Razer’s Project AVA is what happens when the “AI assistant” stops being a tab in your browser and literally moves into a tiny holographic capsule on your desk. It is part gaming coach, part productivity buddy, part “anime waifu in a jar,” and very deliberately aimed at making AI feel like a character you live with rather than a tool you occasionally open.​

Project AVA started life at CES 2025 as something much drier: an esports AI coach that lived on your screen, mostly represented by code diagrams and gameplay overlays. In 2026, Razer has turned that backend into a physical object—a small cylinder with a 5–5.5‑inch animated “hologram” inside, sitting next to your monitor and watching both you and your display. The capsule has a camera on top, far‑field microphones, and a USB link to your PC so AVA can see your screen, hear your voice, and constantly build a model of what you are doing, from tweaking a Battlefield loadout to poking at a spreadsheet at 2 am.​

The hook is personality. Instead of a faceless voice, AVA appears as a 3D character rendered in real time: Razer‑exclusive avatars like Kira, a very on‑brand anime‑style companion, Zane, a more classic “gaming wingman,” and more office‑ready looks for people who want something less cosplay and more corporate. Razer says these avatars track your eyes, mimic facial expressions, and lip‑sync as they talk, creating a sense that the character is actually looking back at you rather than just looping canned animations. During CES demos, AVA sat in its glass tube, glowing Razer green, commenting on weapon choices, suggesting gadgets, and offering the kind of backseat‑gamer advice you usually get for free on Discord—only this time, the backseat gamer lives on your desk and never logs off.​

Under the flair is a fairly serious AI stack. Razer pitches AVA as “AI agnostic,” with current demos running on xAI’s Grok but with plans to support other large models later. The companion uses a learning engine that remembers preferences and routines—what you play, when you work, how you like your loadouts or even your outfits—and then surfaces suggestions in context, whether that’s recommending frag grenades over stun grenades or nudging you about a meeting you are about to miss. With PC Vision Mode, AVA can actually read the screen in front of you, so instead of generic chatbot answers, you get advice tied directly to what is open right now: strategy tips during a match, formula help in a sheet, or live translation while you hover over another language.​

Razer’s marketing is keen to stress that AVA is “more than gaming.” In the company’s own scenarios, it becomes a desk generalist: handling schedules, brainstorming article ideas, analyzing data, suggesting dinner plans, tracking your mood, and even giving wardrobe tips before you head out. That blend of gaming culture and lifestyle assistant is what nudges AVA into a more personal space—less like asking a smart speaker for the weather, more like talking to a character that has slowly learned your tells, your deadlines, and your bad habits.​

At CES, that intimacy comes with some obvious trade‑offs. To be genuinely “context aware,” AVA’s camera and mics need to be on; it is, by design, a device that watches your every move so it can coach, consult, and cajole. In noisy show‑floor demos, voice recognition was hit‑and‑miss, and some observers raised eyebrows at just how close this feels to putting a gamified, brand‑skinned version of HAL 9000 on your desk, complete with mood tracking and a memory of everything you ask it. Razer is positioning that watchfulness as a feature—“full contextual awareness” and “human‑like vision”—but for privacy‑conscious users, AVA will inevitably prompt questions that go beyond “will this help my K/D ratio?”​

What makes Project AVA interesting, though, is how clearly it captures the 2026 AI zeitgeist. AI companions are suddenly everywhere, from text‑only chatbots to voice‑first apps and deeply customizable “friends” that live in your phone, and Razer is effectively arguing that the next step is to give them a body—even if that body is five‑and‑a‑half inches tall and lives under tempered glass. Preorders require a small fee just to reserve a slot, and the device is still positioned as a concept‑turned‑early product, but the direction of travel is obvious: AI that doesn’t just respond when called, but sits in your peripheral vision, quietly learning until you forget what your desk looked like without it.