Cloud Gaming Bills by the Hour. Your Casual Catalogue Doesn't.

Cloud gaming bills by the hour and its cost scales with time played, which is why it cannot replace the casual HTML5 catalogue on a telecom or web games portal. Both can be worth buying. They are not substitutes, and they should not come out of the same budget line.

Every operator with a content budget has sat through the cloud gaming pitch by now. It arrives with a market forecast, a latency map, and an implicit suggestion: casual games were the thing you ran while waiting for bandwidth to catch up, and now it has.

The bandwidth did catch up. The economics did not move in the direction the pitch implies. And the clearest way to see that is to start with the numbers in the deck, because they are the weakest part of it.

📊 Two Firms Sized This Market This Year and Landed Four Times Apart

Mordor Intelligence puts the 2026 cloud gaming market at USD 6.23 billion, reaching USD 21.62 billion by 2031 at a 28.25% CAGR. Fortune Business Insights puts 2026 at USD 23.79 billion, on the way to USD 159.26 billion by 2034 at 26.8%.

Same year. Same named category. A gap of roughly 3.8x.

They do not agree on where the market is, either. Mordor gives Asia Pacific 38.45% of 2025 revenue. Fortune gives North America 43% for the same year. Those two statements cannot both describe the same thing.

This is not a reason to write the category off. It is a reason to refuse the market-size slide as a decision input. Nobody agrees what counts: whether "PC in the cloud" subscriptions belong in the total, whether game spend that happens to flow through a streamed session belongs, whether an operator's bundle revenue is counted once or twice. When the definitional question is unresolved, the headline total is decoration.

The numbers you can actually act on are your own — activation rate on the bundle, cost per activated user, and what happens to that user's churn. Nothing on the slide substitutes for those.

📉 The Forecast Your Vendor's Deck Inherited Is Five Years Old

In September 2021, Omdia forecast that cloud gaming would "nudge $12bn by 2026". The category generated $3.7 billion in 2021 at the time of writing, and its share of total consumer games spend was expected to nearly triple, from 2.2% in 2021 to 6.1% by 2026.

Five years later, the lower of the two 2026 estimates above is roughly half that forecast and the higher one is nearly double it. Meanwhile one of the services that defined the category when the forecast was written — Google Stadia — shut down in January 2023.

None of that makes Omdia wrong to have published a forecast. It makes a five-year cloud gaming projection a genre rather than evidence. If a vendor's business case rests on one, ask what happens to the case if the number lands at half.

⏱️ The Playtime Cap Tells You More Than Any Forecast

Here is a fact worth more than every projection in this post. Per NVIDIA's own GeForce NOW FAQ, Performance and Ultimate members now get 100 hours of monthly playtime, with up to 15 unused hours rolling over, and additional hours available to buy. Members grandfathered into unlimited playtime keep it only until their first billing date after 15 January 2026.

Read the business model, not the policy. The clear leader in "PC in the cloud" introduced a meter because the underlying cost is a meter: a GPU in a data centre, allocated to one player, for the length of the session. The tenth hour costs roughly what the first one did.

Now put a web portal next to that. A licensed HTML5 game is a bundle of static files. It reaches a CDN edge once, gets cached, and then executes on hardware the player already bought. Your marginal cost for the tenth hour someone spends in a match-3 title is effectively nothing.

This is not an argument about which product is better. It is an argument about which business can tolerate a heavy user. On a portal, your most-retained player is your best asset — more sessions, more ad impressions, more reason to renew a subscription. On an hour-metered streaming service, the heaviest user is the worst-margin customer. That asymmetry is why the cap exists, and it is not going away.

📶 45 Mbps Sustained Versus a Few Megabytes Once

NVIDIA publishes the connection requirements: 15 Mbps for 720p at up to 60 fps, 25 Mbps for 1080p60, 35 Mbps for QHD at 120 fps, 45 Mbps for 4K at 120 fps — and less than 80 ms of network latency to an NVIDIA data centre. Sustained, for the whole session.

Three consequences get under-priced in almost every operator business case:

• It is a concurrency problem, not a traffic problem. Portal bandwidth scales with plays started. Streaming bandwidth scales with sessions held open, times their length. Those forecast very differently.
• The latency figure is a map. "Under 80 ms to a data centre" draws a hard boundary, and a large share of the addressable base for carrier portals across MENA, South Asia, Southeast Asia and Africa falls outside it. A cached HTML5 bundle needs a tolerable connection once, then stops caring.
• It collides with the data plan. Streamed gaming is about the most efficient way to consume a capped mobile allowance. If your bundle exists to make a tariff feel generous, an hour-hungry product does the opposite unless it is explicitly zero-rated.

For operator game portals in particular, that third point decides the whole thing. Bundling a service that eats the customer's own allowance is a support ticket generator, not a loyalty play.

🔌 Amazon Just Demonstrated the Platform Risk

In April 2026, Game Developer reported that Amazon had ended individual game purchases on Luna, with titles bought directly on the platform ceasing to work on 10 June 2026. Access to external stores — EA, GOG and Ubisoft — was removed, along with third-party subscriptions including Ubisoft+ and Jackbox Games. Affected users were offered a complimentary Luna Premium subscription as a thank-you rather than refunds.

Set aside the consumer question. The operator lesson is structural. When you bundle somebody else's streaming service, you do not control the catalogue, the entitlements, or the commercial terms. A retail partner can find, at a quarter's notice, that a meaningful slice of what they marketed no longer exists — and that the remedy offered to their customers was decided by someone else.

A licensed catalogue is not risk-free either. It has a term, a scope and an expiry, and those matter. The difference is that you negotiated them and they sit in a document that nobody rewrites mid-quarter. If your portal is the product, that distinction is the whole business.

🎮 Where Cloud Gaming Genuinely Earns Its Place

The category is not a science project. Xbox Cloud Gaming left beta and is now offered across Xbox Game Pass Essential, Premium and Ultimate, per Xbox Wire in December 2025. That is a maturing consumer product with real distribution behind it.

It is the right tool in specific situations:

• Console-class demand without console hardware. A household that wants a particular big-budget title and has fibre but no box is the textbook case.
• Network differentiation. If you are selling fibre or 5G and the network is the product, a streaming service is a demonstration of it. That is a marketing budget, honestly named.
• Removing the download. Trials and demos of very large games work better streamed than as a 90 GB install.
• Screens with no usable GPU and a good fixed line. Smart TVs and set-top boxes on wired broadband are a reasonable fit.

What it does not do is fill four idle minutes on a prepaid handset.

🎯 The Two Products Answer Different Questions

Cloud gaming answers: "I want to play this specific expensive game and I don't own the hardware." High intent, planned session, willing to sign in, tolerate a launcher, and find a controller.

A casual catalogue answers: "I have four minutes and a phone." No intent toward any particular title. That session does not survive an install, an account creation, or a visible loading bar.

The difference propagates through every number that matters:

• Acquisition. Cloud requires a subscription decision before the first frame renders. A web game requires a tap.
• Session shape. Long, scheduled and controller-driven, versus short, incidental and one-thumbed.
• Failure mode. Cloud fails loudly — stutter, disconnection, input lag — and on an operator bundle the blame lands on your network. Casual fails quietly; the player just leaves.
• Addressable base. Cloud reaches the subset of your customers inside the latency and bandwidth envelope. Casual reaches anything with a browser, including the four-year-old Android handset that still makes up most of the install base in high-growth markets.

If the KPI you are judged on is daily actives on a portal, minutes inside an app, or churn on a low-ARPU prepaid base, the second column is the one that moves.

🚫 Five Ways Operators Get This Comparison Wrong

1. Funding the cloud bundle out of the content budget. They do different jobs. Cloud is network differentiation; the catalogue is daily engagement. Merging the lines starves the one that produces sessions every day in favour of the one that produces a press release.
2. Quoting market size as if it were addressable demand. When two published 2026 figures sit 3.8x apart and disagree on the leading region, neither belongs in a business case as a growth assumption.
3. Assuming the bundle markets itself. An entitlement sitting unredeemed inside a tariff is pure cost. Measure activation and second-month usage, not attach rate — attach rate counts people who never pressed play.
4. Ignoring the data-plan collision. Selling an hour-hungry service into a capped plan without zero-rating produces bill shock, not loyalty.
5. Calling it a games strategy. It is one shelf, aimed at a minority of the base, stocked with titles casual players are not asking for. It does not replace a catalogue and should not be presented internally as though it does.

🎮 Where a Licensed Catalogue Fits

Forestry Games has licensed games since 2017 and carries a catalogue of 1,049 titles spanning HTML5 and Android APK games, with titles also published on Google Play and the App Store. The company develops HTML5 games in-house, works with branded IP, and holds brand partnerships including Disney, Nickelodeon, Cartoon Network and Warner Bros.

For operators weighing this decision, the practical relevance is the cost shape described above: a licensed catalogue is bought once against defined scope, then served from your own edge with no per-hour meter behind it. That works whether you are running a white-label portal, a carrier VAS bundle, or a games section inside an existing app. The catalogue is browsable if you want to size the shelf before talking terms.

🧭 What to Ask Before You Sign a Cloud Gaming Bundle

None of this argues against buying cloud gaming. It argues against buying it on the strength of the forecast slide. Six questions worth putting to the vendor before the commercial conversation:

1. What share of our subscriber base sits inside the 80 ms latency envelope? Ask for the map before the market size.
2. Who pays for the data, and is zero-rating written into the agreement or assumed?
3. What happens to customer entitlements if you restructure the offer? Amazon's June 2026 Luna cut-off is the precedent to name in the room.
4. What activation target is this priced against, and what happens at renewal if we miss it?
5. What is the cost per activated user per month, placed directly beside the cost per monthly active user on our existing portal?
6. Which metric is this supposed to move — churn, ARPU or engagement — and what evidence exists that it moved that metric anywhere else?

If the answers hold up, buy it, and run it beside the casual catalogue rather than instead of it. If they do not, you have just saved the budget that was going to fund the thing your daily active users actually open. The clearest signal in this entire market is the one NVIDIA published in its own FAQ: when the cost of serving a player scales with the hours they play, somebody eventually puts a cap on the hours. Build the part of your portal where that never has to happen.