The Future of Game Engines & Infrastructure


Overview of the space & predictions for where it’s headed

“Video games now write the first draft of history…For the past 30 years, the common wisdom has been that games are a land unto themselves. What happens there is only applicable to games, and nowhere else.” — Facebook’s Oculus sale proves that videogames & the game engines that underly them now write the first draft of history

As we’ve seen in the past decade, gaming is becoming impossible to ignore. Games now seep further than ever into social experiences, fitness, art, and education. Many of the core technological developments around game engines & architecture are being carried into industries from fashion & film (3D garment simulation technology for fashion and film/gaming), to architecture (rendering visualizations), and the automotive industry (rendering showroom graphics). Unreal even refers to itself as “the 3-D operating system for all”. Games that used to require a specialized console and controller with specialized graphics processing are now often accessible with a tap or two in the app store on our phones that are more powerful than prior top tier consoles just years ago.

While many have discussed the application layer of games, we believe the core infrastructure is often understudied when trying to understand both the future of gaming and the many other fields gaming will inevitably expand into. The evolution of game engine infrastructure has grown in two primary directions: towards light-weight & addicting games, and towards exceedingly deep gaming experiences — which branch into games created by big studios with deep pockets and indie games (much like the division of big hits in the film industry).To contextualize the three core areas of innovation around game engines that I think are driving change (machine learning, monetization, and cloud gaming) — I’ll connect their new applications to the ways in which they expand either the addictiveness of depth of modern games.

Cloud Games Creates more Opportunities for Continuous Gaming

The spread of cloud gaming enables greater scale and complexity of the average game, as well as more continuous play for addicting games. Cloud gaming represents a new age of game potential, far less constrained by hardware. Local hardware constraints used to mean game creation was intensely limited by the power of the devices gamers owned and compute used to happen locally — now the significantly more compute in the cloud.

The popularity of cloud gaming has made it easier to decentralize these aspects of the game. Cloud gaming also enabled cross play, which in turn opened up the total addressable market for many to play with their friends. This reduced the power consoles have within the market (hence why many studios aren’t as reliant on consoles anymore, especially when Fortnite demanded cross platform player-matching from Playstation and Xbox). This is evidenced by companies like Parsec Gaming (60 FPS interactive streaming for gaming), Bethesda Softwork’s Orion (which is a game & software agnostic software that reduces latency and increases bandwidth), and many more outlined within this overview graph.

Historically, the cost of on prem server purchase, maintenance, projections, and capacity planning for a client-server model was prohibitive to everyone but large game studios. But over time, costs have been driven down and any developer/publisher can request resources on demand. As the depth of gaming increased, many companies realized in the early 2000s that clusters of commodity hardware that service these needs were both cheaper and more scalable. As gaming shifted towards digital distribution (drastically lowering marginal cost to the studios), more games became free to play, changing the economic model. This lowered the barrier to entry for gaming, with a growing set of initially free-to-play games which later introduced locked levels, powerups, or other paid privileges once players were hooked. The lowered maintenance cost creates a space for gaming as a service: on demand, continual game play, with the added bonus of a continual revenue stream.

As more and more players are online at any given time, we’ve seen added layers of more ambient gameplay — gaming as a space, rather than a linear experience. It’s notable that this allows for the extension of the social world that can be emulated within games — the Travis Scott concert hosted within Fortnite for example, hosted over 12.3 million audience members at once in the game. Massively multiplayer games like festivals and events are part of the expanding norm that cloud gaming enables. In this metaverse, gameplay can be picked up from one device, seamlessly continuing from a console to mobile phone.

Pioneers such as Onlive in 2003 (which allowed gameplay on computers, smartphones, and smart TVs, but failed to secure top-tier games on their platform) and Gaiki in 2008 (which demonstrated World of Warcraft running on an iPad, and was acquired by Sony in 2012) saw early success with multi-million user bases.

The growth of cloud gaming (and the resulting change in what is considered a ‘gaming device’) allows for more robust and persistent experiences of a world within a game across devices — console and mobile (notifications, touch, or even text) even smart speakers or mirrors. Although this type of play is in its early stages, there are also notable debuts of mixed success, such as that of Xbox xCloud, where the transfer of gameplay from console to mobile with the same controller and without much redesign appeared to be rather forced (see more about Microsoft’s game streaming service: Project xCloud here). Other major players that have done the same include Playstation, Steam, and slew of other companies.

Cloud compute also enables scaled multiplayer gaming, paving the way for the dramatic growth of esports. Fortnite’s growth is noteworthy in this area, with cloud gaming ultimately enabling 100 players to be live on one shared map, yet for the average game only 10–15 players need to be rendered at once.

I should note that cloud gaming also presents bottlenecks such as unfairness presented by real-time video consumption delays and more broad latency issues. There is certainly potential for increased optimization of each individual player’s virtual machine such that inner-player delay is optimized.

Within this space we’re eager to hear from companies working to optimize mobile performance for gaming as well as platforms that include social features that are currently more optimized for devices with greater screen space (chat, live view of other friends playing, etc). We believe there’s great creative potential for imagining how the future of deeper gaming experiences will creatively continue between all of a player’s devices — creating a more robust and continuous environment that makes use of the unique features of each device.

Monetization is Becoming Increasingly Continuous

Monetization represents the second and arguably most notable area of change within game infrastructure at the moment. As gaming becomes more persistent, monetization becomes increasingly continuous — the barrier to being a part of a game’s culture and community is now lower than ever. The central goal of game creators is largely shifting from getting some critical mass to try a game to keeping players interested and minimizing churn. Incentives have largely shifted from “acquire users” to “retain users”.

With the popularization of free-to-play games, most of the monetization has shifted to in-game purchases. To lock users into persistent games with lower barriers to entry, you need an in-game economy: a system of monetization that the user feels compelled to build wealth and value within (see this nearly exhaustive piece on the ways virtual economies are exploding). Around larger game studios, Fortnite and Apex legends represent popular versions of this model (and Counter Strike & League of Legends represent earlier versions of this), with in-game items such as skins, weapons, cars, and loot boxes dominating in-game economies. Out of Fortnite’s 2.4 billion in revenue, over 1 billion was generated from the sale of in-game items. The shift in game engine architecture around monetization seems to indicate potential for companies like dmarket (where the internal team promises to help game developers outsource the building of their skins-based in game market), Blacknut (a subscription-based cloud gaming platform which allows users to play multiple games on-demand, without buying them separately), Pragma (a set of tools that enable multiplayer and social games with a more horizontal back-end approach) and forte (where the company is centered around building infrastructure for decentralized ingame economies).

The value of a game also generally increases as potential the game as a continuous, persistent social hub increases. Parallel to the idea of the metaverse, and of continuously deeper games, a world created within a game feels much more rich as the potential for a full range of human expression expands. In worlds external to gaming, as the amount of time spent online broadly increases, and search engines for nearly any product are losing consumer trust, more and more purchasing is tied to the reputation of influencers and the trust they build around their whole person. This means that influencers often have the ability to: drive massive attendance from certain generations to in-person and online events, encourage purchases through intentional or passive endorsement via videos of their experience with the merchandise, and broadly direct trust and therefore attention to a game. We’ve also seen this dynamic play out with our investment in Cryptokitties. As gaming shifts to a more ambient experience, I think the presence of an influencer in a game or limited access game space can begin to resemble the gacha game mechanics — where players pay for access to a specific space and consistently test the slot machine to see what type of person they might be rewarded with.

Within this space we’re interested in companies building tools to aid the creation and upkeep of in-game economies, cooperatively owned economies, and in addition to founders with strong new views on what’s essential to a successful and thriving in-game economy. As it relates to self expression and influence, we’re looking for companies with a strong view on how building influencer support and social value around the game as a broader social hub can create stronger social loops around participation in in-game economies.

Machine Learning is Enabling Deeper Games & More Intelligent Tooling

We strongly believe that machine learning (whether that means procedural content generation, computer vision, or other areas of deep learning) like any other creative tool, doesn’t remove power from creators, but enables a new generation of broader creative ideas to exist due to scale and pace of development improvements.

PointFlow: 3D Point Cloud Generation with Continuous Normalizing Flows

We call this paradigm computational creativity, and have previously partnered with companies such as Runway as well as Shadows within this thesis area (see here for a deeper dive on how we’ve experimented with ML and animation at Shadows). We continue to believe ML will proliferate into many large industries, and believe that gaming is likely next.We’re searching for a company that will continue this trend within the gaming space — applying creativity and automation with the intent of giving creators greater scale. From more infrastructure-centric aspects like skill-based matchmaking or AI opponents, to creative pursuits like texture or level generation, machine learning gives creators scale, and we believe this will only become more true as consumers beg for more depth, and developers crave creating more unique experiences. We see creative content generation growth enabled by ML already reflected in the work of Modl.Ai (content generation and AI opponents),

Additionally, it’s easy to see how the growth of ML applied to simulated/virtual environments will have compounding effects for the world outside of gaming. This area of gaming represents a broad opportunity for testing and modeling human behavior that will inevitably be integrated into other fields, To take just one example, Open AI has also invested time into research in the area of agent hide + seek in generated worlds/environments — which reinforces high potential returns at the intersection of emergent agent behavior and intelligence research, even relative to the generally high potential of ML.ML applied to world and infrastructure generation shouldn’t entirely be understood as simple scaling of creative processes, but as the possibility of creating broader engaging worlds and experiences that only the real world previously represented. We’ve seen this already with our investment in AI.Reverie, a synthetic data company who’s simulations were built using Unreal Engine, and in our participation in the broader autonomous vehicle industry where simulations have become a prominent solution to training for edge cases (a team at Princeton even detailed the pros of using GTA V as a testing environment).

Within this space we’re eager to find companies that understand the power of interfaces that push forward the users vision of the possible use cases of ML combined with stereotypical game building tooling (similarly to the way we saw Runway push forward creative use cases for ML) . We believe that when exploring new use cases for ML applied to fields, the assumption is often that the use cases drive beautiful interfaces/platforms, rather than that well designed interfaces on broad platforms can actually push an informed game designer to find new applications to their field (see: the myth of the infrastructure phase).


Many pieces of game engines (server optimization, physics engines, graphics rendering, etc) have been rather silo-ed from other industries until now, and we are at an inflection point where much of the core technology is being re-applied and improved for external use-cases. Even if you’re not interested in gaming itself, there’s a hidden world of brilliant engineering that holds massive potential for everything outside of gaming. Most recently, Unity discussed this at length in their S-1 as a core future value driver to the company, as they’ve heavily penetrated the mobile game market.

What used to be a rather insular community has suddenly seen core game development principles applied to a vast amount of industries. To take just a few examples: personal development apps like Plant Nanny & Fortune City have been gamified. Interactive narrative game engine software is being used to produce interactive film (like Netflix’s interactive Minecraft Story Mode and Twine software for Black Mirror Bandersnatch). 3D rendering capabilities developed for games are being reused to digitally design and render fashion to be produced physically, or to be represented on TV (CLO Software & Marvelous render clothing for Love, Death, and Robots) — even used to render the entire city of Shanghai based off of live data sources.

We came away with a few core assumptions that we’d love to have challenged:

  • Many of the core functions of game engines are being unbundled as a direct result of the massive changes currently occurring in the industry.
  • The robustness of gameplay (again, as depth and specialization increase & barrier to entry decreases) creates robustness of behavioral data that is unparalleled and represents great potential we think will see much greater use in the next 3 years.
  • Machine learning will continue to be applied to various creative processes to give creators greater scale — gaming will be no different.
  • Game infrastructure companies targeting unbundling for middle-sized games who use hyper specialized tooling or monetization as a wedge to challenge monolithic players of unity and unreal are poised for great success.

Finally: we want to talk to companies that challenge these assumptions or have strong views on this type of unbundling!

With over 200 million casual gamers, and rapidly growing — game engine infrastructure is more relevant to the general population than ever. In quarantine especially, inspecting the underlying technology that upholds the growth of gaming for a digital gathering space is essential. Infrastructure surrounding cloud gaming, new payment frameworks, and ML is at the core. In the coming years companies founded in these areas will set the foundation for the next wave of gaming to enter an even broader mainstream than today.”

Thanks to Michael DempseyAaron LewisJeff LindsayNick Nikolov, and of course Santi Ruiz for all of the edits and thoughts!

As always — feel free to tweet or message me questions, thoughts, disagreements, or pitches on twitter or at


Things I’m thinking about external to this piece

  • Game engines are philosophically interesting because building a 3-D operating system for all requires having some philosophy of the best way to render infinite details of the real world, which naturally demands prioritizing certain tools, functionalities, details or representations. It’s interesting to see the way engines get reused for certain industries and how their underlying philosophies about how to reduce the detail of the world to best render it overlap.


  • The very beginning of this piece started as an exercise I completed when interviewing at Compound (Prompt: give an overview of the future of game engines from an investor’s standpoint, within three days)! If you want to see what I originally submitted, you can view it here (it is very rough and many of the details are transcribed from phone calls with people in the gaming space, and are probably error-ridden).
  • At another point in time I massively overestimated my writing capabilities and started to write a massive thesis — there’s about 40 hodge-podge pages and if you’re curious you can also view that here.