The Google Home Speaker went on sale June 25, 2026, six years after the Nest Audio and seven after the Nest Mini — two devices that are now discontinued. Priced at $99.99, the speaker is the first Google audio hardware built from the ground up for the Gemini AI assistant, and it carries an architectural change that distinguishes it from every Google smart speaker that came before: a dedicated Neural Processing Unit that runs Gemini Nano inference locally on the device, eliminating the cloud round-trip that defined all prior Nest speakers. The practical result is response times Google describes as under 100 milliseconds for standard voice commands — and a meaningful shift in the privacy picture for always-on listening devices, at a moment when Google reached a $68 million class action settlement in January 2026 over prior speakers’ recording of private conversations.

The catch, as first-wave reviews published June 24 make clear, is that the most compelling Gemini features are locked behind a $10-per-month subscription. The hardware itself is capable and competitively priced — but buyers who stop paying after the bundled six-month trial expires will find the speaker substantially less interesting than the one Google demoed on stage.

How the On-Device NPU Changes the AI Architecture

Every Google smart speaker before this one sent voice queries to remote data centers for processing. That cloud round-trip introduced latency — typically several hundred milliseconds — and meant the device was completely non-functional without an active internet connection. Google’s previous Nest speakers were cloud endpoints masquerading as smart home hubs: the microphone picked up a wake word, the audio traveled to Google’s servers, the answer traveled back.

The Google Home Speaker breaks that dependency for standard queries. A quad-core ARM Cortex-A55 processor running at 2.0 GHz is paired with a dedicated Neural Processing Unit — a specialized hardware accelerator designed for the matrix multiplications and tensor math that neural network inference requires. The NPU runs Gemini Nano, Google’s compact on-device model, locally. The same NPU handles background noise suppression, reverb cancellation, and acoustic echo cancellation simultaneously, which is the processing that determines whether a smart speaker can reliably pick up a command across a room when music is playing. Google says this architecture produces response times under 100 milliseconds for standard voice commands — compared to the several-hundred-millisecond round-trips of prior Nest speakers.

The privacy implication is concrete: basic voice processing no longer requires data to leave the device. That distinction matters in context. Google agreed in January 2026 to a $68 million settlement of a class action lawsuit, Case No. 4:19-cv-04286 in the Northern District of California, over allegations that Google Assistant on prior Nest and Home devices activated and recorded private conversations without users’ intentional triggering — so-called “False Accepts” — and shared that audio with third-party review vendors. The new speaker’s on-device inference does not send basic voice commands to Google’s servers. Advanced features that do require cloud connectivity — Gemini Live, Camera History Search, and Home Briefs — are separately disclosed and require a Google account and internet connection to function.

A physical hardware switch on the back of the device cuts electrical power to all three far-field microphones. This is a hardware mute, not a software toggle — the microphone cannot be re-enabled by voice command, only by physically flipping the switch back.

What Gemini for Home Can Do Without a Subscription

The base Gemini for Home experience, included without any subscription, handles natural multi-step voice commands, smart home device control, general knowledge questions, music and podcast playback, timers, and calendar access. The significant change from Google Assistant is architectural: rather than matching specific command phrases to predetermined responses, Gemini for Home is a large language model that parses intent from natural speech. Users can correct themselves mid-sentence, chain multiple commands in a single request, and ask follow-up questions without repeating “Hey Google” each time, because the device holds short-term conversational context across a session. Google has also added 10 natural-sounding voices available across all supported languages.

Continued Conversation — which keeps the microphone briefly active after a response to catch follow-up questions — is now available in all supported languages for the first time, not just English.

What Gemini Live and the Subscription Add

Google Home Premium, at $10 per month for the Standard plan, unlocks three features unavailable to free users.

Gemini Live allows open-ended, free-flowing conversations where users can brainstorm, interrupt, and change topics naturally — closer to a conversational AI interface than a command-response assistant. Engadget’s Nathan Ingraham flags this as the most valuable feature and the one most likely to disappoint buyers who skip the subscription.

Camera History Search lets users ask the speaker about activity captured on connected Nest cameras — for example, whether a package arrived, whether the back gate is open, or whether a pet got on the couch during the day.

Home Briefs delivers an AI-generated summary of household activity when users return home or ask what happened while they were away.

Every Google Home Speaker purchased before September 30, 2026 includes six months of Google Home Premium Standard at no additional cost — a $60 value. After the trial, buyers must decide whether the most compelling features justify a recurring $10 monthly charge.

The Engadget review also noted that the ability to build a smart home routine simply by describing it in natural language — without navigating menus in the Google Home app — is gated behind the subscription. Ingraham wrote that locking features like Gemini Live and conversational routine-building behind the paywall is “definitely a bummer.”

Thread 1.3 Makes This a Smart Home Hub, Not Just a Speaker

Beyond its role as a voice assistant, the Google Home Speaker doubles as smart home infrastructure. It ships as a Matter controller and includes a Thread 1.3 border router — two roles that convert it from a peripheral endpoint into the networking backbone of a modern smart home.

Thread is an IPv6-based, low-power mesh networking protocol built on the IEEE 802.15.4 radio standard, designed specifically for battery-operated and low-power IoT devices — door locks, sensors, thermostats, and switches. A Thread border router bridges that mesh network to the broader IP network, which is what allows Thread-enabled devices to communicate with the internet and with apps on a phone. Before a device with a Thread border router is present on a home network, Thread-based smart home devices have no internet connectivity at all. By bundling the border router into a $99.99 speaker, Google makes that infrastructure available at the entry level for the first time.

Matter, developed by the Connectivity Standards Alliance and backed by Apple, Amazon, Google, and Samsung, is an application-layer standard that governs how smart home devices communicate with each other regardless of manufacturer. A Matter controller can manage any certified device across all supported brands — which means the Google Home Speaker, combined with its Thread border router, can serve as the central hub for a genuinely cross-ecosystem smart home, including devices designed for Apple HomeKit and Amazon-certified hardware.

Additional connectivity includes Wi-Fi 6 (802.11ax, dual-band 2.4 GHz and 5 GHz) and Bluetooth 5.4. Two Google Home Speakers can be paired with a Google TV Streamer for a home theater surround configuration.

Audio Hardware: Better Than the Nest Mini, a Step Back From the Nest Audio

Audio performance is the area where reviewers diverged most. The speaker uses a single 58mm full-range driver configured for 360-degree sound — double the 40mm driver in the Nest Mini — which Google says delivers 2.5 times stronger bass than its predecessor. Three far-field microphones with advanced processing adapt to room acoustics for voice pickup. Engadget’s review confirmed the improvement over the Nest Mini is real, but also noted that the Nest Audio’s separate 75mm woofer and forward-firing tweeter produced richer sound than the single-driver Home Speaker — meaning buyers upgrading from a Nest Audio specifically for audio quality may find the newer device a lateral move, not an upgrade.

Bloomberg’s review, headlined “Impressive AI, Underwhelming Audio Quality,” independently reached a similar conclusion.

How It Competes With the Echo Dot Max and HomePod Mini

The Google Home Speaker enters a three-way market at the $99 price point alongside Amazon’s Echo Dot Max (also $99) and Apple’s HomePod Mini ($99). Each serves its own ecosystem. The Ambient found in direct testing that Gemini handles reasoning and contextual understanding more confidently, while Alexa+ responds faster to straightforward requests. Amazon’s Echo Dot Max retains advantages in third-party skill integrations and multi-room audio breadth. The HomePod Mini integrates tightly with Apple Music, AirPlay, and Siri — but has no Android support and does not function as a Thread border router.

The most significant structural difference in the three-way comparison is the smart home hub role. The Google Home Speaker’s Thread 1.3 border router and Matter controller make it the only device in this price tier that can bridge Thread-based IoT devices to a home network without a separate hub — an advantage for buyers already invested in Thread-compatible smart home devices.

Google enters this competition without a sub-$99 option. The Nest Mini, which retailed at $49, is discontinued. For buyers who want Gemini in a room but not the full $99 price, there is currently no alternative in the Google ecosystem.

Hardware at a Glance

The Google Home Speaker measures 3.4 inches tall with a 4.2-inch diameter and weighs 0.9 pounds. It is covered in a 3D-knit textile made with 37 percent recycled materials. The integrated USB-C power cable measures 1.5 meters and is non-detachable; a 30W USB-C power adapter is included. Available colors are Hazel and Porcelain in all 18 launch markets; Jade and Berry are exclusive to the US Google Store.

Frequently Asked Questions

Does the Google Home Speaker require a subscription?

No subscription is needed for core Gemini for Home features, including smart home control, natural multi-step voice commands, music playback, timers, and calendar access. However, the most compelling AI features — Gemini Live (open-ended conversational AI), Camera History Search (asking the speaker about Nest camera footage), and Home Briefs (daily AI summaries of household activity) — require Google Home Premium at $10 per month after the bundled trial. Every speaker purchased before September 30, 2026 includes six months of Google Home Premium Standard at no cost.

What replaced the Google Nest Mini and Nest Audio?

The Google Home Speaker, launched June 25, 2026, officially replaces both the Nest Mini and Nest Audio, which are now discontinued. The new speaker retails at $99.99 — $50 more than the Nest Mini’s $49 price point. It offers a larger 58mm driver, on-device Gemini AI inference via a dedicated NPU, Wi-Fi 6, and a Thread 1.3 border router that neither predecessor included.

Does the Google Home Speaker work with Apple HomeKit or Amazon Alexa devices?

The Google Home Speaker is a Matter controller and Thread 1.3 border router, which means it can communicate with Matter-certified smart home devices regardless of brand — including devices designed for Apple HomeKit and Amazon Alexa ecosystems. It does not, however, integrate with Apple AirPlay or Siri, and it does not run the Amazon Alexa voice assistant. Cross-brand smart home device control works through the Matter standard; cross-platform voice assistant control does not.

Why does on-device AI processing matter for a smart speaker?

Prior Google Home and Nest speakers sent every voice query to Google’s data centers for processing, which introduced latency of several hundred milliseconds and required a live internet connection for every command. The Google Home Speaker runs Gemini Nano inference on its dedicated NPU, producing response times Google describes as under 100 milliseconds for standard commands and enabling basic operation without internet connectivity. This architectural change also means basic voice processing no longer transmits data to Google’s servers — a notable shift in context after Google settled a $68 million class action in January 2026 over prior devices’ unintended recordings of private conversations.