Chapter 7
A VIRTUAL TREE FALLS IN A VIRTUAL FOREST. IN the previous two chapters, I explained what is required for the tree to be rendered, and for its fall to be processed and then shared and thus known to any observer. But what is this tree? Where is this tree? What is the forest? The answer is data and code.
Data describes the attributes of a virtual object, such as its dimensions or color. For our tree to be processed by a CPU and rendered by a GPU, this data needs to be run by code. And if we want to chop down that tree and use its wood to build a bed or light a fire, that code must be part of a much broader framework of code* that operates the virtual world.
The real world isn’t altogether different. The laws of physics are the code that reads and runs all interactions—from the reasons a tree falls to how this produces vibrations in the air that travel to a human ear causing nerves to relay information via electrical signal across various synapses. Similarly, a tree “seen” by a human observer means it reflected light produced by (usually) the sun, light that is in turn received and processed by the human eye and brain.
But there is a key difference: the real world is fully preprogrammed. We can’t see X-rays or echolocate, but the requisite information already exists in the world. In a game, X-rays and echolocation require data and a lot of code. If you go home, mix ketchup and petroleum, and then try to eat it or paint with it, the laws of physics take care of the results. For a game to manage the same interaction, it needs to know in advance what ketchup and petroleum do when they’re combined (and probably in generic ratios), or it needs to know enough about the two for the game’s logic to figure it out, assuming the game is capable.
A virtual world’s logic might say that petroleum cannot be mixed with anything. Or that it can only be mixed with oil. Or that if it’s mixed with anything at all, it produces unusable sludge. But a more complicated result requires considerably more data and for the virtual world’s logic to be far more comprehensive. How much petroleum can be added to ketchup before it’s inedible? How much ketchup can be added to petroleum before it’s unusable? How does the resulting substance’s color and viscosity change based on the ratio of one ingredient to the other?
The fact that so many of these permutations have little value is, in fact, hugely valuable to those who produce virtual worlds. Because the hero of The Legend of Zelda does not need to go to space, no space-based physics are required. Call of Duty’s players don’t need kayaks, enchantments, or baking goods; the game’s developer did not create the relevant code. Nintendo and Activision could focus more data and code on what their virtual worlds need and benefit from, rather than infinite permutations that have limited practical value to their games.
For all its efficiency, this approach introduces obstacles to building Metaverse-like virtual worlds, and especially in establishing interoperability among them. In Microsoft Flight Simulator, for example, a pilot can land a helicopter beside a football field, but there’s no football game he can then watch, let alone join. For Microsoft to offer such functionality, it would need to build from scratch its own football system, even though many developers have already done so and, with years of experience, are likely better at it, too. While MSFS could instead try to integrate into these football-specific virtual worlds, the data structures and codes on every side are likely incompatible. In the networking and compute chapters above, I discussed the fact that users’ devices are often performing the same work. But if a comparison can be made, developers must be said to be even worse. They are constantly building and rebuilding everything from a football field to a football and even rules for how a football flies in the air. What’s more, this work is getting harder every year as virtual world builders look to take advantage of more sophisticated CPUs and GPUs. According to Nexon, one of the world’s largest video game publishers, the average number of staff credits for an open world action game (think The Legend of Zelda or Assassin’s Creed) has grown from roughly 1,000 in 2007 to over 4,000 by 2018, with budgets growing by a factor of 10 (roughly two and half times faster).1
To hear trees fall, to have them fall near football fields, and to have the sound of their falling added to the roars of a crowd responding to a game-winning touchdown requires lots of programmers writing lots of code for handling vast amounts of data and all in the same ways.
Now that we’ve covered the networking and computing power required to share, run, and render the data and code required for the Metaverse, we can turn to these latter concepts.
Game Engines
The concept, history, and future of the Metaverse are all intimately tied to gaming, as we’ve seen, and this fact is perhaps most obvious when we look at the basic code of virtual worlds. This code is typically contained in a “game engine,” a loosely defined term that refers to the bundle of technologies and frameworks that help to build a game, render it, process its logic, and manage its memory. In a simplified sense, think of the game engine as the thing that establishes the virtual laws of the universe—the ruleset that defines all interaction and possibilities.
Historically, all game-makers built and maintained their own game engines. But the past fifteen years have witnessed the rise of an alternative: licensing an engine from Epic Games, which makes the Unreal Engine, or one from Unity Technologies, which makes an eponymous engine.
Using these engines has a cost. Unity, for example, charges every individual developer that uses it an annual fee. This fee ranges from $400 to $4,000, depending on the features required and the size of the developer’s company. Unreal typically charges 5% of net revenues. Fees aren’t the only reason to build your own engine. Some developers believe doing so for a given game genre or experience, such as realistic and fast-paced first-person shooters, ensures that their games “feel better” or perform better. Others worry about the need to rely on another company’s pipelines and priorities, or fret that their vendor has such a detailed view into their game and its performance. Given the concerns, it’s common for large publishers to build and maintain their own engines (some, such as Activision and Square Enix, even operate a half dozen or more).
Most developers, however, see strong net positives to licensing and then customizing Unreal or Unity. Licensing allows a small or inexperienced team to build a game with an engine more powerful and extensively tested than they could ever build—and that is less likely to fail and can never run over budget. In addition, they can focus more of their time on what will differentiate their virtual world—level design, character design, gameplay, etc.—instead of the basic technology required for it to run. And rather than hiring a developer and then training them to use or build on a proprietary engine, they can instead appeal to the millions of individual developers already familiar with Unity or Unreal and immediately put them to work. For similar reasons, it’s also easier to integrate third-party tools. An independent start-up that makes, say, face-tracking software for video game avatars doesn’t design their solution to work with a proprietary engine they’ve never used, but instead to work with those chosen by the largest number of developers.
A good analogy is designing and building a house. Neither the architect nor the decorator designs proprietary lumber dimensions, assembly hardware, measurement systems, blueprint frameworks, or tools. Not only does this make it easier to focus on the creative work, but it makes it easier to hire carpenters, electricians, and plumbers. If the house ever requires a renovation, another team can more easily modify the existing structure because they don’t need to learn new techniques, tools, or systems.
This analogy has a key shortfall, however. Houses are built once and in one place. Games, conversely, are designed to run on as many possible devices and operating systems as possible—some of which haven’t yet been developed, let alone released. As a result, games must be compatible with, say, different voltage standards (e.g., the UK’s 240 volts and America’s 120 volts), measurement systems (imperial and metric), conventions (aerial telephony wires and buried ones), and so on. Unity and Unreal build and maintain their game engines so that they’re not just compatible with but optimized for every platform.†
In a sense, we can think of independent game engines as a shared R&D pool for the industry. Yes, Epic and Unity are for-profit companies, but instead of every developer sinking part of their budget into proprietary systems to manage core game logic, a few cross-platform technology providers can concentrate a portion of their budgets into a more capable engine that supports, and benefits, the entire ecosystem.
As the major game engines developed, another type of independent gaming solution emerged: live services suites. Companies such as PlayFab (now owned by Microsoft’s Azure) and GameSparks (Amazon) operate much of what a virtual world needs to “run” online and multiplayer experiences. This includes user account systems, player data storage, processing in-game transactions, version management, player-to-player communications, matchmaking, leaderboards, game analytics, anti-cheat systems, and more, all of which work across platforms. Both Unity and Epic now have their own live services offerings, too, which are available at low-to-no cost and are not limited to their engines. Steam, the world’s largest PC game store and a key point of discussion in Chapter 10, offers its own live services product, Steamworks.
As the global economy continues to shift to virtual worlds, these cross-platform and cross-developer technologies will become a core part of global society. In particular, the next wave of virtual world builders—not game-makers, but retailers, schools, sports teams, construction companies, and cities—are likely to use these solutions. Companies like Unity, Unreal, PlayFab, and GameSparks are in an enviable position. Most obviously, they become a sort of standard feature, or lingua franca, for the virtual world—think of them as the “English” or “metric” of the Metaverse. Just as it is likely that you use some English and some knowledge of the metric system when traveling internationally, odds are that if you’re building something online today, irrespective of what it is you’re building, you are using—and paying—one or more of these companies.
But more importantly, who better to establish common data structures and coding conventions across virtual worlds than the companies that govern their logic? Who better to facilitate exchanges of information, virtual goods, and currencies between these virtual worlds than the companies that power the same inside them? And who better to create an interconnected network of these virtual worlds, as ICANN does for web domains and IP addresses? We’ll come back to these questions and this presumptive answer, but first we need to consider a path some think of as the easier, and best, way to build the Metaverse.
Integrated Virtual World Platforms
As both independent game engines and live services suites developed over the past two decades, other companies combined these approaches into a new one: integrated virtual world platforms (IVWPs) such as Roblox, Minecraft, and Fortnite Creative.
IVWPs are based around their own general-purpose and cross-platform game engines, similar to Unity and Unreal (Fortnite Creative, or FNC, which is owned by Epic Games, is built using Epic’s Unreal Engine). However, they are designed so that no actual “coding” is required. Instead, games, experiences, and virtual worlds are built using graphical interfaces, symbols, and objectives. Think of it as the difference between using the text-based MS-DOS and visual iOS, or designing a website in HTML versus creating one in Squarespace. The IVWP interface enables users to create more easily and with fewer people, less investment, and less expertise and skill. Most Roblox creators, for example, are kids, and nearly 10 million users have created virtual worlds on Roblox’s platform.
In addition, every virtual world built on these platforms must use the platform’s entire live services suite—its account and communication systems, avatar database, virtual currency, and more. All of these virtual worlds must be accessed through the IVWP, which therefore serves as a unified experiential layer and a single installer file. In this sense, building a world on Roblox is more like constructing a Facebook page than a Squarespace website. Roblox even operates an integrated marketplace for developers where they can upload anything they custom-made for their virtual world (e.g., a Christmas tree, a snowed-on tree, a barren tree, a pine bark texture) and license it to other game makers. This provides developers with a second source of income (developer-to-developer rather than just developer-to-player), while also making it easier, cheaper, and faster for others to build their virtual worlds. The process also drives further standardization of virtual objects and data.
Though it’s easier for a developer to build a virtual world using an IVWP than a game engine like Unreal or Unity, it’s harder to build an IVWP than a game engine in the first place. Why? Because for an IVWP, everything is a priority. An IVWP wants to enable creators’ creative flexibility while also standardizing underlying technologies, maximizing interconnectivity among everything that’s built, and minimizing the need for training or programming knowledge on the part of creators. Imagine if IKEA wanted to build a country as dynamic as the United States, but force all buildings to use IKEA prefabs. Additionally, IKEA would be in charge of the new country’s currency, utilities, police, customs, and more.
A good way to understand how hard it is to operate an IVWP was provided to me by Ebbe Altberg, the former CEO of Second Life. In the mid-2010s, one of the platform’s developers created a business selling virtual horses, alongside a subscription to virtual horse feed. Later, Second Life upgraded its physics engines, but a bug resulted in horses sliding past their feed whenever they tried to eat it. As a result, the horses starved and died. It took time for Second Life to even know this bug existed, and more time still to fix it, then to provide the appropriate redress to those affected by it. Still, such events disrupt Second Life’s economy, while also producing distrust in the market, which harms both buyers and sellers. Finding a way to constantly improve functionality, while continuing to support old programming and without errors, is an extraordinary task. Game engines also face a version of this problem. However, when Epic updates Unreal, it is up to every developer to deploy this update, and they can do so at the time of their choosing, after extensive testing, and without worrying about how that update affects their interactions with other developers. When Roblox pushes an update, it automatically reaches all of its worlds.
At the same time, the fact that a “virtual IKEA” is built on programming, not particle board, means that its potential is bound not by literal physics, but the nearly boundless potential of software. Anything made in Roblox, by the Roblox Corporation or its developers, can be endlessly repurposed or copied at no marginal costs. They can even be improved upon. Every developer in an IVWP is effectively collaborating to populate an ever-expanding and increasingly capable network of virtual worlds and objects. As this network improves, it becomes easier to attract more users and more per-user spending, which leads to more network revenue, and then more developers and investment, and thus further improvements to the network, and so on. This is the benefit of pooling not just engine R&D, but, well, R&D for everything.
But what does this look like in practice? The Roblox Corporation offers the best answer at the moment, given that Fortnite Creative is managed by Epic Games, which remains private, and Minecraft’s financials are not disclosed by its owner, Microsoft.
Start with engagement. By January 2022, Roblox was averaging more than 4 billion hours of usage per month, up from roughly 2.75 billion a year earlier, 1.5 billion the year before that, and 1 billion at the end of 2018. This excludes time spent watching Roblox content on YouTube, which is the world’s most used video site and reports that gaming content is its most watched content category, and Roblox its second most popular game (Minecraft, another IVWP, ranks first). As a point of contrast, Netflix is estimated at 12.5 billion to 15 billion hours of use per month. All of the top Roblox games, such as Adopt Me!, Tower of Hell, and Meep City, come from independent developers with little to no prior experience and staffs of 10 to 30 (having started with one or two). To date, these titles have been played 15 to 30 billion times each. In a single day, they’ll reach half as many players as Fortnite or Call of Duty—and half as many as titles like The Legend of Zelda: Breath of the Wild or The Last of Us do in their lifetimes. And as for populating the platform with a wide range of virtual objects? 25 million items were made in 2021 alone, with 5.8 billion being earned or bought.2
Part of Roblox’s surging engagement is driven by its growing userbase. From Q4 2018 to January 2022, average monthly players increased from an estimated 76 million to more than 226 million (or 200%), while average daily players grew from around 13.7 to 54.7 (or 300%). You’ll note that daily players grew more than the monthly userbase, and engagement grew by an even larger volume (400%). Not only is Roblox becoming more popular overall; it’s becoming more popular with its users, too. We can see similar evidence of Roblox’s network effects in its financials. Roblox’s revenues are up 469% from Q4 2018 to Q4 2021, while its payments to on-platform world builders (i.e., developers) have grown 660%. In other words, the average Roblox user is spending more per hour than ever before and generating revenues faster than ever before, and with growth in these two metrics exceeding the already impressive growth in users, which is then exceeded by the growth in compensation to developers. Furthermore, Roblox’s growth has been disproportionately concentrated among older audiences. At the end of 2018, 60% of daily users were under 13. Three years later, only 21% were. Put another way, Roblox ended 2021 with nearly two and half times as many players over 13 as the service had under 13 in 2018.
The most impressive aspect of Roblox Corporation’s flywheel may be its investments in R&D. In the first quarter of 2020, the last before the COVID-19 pandemic, the company generated roughly $162 million in revenue and invested $49.4 million in R&D. That means 30 cents of every dollar spent on Roblox went back into the platform. Over the following seven quarters, Roblox’s revenue surged more than 250%, totaling $568 million in the fourth quarter of 2021. However, Roblox did not divert this revenue to profits, nor any alternative uses. Instead, it continued to reinvest in R&D—at roughly the same rate as before. As a result, the company spent more on R&D in Q4 2021 than it generated in revenue in Q1 2020. In 2022, Roblox R&D may top $750 million and by the end of the year, it may approach $1 billion on an annualized basis.
As points of contrast, consider Rockstar’s Grand Theft Auto V and Red Dead Redemption 2. GTA:V is the second-best-selling game in history, with over 150 million copies sold (Minecraft ranks first with nearly 250 million). RDR2 was the best-selling title made for the eighth generation of consoles (i.e., PlayStation 4, Xbox One, Nintendo Switch), with 40 million copies sold. The two games are also believed to be among the most expensive game productions ever, with final budgets estimated at $250 million to $300 million and $400 million to $500 million, respectively, which includes more than half a decade of development each, plus extensive marketing and publishing costs. Or compare Roblox’s R&D budget to that of Sony’s PlayStation group, which topped $1.25 billion in 2021 and spanned close to a dozen gaming studios, its cloud gaming division, live services group, and hardware division. That same year, Epic Games’ Unreal Engine is believed to have generated less than $150 million in revenue. Unity’s engine brought in much more—roughly $325 million—but still came in 20% short of Roblox’s R&D.
Roblox’s R&D investments are diverse, spanning improvements in developer tools and software, server architecture to synchronize high concurrency simulations, machine learning to detect harassment, artificial intelligence, rendering for virtual reality, motion capture, and more. That Roblox can invest so much into its platform is astonishing. In theory, every additional dollar enables developers to produce more compelling virtual worlds, which attracts more users leading to more revenue—which enables not just more R&D by Roblox, but also by the independent developers who make these worlds, investment that again drives more user engagement and spending on Roblox, leading to more R&D by the company.
Many Virtual Platforms and Engines, Not Many Metaverses
Think back to the definition of the Metaverse I laid out in Chapter 3: “A massively scaled and interoperable network of real-time rendered 3D virtual worlds that can be experienced synchronously and persistently by an effectively unlimited number of users with an individual sense of presence, and with continuity of data, such as identity, history, entitlements, objects, communications, and payments. Some might read this definition and think Roblox is pretty close. It cannot be experienced synchronously and persistently by an effectively unlimited number of users; no real-time rendered virtual world can, at the moment. And when that does become possible, it will surely be true for Roblox. However, Roblox is unlikely to meet my definition in one key way: most virtual works will exist outside of it. This makes it a Metagalaxy, rather than the Metaverse.
But could Roblox become the Metaverse? What if Epic’s IVWP Fortnite Creative, game engine Unreal, and live services suite Epic Online Services, along with its other special projects, were combined—would the result be the Metaverse? If you squint, you might be able to imagine these companies, or one like it, subsuming all virtual experiences, thereby becoming a Metaverse-sized Metagalaxy. And it is notable that some form of this process is what happens in Snow Crash and Ready Player One.
The current state of technological progress, however, suggests another outcome. Why? Because as fast as these virtual giants are growing, the number of virtual experiences, innovators, technologies, opportunities, and developers are all growing faster.
While Roblox and Minecraft are among the most popular games in the world, their reach is modest when considered in the broadest terms. These two supposed titans have 30–55 million daily active users, a fraction of the global internet population of 4.5–5 billion. In effect, they are still at the ICQ stage of virtual words; billions of users and millions of developers have yet to even try them. It’s easy to assume that Roblox or Minecraft will be the primary beneficiaries of this growth, yet history cautions us to be skeptical.
When Microsoft acquired Minecraft developer Mojang in 2014, the title had sold more copies than any other game in history, and also had more monthly active users—25 million—than any AAA video game in history. Seven years later, Minecraft had grown nearly five times in monthly users, but also had ceded its crown to Roblox, which had grown from fewer than 5 million monthly users to over 200 million. Furthermore, the new king boasts nearly twice the daily users as Minecraft had monthly. What’s more, this period included the launch of many other IVWPs. Fortnite didn’t launch until 2017, with FNC coming a year later. Another battle royale, Free Fire, which also counts more than 100 million daily active users globally, released its creative mode in 2021. Although it launched in 2013, Grand Theft Auto V spent much of the past decade transforming from a single-player game into a makeshift IVWP in Grand Theft Auto Online. Sometime over the next few years, the title’s much-anticipated next sequel will release and will doubtlessly take advantage of the successes and learnings from Roblox, Minecraft, and FNC.
As long as there are billions, or even tens of millions, of players left to adopt IVWPs, more will come to market. Krafton, one of South Korea’s largest companies and the creator of PUBG, the first and most popular mainstream battle royale, is surely working on its own offering. In 2020, Riot Games, which makes the most successful game in China, League of Legends, bought Hypixel Studios, which previously operated the largest private Minecraft server before shutting down to develop their own Minecraft-like platform.
Many new IVWPs are being developed around different technical premises, too. At the end of 2021, even the largest of the blockchain-based IVWPs, which includes Decentraland, The Sandbox, Cryptovoxels, Somnium Space, and Upland, had less than 1% of Roblox’s and Minecraft’s daily active users. However, these platforms believe that by allowing users more ownership over their in-world items, as well as a say in how the platform is governed, and a right to share in its profitability, they will be able to grow far more quickly than traditional IVWPs (more on this theory in Chapter 11).
Facebook’s Horizon Worlds is not limited to immersive VR and AR, but it is focused in those areas, which contrasts with Roblox, which is available in immersive VR but prioritizes traditional screen interfaces, such as an iPad or PC screen. Upstarts such as Rec Room and VRChat are also centered on immersive VR world creation, and are rapidly accumulating users. With valuations around $1 billion–$3 billion each at the end of the 2021, the two platforms remain small. But at the start of 2020, Unity Technology and Roblox Corporation were valued at less than $10 billion and $4.2 billion respectively. Two years later, both have valuations that exceed $50 billion. Niantic, the maker of Snap and Pokémon Go, is working on its own augmented reality and location-based virtual world platforms.
These competitors might falter, but it’s more likely that they grow alongside and potentially displace current market leaders. Take Facebook, as an example. The social networking giant entered 2010 with more than half a billion monthly active users, but has failed to subsume any of the hit social media platforms which emerged in the decade. Snapchat launched in 2011, with Facebook launching its own Snapchat-like app (or “clone”) in 2013, called “Poke,” which was shuttered a year later. In 2016, Facebook launched “Lifestage,” its second Snapchat clone, with was also closed after 12 months. That same year, Facebook’s Instagram app also copied Snapchat’s signature “Stories” format, with Facebook’s main app adding the feature the following year. Then in 2019, Instagram launched its own dedicated Snapchat-like app, “Threads from Instagram,” though almost no one noticed. Facebook Gaming, the company’s Twitch competitor, launched in 2018, as did Facebook’s TikTok competitor, Lasso. Facebook Dating released in 2019, with Instagram adding a TikTok-like feature named “Reels” in 2020. Facebook’s efforts have undoubtedly curbed these services’ growth, yet each service is larger than ever and still expanding. By the end of 2021, TikTok had more than billion users and was reportedly the most visited web domain of the year, with Google and Facebook rounding out the top three.
Though the top integrated virtual world platforms are mighty and fast-growing, they also represent a far smaller portion of the gaming industry than Facebook does in the social web. In 2021, the combined revenues of Roblox, Minecraft, and FNC represented less than 2.5% of gaming revenues in 2021, and reached fewer than 500 million of an estimated 2.5 billion–3 billion players. Moreover, they’re dwarfed by the major cross-platform engines. Roughly half of all games today run on Unity, while Unreal Engine’s share of high-fidelity 3D immersive worlds is estimated at between 15% and 25%. Roblox’s R&D expenditures may exceed that of both Unreal and Unity, but this ignores the billions in additional investment made by licensors of these engines. The two most popular games in the world, excluding low-fidelity casual titles such as Candy Crush, are PUBG Mobile and Free Fire, both built on Unity. Most important may be the reach of Unreal and Unity’s developers. While millions of users have made a Minecraft mod or a Roblox game, the number of professional developers using these IVWPs is estimated in the tens of thousands. Epic and Unity count millions of active and skilled developers. And scores of proprietary engines, such as Activision’s IW (Call of Duty) and Sony’s Decima (Horizon Zero Dawn and Death Stranding) continue to receive investment and the games that use them are more popular than ever.
The growing value of virtual worlds and the Metaverse increases the incentives for a developer to in-source their technology stack, as this approach provides a greater opportunity for technical differentiation and greater control over their technology overall, reduces their reliance on third parties who might become competitors,‡ and increases profit margins. Of course, many of these developers will still use Unreal or Unity as a game engine, or GameSparks or PlayFab for live services. However, these providers enable a developer to “pick and choose” what they like, and also to customize much of what they license. Unlike IVWPs, they also allow a developer to manage its own account systems and operate its own in- game economies. These services are also much cheaper, too. Roblox pays a developer less than 25% of the revenue a player spends on their game.§ Epic’s Unreal Engine, conversely, takes only a 5% royalty on revenues. The total cost of Unity’s engine is likely to be less than 1% of a successful game’s revenue. Roblox does take on additional expenses for its developers, such as costly server fees, customer service, and billing, but in most cases a developer will still have higher profit potential by building a standalone virtual world, rather than one inside an IVWP. As such, we should assume that no matter how much more successful Roblox or Minecraft become, they will power only a minority share of all games.
While games and game engines are central to the Metaverse, they don’t come close to encompassing it. Most other categories have their own rendering and simulation software. Pixar, for example, builds its animated worlds and characters using its proprietary Renderman solutions. Most of Hollywood, meanwhile, uses Autodesk’s Maya software. Autodesk’s AutoCAD, along with Dassault Systèmes’ CATIA and SolidWorks, are the primary solutions used to build and design virtual objects that will then be made into real-world ones. Examples include cars, buildings, and fighter jets.
In recent years, Unity and Unreal have made inroads into non-gaming categories, including engineering, filmmaking, and computer-aided design. In 2019, as discussed earlier, the Hong Kong International Airport used Unity to build a “digital twin” that could be connected to myriad sensors and cameras throughout the airport to track and evaluate passenger flows, maintenance, and more—all in real time. The use of “game engines” to power such simulations does make it easier to produce a Metaverse which spans both the physical and virtual planes of existence. However, the success of the Hong Kong airport endeavor and other simulations like it means more competition, as Autodesk, Dassault, and others respond by adding their own simulation functionality. And just as Unreal and Unity don’t provide all the technology required to build or operate a game, they’re not sufficient in other domains either. Many new software companies are emerging that take the “stock” editions of these engines and “productize” them for civil and industrial architects, engineers, and facility managers, while also adding their own custom code and functionality. One example is Disney’s Industrial Light & Magic (ILM) special effects division. Since using Unity to film Disney’s The Lion King (2017) and Unreal for the first season of the TV show The Mandalorian (2019), ILM has developed its own real-time rendering engine, Helios. The fact that even the most avid Star Wars fans failed to notice any impact from the switch from Unreal to Helios for The Mandalorian’s second season further suggests how many different rendering solutions and platforms will be built in the years to come.
As measured by the number of assets created, the fastest-growing category of virtual software may be those that scan the real world. Matterport, for example, is a multi-billion-dollar platform company whose software converts scans from devices such as iPhones to produce rich 3D models of building interiors. Today, the company’s software is primarily used by property owners to create vivid and navigable replicas of their real estate on sites such as Zillow, Redfin, or Compass, affording would-be renters, as well as construction professionals and other services providers, a better way to understand the space than allowed by blueprints, photographs, or even live tours. Soon we might use such scans to determine the placement of a wireless router or plant, test out a selection of different lamps (each one purchasable through Matterport), or to operate our entire smart home, including electricity, security, HVAC, and more.
Another example is Planet Labs, which scans nearly the entire earth via satellite each day and across eight spectra bands, capturing not just high-resolution imagery, but details including heat, biomass, and haze. The company’s goal is to make the entire planet, in all its nuances, legible to software and to update its data on a daily to hourly basis.
Given the pace of change, level of technical difficulty, and the diversity of potential applications, it’s likely that we will end up with dozens of popular virtual worlds and virtual world platforms, with many more underlying technology providers. This is a good thing, to my mind. We should not want a single virtual world platform or engine operating the entire Metaverse.
Recall Tim Sweeney’s warning about the scope of the Metaverse: “This Metaverse is going to be far more pervasive and powerful than anything else. If one central company gains control of this, they will become more powerful than any government and be a God on Earth.” It’s easy to find such a statement hyperbolic, and it may be. Yet we already worry about how the big five technology companies—Google, Apple, Microsoft, Amazon, and Facebook, each one valued in the trillions—manage our digital lives, influencing how we think, what we buy, and more. And right now, most of our lives are still offline. While hundreds of millions of people today are hired through the internet, and work using their iPhones, they don’t literally perform their work inside iOS or by building iOS content. When your daughter attends school via Zoom, she accesses Zoom and her school through her iPad or Mac, but the school isn’t operated inside the iOS platform. In the West, e-commerce’s share of addressable retail spend now hovers between 20% and 30%, but most of this spend is for physical goods, and retail is just 6% of the economy. What happens when we shift to the Metaverse? What happens when a corporation operates the physics, real estate, customs policies, currency, and government of a second plane of human existence? Sweeney’s warning starts to sound less hyperbolic.
From a purely technological perspective, we shouldn’t want the evolution of the Metaverse to be tied to the investments and beliefs of a single platform. The company Sweeney is imagining would surely prioritize its control over the Metaverse rather than what’s best for its economies, developers, or users. It would surely maximize for its share of the profits, too.
But if we don’t have a single Metaverse platform or operator—and if we don’t want one, either—then we need to find a way to interoperate between them. Here we return, once again, to trees. As you’ll see, I wasn’t kidding when I said the existence of a virtual tree is harder to ascertain than that of a real one.
* The tree may itself be code that brings together many smaller virtual objects, such as leaves, trunks, branches, and bark.
† As you’ll recall from the discussion of GPUs and CPUs, the fact that Unreal or Unity is compatible with most gaming platforms does not necessarily mean that a given experience can run on them.
‡ Epic Games’ history with Fortnite is a good example of this concern. As the highest-grossing game in the world from 2017 through 2020, Fortnite has obviously cannibalized players, player hours, and player spend from other games—some of which are made by publishers other than Epic but which used Epic’s Unreal Engine. In addition, the version of Fortnite which is so popular today—its “battle royale”—was not the original version of the game. When the title launched in July 2017, it was a cooperative survival game in which players worked to defeat zombie hordes. It wasn’t until September 2017 that Epic added its battle royale mode, which closely resembled that used by the hit game PUBG, which, notably, licensed the Unreal Engine. The publisher behind PUBG subsequently sued Epic for copyright infringement, though the suit was later dropped (it’s unclear if a settlement was made). In 2020, Epic launched its own publishing arm to release games made by independent studios, thereby placing the company in even greater competition with some of the publishers which occasionally licensed Unreal.
§ There is some flexibility here—and most analysts expect this payout ratio to go up over time. More on this topic in Chapter 10.