It Ain’t Just AI: What We Saw At Google Cloud Next
Google conjured a “Christmas in July” atmosphere by scheduling Google Cloud Next for April 9–11, some four months earlier than usual, to present its AI offerings ASAP and upstage its rivals. With some 30,000 people jammed into a Las Vegas hotel and convention center, expectations were clearly high — and Google gave them plenty to consider.
At third place in public cloud market share, well behind AWS and Microsoft Azure, Google is using the generative AI (genAI) moment to project itself into enterprise IT discussions about when and where to scale out genAI. The technology is key to Google Cloud’s “land and expand” strategy, as otherwise it wouldn’t get a seat at the table among many enterprise customers at the scale of its main competitors. The opening keynote showcased Gemini for Google’s retail industry customers, where Google Cloud first got a foothold in part because of the “anyone but AWS” policy by many big retailers determined not to spend a dime with archrival Amazon.
While Google execs doubtless faced pressure to highlight Gemini’s potential to move on from the consumer version of its controversial public debut in search, attendees could have missed the significance of Google’s other AI-related announcements, such as previewing vector support for customer favorites like BigQuery and Bigtable or the AI Hypercomputer, a pattern for building your own AI platform based on Google Kubernetes Engine. These integrations — call them genAI platform enablement — keep Google well apace with its competitors in this quickly evolving ecosystem. Customers featured in keynotes and sessions were enthusiastic but still tentative, focusing more on what they were expecting and hoping for with regard to genAI, not what they’d already achieved.
Amid the intense competition around foundation models, Google is making Gemini its flagship family of models. Google hopes that these models, currently some of the only multimodal models commercially available, and the ongoing enhancements to predictive and generative AI in Vertex will please longtime data and analytics customers and attract interest from new ones.
Google Cloud Next also highlighted a few customers with more mundane cloud migration stories, where it has significant catch-up to do with its peers. This ordinariness was noteworthy, as it showed a mainstreaming of Google Cloud beyond the trophy, big-logo accounts such as Deutsche Bank, Goldman Sachs, and Ford and ultra-technical users like the CERN atomic particle research lab. Many of those plain-vanilla workloads will, in a year or so, run on the newly announced Arm-based Axion chip, which puts Google into the race, along with Azure, to close the gap on AWS’s years-long lead in Arm alternatives.
Further up the stack, the Google Cloud Next announcements were uneven. Google Workspace, embraced by many medium-sized businesses, doesn’t have the AI-infused offerings of Microsoft 365, and AppSheet doesn’t match the Power Platform offerings in depth or strategic prominence. The developer tools and AI-assisted coding offerings, thanks to a Gemini-driven assist, could give Google Cloud a broader reach among devs, especially if Google continues down its path of specifically AI-oriented developer environments to compete with the rapidly proliferating Copilot tech from Microsoft and AWS’s efforts with Q.
Here’s a closer look at the announcements.
Generative AI models: Google moved forward with opening up the Gemini model family to more customers. Gemini 1.0 has finally launched into GA, while 1.5 Pro is in public preview. Notably, the 1.5 Pro model is one of the only models on the market that can provide a context window of up to 1 million tokens. To support the leveraging of these models into applications, Google continues to expand its features for prompt engineering and management. Google also added features into Imagen 2 that bring it up to par with other image generators on the market such as inpainting and outpainting. Additionally, Google introduced a “Text-To-Live” image generator, which essentially generates a brief, looped GIF-like animation. Despite Google showing off how companies are using Imagen today, enterprises should still be extremely cautious about this due to the uncertain copyright and IP status of both the training data and the generated images.
AI infrastructure: Google continued to roll out AI infrastructure enhancements, with Cloud TPU v5p becoming generally available, and on the NVIDIA front, the A3 Mega compute instance is set for availability in May. Google is also promoting an architecture it calls the AI Hypercomputer that embraces both NVIDIA GPUs and Google’s TPU on Google Kubernetes Engine and Compute Engine, along with optimized storage, multiple models and frameworks, and variable consumption offerings. As a framework rather than a product, the AI Hypercomputer is aimed at customers who want to go beyond AI managed services offerings from Google or its rivals.
Data/analytics: Google Cloud has often gotten in the enterprise door through its data and analytics offerings, with BigQuery gaining a following. Google now seeks to boost BigQuery’s attractiveness by linking it to the Gemini model. Available in private preview, it offers what Google calls “AI-powered experiences for data preparation, analysis, and engineering.” The message to current BigQuery customers: Stick with Google for generative AI while appealing to prospects to consider Google for their broader genAI buildout.
Application development: A highlight of the developer keynote came when Google’s Developer Advocate Chloe Condon was joined on stage by Vercel CEO Guillermo Rauch. In a punchy demonstration, they showed the potential of bringing together Gemini Code Assist and frameworks like next.JS to build a dynamic flight booking application on the fly.
The most important announcement for cloud-native development on Google Cloud was far less flashy: Cloud Assist combined with Code Assist and Context Windows can move AI-assisted development in a far more practical direction for enterprises struggling with the nuances of platform optimization and cloud-native design patterns. These capabilities can help developers focus on coding differentiating business logic rather than infrastructure complexity and deployment pipelines.
Low-code: Despite the overwhelming data on growth in both the substantial low-code platform market and enterprise citizen developer strategies, no AppSheet or other low-code announcements were presented from the keynote stage. But announcements in the AppSheet product breakouts were encouraging and substantive. Most important is the decision to expose Google’s library of Integration Connectors for AppSheet development, which enables new significant use cases (e.g., apps requiring SAP integration) and begins to close a critical platform gap.
Also of interest was the announcement of AI-assisted app creation — not because the idea’s unique but because early demos suggest a thoughtful and well-integrated design. Other announcements (for central platform administration and policy features, genAI-enabled data extraction from photos, and Google Forms integration), while less significant, will prove useful to customers. For the time being, our view (client-only access) remains: AppSheet is an elegant product and could become a powerful weapon in the Intergalactic Cloud Wars — if it is matured, grown, deeply integrated into Google Cloud, and effectively combined with Google’s burgeoning enterprise credibility in genAI.
FinOps: Like the other major cloud providers, Google Cloud is building a plethora of cloud cost management tools for the end user. Following on the heels of its newly announced generative AI capabilities, FinOps consulting services, and FinOps certification for Gemini, the company addressed gaps by announcing features that put its management tools closer to par with AWS and Azure’s tools at Next. Specifically, Google announced daily cost data updates, visibility for Cloud Storage costs, wider CUD coverage and analysis, CUD tracking, and anomaly detection and tracking.
In line with the company’s go-to-market strategy of taking on a multicloud-friendly approach, Google Cloud is also a major player in FOCUS, the open-source project to standardize public cloud billing constructs. Rivals will take note of these moves as cloud customers increasingly focus on FinOps to contain costs for garden-variety workloads to allocate more spending on potential differentiators around AI.
Security: The event’s generative AI theme also permeated the security announcements. Google is planning to use its Gemini/Vertex generative AI capabilities in most areas of Security Command Center Enterprise, including Cloud Workload Security (including cloud security posture management, cloud infrastructure entitlement management, cloud workload protection, and shift-left scanning/vulnerability management). Use cases include large language models for natural language incident descriptions, response playbook creation, automatic policy recommendations, and copilot functionality in policy authoring.
It also announced its new Security Command Center Enterprise, bringing together its cloud security offerings with its SecOps offerings, which has the potential to provide more context for incident response. Overall, the security announcements were light this year, and Forrester expects to see more security announcements from Google later this year, likely at RSA Conference or mWISE.
Workspace: Google Vids, a companion app to Sheets, Docs, and Slides, is a new authoring app designed for video content. Gemini provides AI-powered video generation using text prompts and can use existing documents as a baseline, including options for AI narration.
AI also popped up in classification labels for documents in Google Drive, a $10/user/month add-on that uses AI to apply security classification labels to documents, albeit with up-front model training.
Another AI infusion: Gemini for Meet and Chat is a separate offering designed for employee segments that may not be active users of Workspace’s other capabilities but that rely on chat and meetings for their work. Also priced at $10/user/month, this add-on brings genAI capabilities such as meeting summarizations, captions and translations, and note-taking into Google Meet and Google Chat in an apparent challenge to Microsoft’s new Teams.
Devices: Google’s new Pixel 8 Pro now has Gemini Nano running locally on-device, the company’s first foray into AI devices with full integration across Workspace. While Google was silent on AI on Chromebooks hardware as major Windows competitors embrace the AI PC — a current gap — customers at the event did indicate continued faith in Google’s positioning of end-user computing as moving primarily toward the web.
Google also announced a new premium SKU for the Chrome browser — Chrome Enterprise Premium — that offers more advanced security features with the intention of discouraging companies from embracing third-party enterprise browsers.