The Steam Deck quietly ate the handheld cloud category

Around the Stadia launch, the cloud-handheld pitch was clean. A 6-inch device with a great screen, no internal GPU, and a thin client. Battery would be great because there's no rendering happening locally. The library would be your full Steam or Game Pass catalogue. Storage requirements would be zero.

Logitech G Cloud and the Razer Edge both shipped products on this thesis. Both struggled. The Logitech G Cloud sold modestly, the Razer Edge effectively didn't.

The Steam Deck is a 7-inch handheld with a real GPU, 16 GB of RAM, and 512 GB–2 TB of storage. It runs games locally. It supports cloud gaming as a fallback (Steam Link, GeForce Now via Chrome). But it's not designed around cloud, and the local-rendering experience is the primary one.

Steam Deck sales since launch have been north of 6 million units. The OLED refresh and Steam Deck 2 (announced 2025) have only accelerated demand. The category Valve targeted — portable PC gaming on a curated catalogue — exists at scale.

Three reasons, in order of impact.

First: handheld gaming sessions happen in places with bad WiFi. The bus, the car, the cabin, the park. A local-rendering handheld works everywhere; a cloud-rendering handheld works only where there's a strong network. The first failure mode in your second session sours the whole product.

Second: handheld battery life favours efficient local rendering over cloud streaming. A modern AMD APU at low TDP draws roughly 6–8 W under load. A high-bitrate cloud streaming session also draws 6–8 W (because the WiFi radio is on, the screen is bright, and the video decoder is busy). The battery savings of cloud rendering on a handheld is roughly zero versus an efficient APU.

Third: handheld players run a lot of indies. Hades, Vampire Survivors, Slay the Spire, Balatro. These games run beautifully on a Steam Deck at 60+ fps and are not on any major cloud service catalogue. The library you'd actually want on a handheld is mostly the library Valve has and cloud doesn't.

For AAA single-player on the road. If you're on a 5G hotspot at a hotel and you want to play Starfield at high settings for an hour, cloud-on-handheld is genuinely good. The Logitech G Cloud and Razer Edge both do this fine.

But this is a niche of a niche. The frequency with which a real handheld user wants to play a current-gen AAA title in an environment with both good network and an hour of free time is much smaller than the frequency with which they want to play a 2023 indie with no network.

The handheld category was supposed to be cloud's strongest play. Small device, no local hardware tax, mobile audience. It turned out to be cloud's weakest play because the failure modes — network outages, battery drain, library gaps — all hit harder on a handheld than on a TV or desktop.

If cloud can't win the handheld category, the 'cloud beats local hardware everywhere' thesis is weaker than its proponents thought. The right framing is that cloud wins in specific deployment scenarios (TV in a living room with wired ethernet, second computer in a study, casual play on a tablet) and loses in others (handheld with variable network, competitive gaming, modded games).

Steam Deck 2 ships in late 2026 on a refreshed APU. It will likely add hardware AV1 decode, which makes it the best handheld cloud gaming device in addition to the best local one. The cloud services that adapt to this (a Steam Deck-optimised GeForce Now client, for instance) will see a usage bump.

But this doesn't reverse the category narrative. The Steam Deck is a local-first handheld with cloud-as-fallback. The cloud-first handheld category is over, and the products that bet on it have lost.