Why GeForce Now keeps winning — and what could change that

GeForce Now has been our top pick for cloud gaming for three years running. That's not a marketing statement — it's the result of running the same lab benchmark across all five major services on the same hardware, on the same network, twice a year. GFN keeps coming out on top, and the margin is widening rather than shrinking.

This piece is about why, and about the three specific things that could change that trajectory.

Every other major cloud gaming service buys GPUs from NVIDIA and runs them. NVIDIA runs its own. That sounds like a minor supply-chain detail until you realise that it means GeForce Now ships new GPU architectures at the same time as the consumer cards do, while every competitor is on a 1–2 year lag.

The RTX 4080-class rigs that Ultimate runs on shipped two months after the consumer RTX 4080. Xbox Cloud Gaming's underlying hardware is two GPU generations behind that. Boosteroid is three. The performance gap compounds.

NVIDIA also makes the NVENC hardware encoder that does the streaming compression server-side, and ships GPU-accelerated decoders on the consumer client end. They co-design these.

The practical effect is the lowest end-to-end input latency in the industry — about 23 ms of overhead on top of network ping in our lab. Competitors run on either older NVENC silicon (Boosteroid, Luna), AMD encoders (Xbox Cloud's recent rigs), or proprietary Sony encoders (PS Plus Premium). All of them measure 8–20 ms slower than GeForce Now in our tests.

8–20 ms doesn't sound like much. It's the difference between cloud-streamed competitive Counter-Strike feeling viable and feeling cursed.

Our region latency lab piece documented this in detail: GeForce Now wins 18 of 24 cities we measured worldwide. NVIDIA has invested in physical data centre presence — Frankfurt, Stockholm, Madrid, Paris, London, Dublin, Warsaw, Tokyo, Seoul, São Paulo, Sydney — in markets that other cloud gaming providers either don't serve or serve via long-haul routing back to a US data centre.

That's a moat. Building a data centre is a 12–18 month capital project. The companies that didn't start in 2022 are still 2 years behind even if they sign a lease today.

We've written elsewhere about why catalog services are a worse long-term bet for serious players. The flip side of that argument is that GeForce Now's model is structurally easier to scale: NVIDIA doesn't need to negotiate every individual publishing deal as a revenue split, they just need to convince publishers to allow their existing Steam/Epic/Xbox PC versions to be streamed.

That keeps GFN's marketing budget free of the 'we just signed Bethesda' wars that consume Game Pass and Premium.

Three plausible scenarios where GeForce Now's lead shrinks meaningfully over the next 18 months:

1. AMD ships hardware AV1 encoders at parity with NVENC. If AMD's MI-class GPUs reach NVENC's encode latency and image quality, AMD-based cloud providers (mainly Xbox Cloud Gaming on its newer rigs) can close the latency gap. There's chatter that the next generation already does. We'll know by H2 2026.

2. Apple changes its mind about native cloud gaming clients on iOS. The whole industry has standardised on Safari PWAs because of Apple's policy. If Apple lets cloud gaming apps ship natively, the playing field on iPhone resets. GeForce Now's PWA polish stops being a moat. We don't expect this in 2026, but the European DMA pressure makes it possible.

3. A major publisher pulls out of bring-your-own deals. If, say, Activision/Microsoft revokes GeForce Now support for Call of Duty (something it has tried before), the structural-advantage argument weakens. Cloud streaming rights are still contested legally — a single bad ruling could fragment the market into 'NVIDIA hardware, but only the games on the list' rather than 'NVIDIA hardware, your whole library.'

Realistically? A vertically-integrated competitor with similar scale that isn't NVIDIA. The candidates are limited. AWS has the data centre footprint but Amazon Luna keeps under-investing in the gaming product. Microsoft has the catalogue and the brand but is structurally tied to the Xbox console business model. Google would be the obvious challenger and they already tried with Stadia.

The conclusion most likely to be true a year from now is: GeForce Now is still on top, but Xbox Cloud Gaming is closer on latency than it is today, and the gap on quality is smaller than the gap on infrastructure. Bet accordingly.