Maximizing Value of the Distributed Edge

Throughout my career I have focused on driving efficiency, high performance, and scale through standardization of the foundation layers of digital infrastructure—data center, hardware, network and infrastructure management. In 2007, I celebrated 14 years at Sun Microsystems by launching the Performance Optimized Datacenter (POD), a modular architecture that enabled any rack of any density to be placed anywhere. This solution had blossomed from internal data center projects including the launch of the Sun Grid in February of 2005. For those who don’t remember those glory days, Sun launched the $1 per cpu-hour grid to enter the burgeoning new industry wave we now call “cloud.” Sun was a year ahead of AWS. While Sun had superiority in hardware, operating system, and enterprise solutions, they missed the market opportunity to provide a platform that would make it easier for customers to migrate to their offering. AWS got it right, and as everyone knows, the rest is cloud history. However, I believe there is one concept from Sun that is even more relevant today. David Douglas  Sun’s first Chief Sustainability Officer, said sustainability is about balancing economics and ecology. They can, and should be, complementary. You don’t have to sacrifice one for the other. I’ve applied that logic ever since.

When I joined eBay in late 2009, they were at the beginning of a turn around. Fast forward three years and we had grown the infrastructure 4x and the company was continuing double digit revenue growth. My team was able to slash the internal unit costs of foundational infrastructure by 50%, while delivering some of the highest performing deployments in the world. Among my favorites were the modular data centers we installed in Phoenix that achieved PUEs of 1.018. That meant >98% of the power went to useful work. How did we do that? Alignment and standardization. The data center, hardware, and network elements acted as a system versus discrete functions operating in silos. At that time, standardization and rack-and-roll had been adopted in many places, including Facebook’s launch of the Open Compute Project (OCP). Much to the dismay of leading OEMs, Facebook decided to release their designs allowing anyone to benefit from their efforts (thank you, Frank Frankovsky!). OCP had a major influence on the industry, but in my humble opinion they missed one thing—price. Adopters could not achieve the same discount rates as Facebook because they didn’t have the same volumes. This gave me an idea. What if we could create a buyers co-op with what we had done at eBay? If others could adopt the standard configs, we could all buy together and have the same benefits. Fast forward to 2014, and the idea had solidified. On a sunny September afternoon in San Jose, my business development partner and I pitched the idea to Bob Swan, eBay’s CFO. He thought the idea had merit but said he couldn’t commit. In two weeks we found out why. On September 30, 2014 eBay Inc. announced that the board had approved splitting eBay and PayPal. As they say, timing is everything. This was not the right time.

Over the next 10 months, we focused everything on executing the split. On July 17, 2015 the gavel fell at the NYSE successfully splitting eBay and PayPal’s $255Bn global commerce engine into two standalone Internet companies. It was an incredible project, but the hardware co-op idea was still in my mind.

I left eBay in 2016 and started developing a company to build an independent bare metal marketplace that would enable a buyers co-op. Over that next six months the idea took shape but one major roadblock continued to surface. Building out capacity for a bare metal marketplace was extremely capital intensive. Partners were interested but the question always came back to who would hold the assets on their balance sheet? Bottom line, competing with cloud companies who have already invested billions in locations all over the world was a non-starter.

Around that time a new opportunity emerged when an industry friend sent me a job description from Uber. I remember reading it and thinking that I couldn’t have written it better myself. What also came to mind was that Uber could be the catalyst for the bare metal co-op I’ve been toying with for years.

I joined Uber in September of 2016 and established the Uber Metal team. Over the next three years, we focused on standardization and scale. The growth was amazing. We quickly surpassed eBay’s portfolio size and continued to drive optimization, replacing all but one legacy data center and refreshing >80% of the server, storage and network capacity. We also established a hardware original design manufacturing (ODM) team that developed our own standard SKUs like Facebook, Google, Amazon and Microsoft. These elements enabled the Uber Metal reference architecture with building blocks for on-prem zones and regions in our hybrid portfolio. In late 2018, the time had come to take our work to the next level. I pitched the bare metal buyers co-op and the potential to enter into a bare metal marketplace. We had the designs, volumes, partner relationships and supply chain capabilities to make this a reality. Unfortunately, like eBay, this was not in the cards. While the idea had merit and Uber’s balance sheet could allow this to happen, it was not part of the larger strategy.

When I left Uber in 2019, I had pretty much given up on the idea. It seemed like ‘three strikes and you’re out’ was an apt saying at that point. Little did I know that in just a few short months, the stars would start to align again and show me a new path. One that had all the right pieces in place.

In the summer of 2019 I reconnected with a long time industry friend, John Cowan, who had a little startup called EDJX. He and James Thomason, a digital infrastructure executive with 13 startups under his belt, co-founded the company along with corporate finance professional Michael Preston. They asked if I would be interested in joining their advisory board. They explained their vision to make it easier for developers to put code and data closer to devices, machines and users than ever before. Needless to say, I was hooked! They were building the next generation edge platform.

EDJX understood that the scale of the edge would require a new approach for data center hardware. Their platform leverages circular economy equipment from a company called ITRenew. Another bonus was that ITRenew hardware was already standardized. It included proven hyperscale designs from the likes of Facebook, Microsoft, and others, even Uber and eBay!  ITRenew takes this equipment, refreshes and recertifies it, and then resells it as standard hardware with a three year warranty. Because it is recertified, they can offer a price advantage of approximately 50% with performance parity compared to new equipment. I was also shocked at their volumes. If they were to sell all the hardware they had in their warehouses as servers, they would be the fourth largest OEM in the world. Economic benefit? Check. And because of the circular economic model, deployment of the ITRenew equipment avoids a ton of new manufacturing. This means more than 70% of the embedded carbon is avoided by repurposing IT equipment. Ecology benefit? Check. But the story doesn’t end there.

In September of 2019, I joined the board of Virtual Power Systems (VPS). VPS is tackling the last big opportunity area in data centers—unlocking stranded power capacity. In February of 2020, I stepped in as CEO and zeroed in on our initial target customers—colocation providers. Why? They have the most stranded power capacity because they are at the mercy of their customers’ SLAs and workloads. As we engaged further, a clear pattern emerged. There are pools of stranded power capacity in every data center lost to safety buffers, low utilization and rarely-used redundant infrastructure. The punchline is that 40% of built data center power capacity isn’t used. Money is being left on the table everywhere. We saw a powerful opportunity in partnering with EdjX and ITRenew to help colocation providers not only increase utilization and efficiency, but also develop new revenue streams in the emerging edge economy and traditional colocation verticals.

In October of 2020, EDJX announced the alliance between EDJX, ITRenew, and VPS to land EdjBlocks in data centers all over the world. Cyxtera was part of the launch as the first major global data center operator to adopt this strategy. The platform will be generally available to the market starting in January of 2021. With a 50% price/performance advantage, a 70% reduction of embedded carbon and the ability to tap into the 40% of stranded capacity to power these deployments, the EDJX platform is destined to be one of the most sustainable deployments in the digital infrastructure industry. To make the deal even sweeter, the EDJX economic model is decentralized, which means partners earn a share of every dollar of revenue generated on the EDJX platform.

From my perspective, the combination of EDJX and ITRenew creates a competitive advantage unlike anything we have seen before. Software developers and CDN users around the world will be able to leverage a cost-optimized, low-carbon, highly efficient platform alternative to cloud. As I write this article more than a decade later, I realize three things. First, David Douglas was right. Sustainability is about balancing economics and ecology. You don’t have to sacrifice one for the other. Second, marketplaces work. You just need to find the right type and the right time for them to thrive. While this new marketplace isn’t a buyers co-op, it is delivering the same result—a standardized, low cost alternative at scale. Third, forget the third—the FOURTH time IS the charm.