter revenue growth came in at a whopping 68 percent. Given its high gross margins, low expenses for sales and marketing, and the upcoming merger with SendGrid, I believe we built a position in Twilio stock at a very attractive entry point. I expect our investment to compound in the 20 percent range over the long term using what I believe are reasonable growth, margin and exit multiple assumptions.
 Quoted from The Everything Store by Brad Stone.