13-12-2025 Cloud Gaming

Wait, you can game on a mac now?

Cloud gaming is incredible. As both a delivery model and business model. Also, as internet speeds increase – what else is possible?

This is a bit of a thought-dump. I found cloud gaming interesting from quite a few angles.

my context

Recently, I bought a gaming PC on Black Friday and refunded it. Despite being a heavy mac/linux user, the motivation was as an investment into my relationships. Friends I treasure are scattered among different cities, and gaming as an activity is a way of spending time with them before we all get too busy in life.

The PC cost $1.6k AUD(~1.1k USD), and even if I only gamed 2 hrs a month – I’d consider it a worthwhile investment.

It was my first time getting a PC and there’s a depth of optionality to digest. CPU, GPU, RAM, SSD, power supply rating, ports, wifi, cooling capabilities, case…etc.

As I delved further in, I reencountered cloud gaming – something I came across previously, but couldn’t fathom with slower internet back then.

Cloud Gaming: a delivery model where you stream games (the output) from a more powerful and compatible computer to your device, which is where your inputs are sent from.

While I weighed multiple configurations and options for my PC, including the target longevity and budget for my predicted usage pattern – cloud gaming just seemed to simplify the decisions required and offered greater flexibility. I didn’t need to have a depreciating unused asset on months I don’t game, I could pay for what I needed.

Nvidia’s GeForce Now (GFN), is a cloud gaming service in Australia via a local provider. A few things made the prospect of it more sensible in my mind:

• Confidence – access and experience with one of the fastest internet in Australia: 1 Gbps (that’s fast here), experiencing that changed what I thought was possible.
• Familiarity – recent experience going to a gaming lounge with friends: a pricey $20/hour.
  • But I got to play the small amount desired without sinking capital into a $500 console.

After trialing GFN, to a satisfactory experience – I realised it was a simpler solution at the moment. It would take a minimum of 4 years on the most expensive GFN plan on a month-by-month basis to reach the price of buying a PC. With reselling PC at around 33% value, then 2.5 years. Make no mistake, buying a PC is still better economically with frequent use/multi-use purposes. However, for my single and infrequent gaming purposes – cloud gaming makes more sense.

Additionally, I knew I wouldn’t be subscribed continuously – which would reduce the overall expenditure and more importantly the opportunity cost.

Applying my junk food philosophy to cloud gaming

My ice-cream/junk food philosophy is to eat the amount I initially wanted and often throw away the remaining, rather than leave it as an easily accessible temptation in the kitchen. Sounds crazy, even heretical – but it has undoubtedly made me a healthier and happier person.

I realised I can apply a similar method with cloud gaming. While long-term subscriptions of bi-annually are cheaper than month-by-month, I am going to opt for month-by-month. Essentially, when my friends and I want to play in a less busy period, I’d pay ~$25-33 to be able to play for that initial session and any others we might have that month. We might not play any more than the first session, and that’s fine too. The point is two-fold:

• Friction is introduced: Gaming can be a vicious-cycle and a massive mental/energy/time-sink with the way it hits your dopamine. By adding friction in that I’d need to repurchase a new month when that month finishes, I reduce the ease and accessibility of personally gaming more than I’d actually want to. My time and energy saved from doing so, easily surpasses the financial discount from a longer-term subscription. There are many other things I want to do.

• More closely match my intended usage pattern: There will be gaps in-between the single-month cloud gaming subscriptions based on when we play. E.g. in weeks: W, W, W, W, 0, W, W, W, W, 0 , 0 , W, W, W, W Meaning, it might not actually be overall more expensive as I pay more closely to my usage pattern.

Why do I not do this for my gym membership? Besides the difficulty in even doing so, that’s because I want to encourage and enable the other end of behaviour.

Cloud Gaming - thoughts that sparked

Alright, here’s some interesting thoughts that sparked.

The Delivery Model

• Cloud Gaming is essentially a PC cafe at scale

  • PC cafes and cloud gaming both give users the value of gaming on a PC they don’t own

  • PC cafes are complicated and less-scalable. You need physical assets beyond computers (screens, tables, chairs, lights…etc), maintenance, frontline staff, venue…etc. Your expenses are greater to provide that service directly to users on-site. The blueprint is not 1-1 applied whenever you expand to a new location.

  • Cloud gaming removes a lot of the complexity in a PC cafe model. It simplifies to primarily a hardware (compute) and software (delivery method) problem.

  • Interestingly there’s still a common physical constraint in that there’s a limited amount of users that can be serviced at once by the GPUs available.

    • Unlike other digital products/services which scale essentially infinitely to meet demand, cloud gaming cannot simply multiply the expensive GPU’s required to run games instantaneously (without some clever engineering) to meet demand.
    • When I worked on scaling AI generated video conversations, there was a budget constraint with Nvidia B200’s – which one provider offered on a monthly commitment basis of ~$2,500 USD/month for a single B200 GPU. Meaning we had to forecast expected our user usage across different timezones and optimise how many conversations we could run on each B200. Due to this steep upfront cost for GPUs, we cannot over-provision too much and we cannot easily scale on-demand. Especially, if we own the assets themselves.
    • Feature-wise, that usually results in a queue mechanism of some sort. For GFN, if too many users try to spin up a session at once, you enter a queue. And there is a session max length of 6-8hrs, assuring the queue will move.

• Internet speed

  • Users need a fast and stable internet for a good cloud gaming experience. You’re streaming user controller input and game audio/video output back.
  • Additionally, latency must be low for the nature of gaming. It would suck to pull the trigger and see the bullet fire 2 seconds later on screen.
  • This requirement prevents those without such internet access from cloud gaming.
  • However, internet speeds are increasing globally. So the viability and market potential of cloud gaming is only increasing every year.
    • Interestingly, I found that Singapore has the #1 rank internet globally. With the base home internet offering 3 Gbps download speed. I wonder if its tied to their human capital and human resource cultivation. It definitely is an enabler for its people considering how digital the world is.
  • I am reminded of someone who said that educational/opportunity uplift in developing nations could be achieved by providing everyone with a cheap smartphone/laptop and fast internet access.
  • For me, I have the option now of not having to own a big pricey PC rig to play games. I.e. I do not need to own the consumer GPU.

• In the future,

  • What other usecases are possible with fast internet? What else could be delivered?
    • the first one that came to mind was VR experiences/gaming.
      • you no longer need a $4000AppleVisionPro,justa$200 headset. Maybe even Google cardboard?
  • Will the market of consumer GPUs shrink? How about compute and storage…etc?
  • Would all we need be cheap laptops?
  • Fundamentally, I think people still like to own stuff (as they should). But there exist some price point and value proposition where it starts to make more sense for some people
    • E.g. why buy a $350 2TB physical SSD to hold your photos when Google Photos costs 15/month and makes your photos accessible everywhere? Or why not get both?

The Business Model

Congrats, you managed to run a massive PC cafe on the cloud.

• A unique dilemma – it’s actually less like a PC cafe and more like a gym membership

  • Similar to a PC cafe, you lose money if no one are using your ‘gaming PCs’. They are depreciating assets you bought upfront.
  • So you want to fill up all your possible seats always.
  • Unfortunately, unlike a physical PC cafe – where people who walk in, might see the place is full, and walk out with a slight :(
    • A digital PC cafe, where you are selling membership like a gym, means that – users walk in, see the place is full, and expect to be able to play ASAP with a growing :(
  • This resulted in a unique pricing strategy, where there is:
    • a maximum amount of hours a user can play in a month
    • a queue where users line up to access an available gaming ‘seat’
    • a maximum session length a user can play continuously, before having to give up their seat and re-queue if desired
  • Unlike a PC cafe, you don’t actually need people to ‘use’ your ‘gaming PCs’ to make money. You can instead, run it like a gym where people pay for access rather than usage.
    • You end up balancing trying to get as many membership sign-ups as possible, while also planning/building/provision the infrastructure to support it.
      • think of your gym adding a floor to the building every time they get 100 more members
      • what ended up happening with GFN is that they removed the ‘Basic’ pricing tier as it ‘sold out’ (the number of people those GPUs could support have been reached)

• Behaviourally, cloud gaming become more profitable per user as time goes on

  • For GFN, each user has a maximum total time of 100 hours per month
    • that sounds differently to a person depending on what stage of life they are at,
      • a 15 y.o. high schooler might consider that too little, with their plentiful free time
      • a 21 y.o. uni student might consider that generous, with their other shifting priorities
      • a 35 y.o. with kids would consider that way more than needed
    • Basically, as the user ages – their actual realised value obtained from the service decreases while the revenue remains constant. Resulting in increasing profit-margin.
      • If a GPU can provide 100 hours of gaming a month, and all users used up their allocated amount – then 1 GPU per user is needed for adequate servicing.
      • If all users used up only half their allocation – then roughly 2 GPUs per user is needed. And so on.
  • This is extremely similar to gym memberships, where many members might not even go to the gym for weeks and months – yet, continue to pay.
    • These users continue to pay sometimes for behavioural reasons and sometimes for optionality (which is value in itself too).
    • Note, as users age and utilise the service less – they generally have increasing disposable income.
      • Hence, their realised value to %disposable income ratio might still remain stable – which I suspect is partially why they might continue to stick to the product.