Google Cloud Debuts New Era of AI-powered Cloud Services with Duet AI

AI took center stage at Google Cloud Next 2023, where Google Cloud CEO Thomas Kurian shared the company’s optimism and vision for a technology that will transform the way businesses work and operate over the next decade. A front-runner in AI innovation for nearly seven years, Google (Nasdaq: GOOGL) has cultivated a strong brand in the space by helping customers draft emails in Gmail, experience immersive locations in Google Maps and improve outcomes in Search, to name a few examples. Google Cloud — the highly strategic and now profitable division within the broader Alphabet corporation has nurtured Google’s AI technology, brand and customer relationships to transition to an enterprise-grade cloud provider and position Google Cloud Platform (GCP) as the best place to run AI, machine learning (ML) and analytics workloads. Now with the latest subset of AI — generative AI (GenAI) — taking the tech market by storm, Google Cloud is prepared to advance this position, using internally developed large language models (LLMs) to bolster the GCP and Workspace portfolios and offer customers a “new way to cloud.”

TBR Perspective

For years, Google Cloud has distinguished itself from legacy peers as a born-in-the-cloud provider, allowing it to commit to the latest frameworks and tooling that help digitally native customers build cutting-edge applications. This core value proposition closely aligned with the central theme of Next, which focused on delivering customers a “new way to cloud.” If the old way is solving a technical challenge through labor-intensive manual processes, then the new way is leveraging highly automated AI-based solutions that allow customers to personalize, secure, predict and ultimately realize business outcomes. Google Cloud is striving to help customers do just that by embedding GenAI throughout its portfolio, which may not only encourage current customers to keep expanding their GCP investments but also attract the legacy enterprises that are ready to unlock the potential of cloud and GenAI.

Kurian prefaced the event with an overview of GenAI, and one piece of Google Cloud’s GenAI strategy quickly became clear: Google is striving to offer the broadest set of LLMs among peers. With Pathways Language Model (PaLM) 2, Codey, Imagen and Med-PaLM, Google offers its own set of generic and industry-specific models, but when combined with partners like Meta (Nasdaq: META) and Anthropic, as well as the open-source community, there are over 100 foundation models available from Google.
This approach reaffirms Google’s commitment to offering customers maximum choice when it comes to the models they use, including the ability to customize models based on internal data. Aside from supporting an array of models, Google will continue to refine PaLM 2, which has expanded from 4,000 to 32,000 tokens, or the equivalent of roughly 85 pages of text. These developments will be important as they power new products and features, including the new cognitive tool taking GCP by storm: Duet AI.


Announced in May, Google Cloud’s Duet AI is described as an “always-on AI collaborator” that works behind the scenes to complete productivity tasks in Workspace and perform developer functions, such as automated coding, in GCP. At Next, Google Cloud announced the general availability of Duet AI in Workspace, officially allowing customers to take advantage of GenAI features, such as describing an image you want with a few keywords and having it populate within Google Slides or automatically classifying rows of data in Google Sheets. These capabilities are all about improving the user experience and strategically helping Google Cloud cross-sell GCP services to the vast base of 10 million paying Workspace customers.


At the event, Google Cloud also announced it is expanding the capabilities of Duet AI on GCP in preview, with general availability slated for later this year. There are several new preview capabilities to unpack from context-aware code generation to automated API publishing within GCP’s Apigee API Management service. One of the capabilities we found most compelling, however, is support for Duet AI with Database Migration Service (DMS), as it speaks to Google Cloud’s strategy of using GenAI to attract more legacy workloads and become a competitive modernization partner for the traditional enterprise.


In the scenario that Brad Calder, VP and general manager of GCP, introduced at Next, customers can use Duet AI to migrate Oracle Database workloads to Google Cloud’s AlloyDB. Currently, much of the database schema and code is automatically converted to AlloyDB, but for functions specific to Oracle (NYSE: ORCL), it can be challenging for developers to make coding changes required for migration. With Duet AI, customers will be able to open the DMS console and to manually code just one Oracle function and have that coding change automatically applied to all other cases. While we do not believe Duet AI will be enough to sway invested, customized customers away from Oracle Database, it will help Google Cloud better compete with Amazon Web Services (AWS) (Nasdaq: AMZN) and Microsoft (Nasdaq: MSFT) for the subset of customers prepared to leave their legacy Oracle systems behind.


It is still early days for Duet AI, but with over 1 million users to date, the service is gaining acceptance throughout both Workspace and GCP. Complementing Google Cloud’s existing Vertex AI solution, which is for more advanced developers to help them better build with GCP IaaS and PaaS, Duet AI is about making AI easily accessible under the hood for line-of-business teams and novice coders. Targeting both these personas, Vertex AI and Duet AI will define Google Cloud’s AI portfolio going forward and potentially help the company attract both digitally native and traditional enterprise customers.

Workload-optimized Infrastructure Is Core to the GCP Strategy

The rise of data-intensive workloads with specific compute requirements has driven Google to re-evaluate its infrastructure architecture over the past couple of years. At Next, Google Cloud mentioned its under-the-radar Titanium architecture, which Google Cloud claims has been powering certain services for years. With the Titanium system, tasks such as block storage and virtual networks are offloaded to dedicated chips — infrastructure processing units (IPUs) codeveloped with Intel — to reduce reliance on the host CPU and help maximize performance. TBR notes this approach mirrors what AWS did over a decade ago with the Nitro System, the core architecture powering all its next-generation EC2 instances. Like AWS, Google Cloud is using its architecture to support customers beyond general-purpose workloads, and considering the compute requirements for GenAI workloads, we can expect Titanium to power all future generations of Google Cloud’s IaaS services.


Google Cloud also launched the v5e tensor processing unit (TPU), which is integrated with popular GCP services, including Vertex AI, and promises more than double the performance per dollar for both AI training and inference workloads compared to the v4 TPU. One other key attribute of the v5e TPU is its versatility, as customers can choose up to eight virtual machines (VMs) and 250 chips, allowing them to manually configure their hardware based on the size of the LLM.

The new TPUs come as Google Cloud competitors, namely AWS, leverage their custom chips to support the compute demands of LLMs and GenAI and offer the market an alternative to what seems to be an increasingly scarce supply of GPUs. Even so, customers clearly value NVIDIA’s AI hardware, and that is why Google Cloud continues to support GPUs and is bringing its A3 VMs, based on NVIDIA H100 GPUs, to the market next month.

Bringing GenAI to Security

In TBR’s own discussions with enterprise customers, we have uncovered several established GenAI use cases across customer service, administration and software development. Security, however, appears to be a budding area of interest, especially considering the current lack of trained security professionals in the market. At least for now, Google Cloud appears to be a front-runner in this space, as the company has its own LLM trained on security use cases (Sec-PaLM 2) to predict threats and help explain behavior surrounding malicious scripts.


Further, Google Cloud has integrated Duet AI with Mandiant Threat Intelligence, which applies GenAI to actually summarize what a hacker is trying to accomplish. Security remains the top concern for customers, and TBR findings reveal that customers generally expect their cloud service providers to do more to address security natively, so steps Google Cloud can take to make security easier for the customer will be well received. The Mandiant acquisition and other investments in security are also key to Google Cloud’s enterprise positioning and efforts to attract customers that do not use GCP beyond analytics and AI.

GenAI Presents New Opportunity in Google Cloud’s Consistent Partner Strategy

Two of Google Cloud’s biggest differentiators have been its partner-first strategy and data analytics capabilities, and the company converged and advanced both during the latest Next event. The rising interest and investment in GenAI capabilities is a ripe opportunity for Google to improve its standing in the cloud platform and infrastructure spaces, and the flurry of partner announcements indicated the company is looking to seize the moment.


The partner announcements reflected the breadth of Google’s ecosystems strategy, impacting programs spanning ISV to alliances to global systems integrator (GSI). Two of the biggest SaaS companies, SAP (NYSE: SAP) and Workday (Nasdaq: WDAY), were part of the new announcements during the event. Workday is now live on GCP, a major win for the platform. The relationship dates to 2021, when Workday named GCP as its preferred cloud partner of choice for specific verticals, including healthcare, financial services and retail.


Even more than just being hosted in GCP, Workday will eventually take advantage of Google of GenAI capabilities, including building new features into the flow of its core human capital management (HCM) and finance offerings. The packaging of core SaaS offerings with GenAI features is an emerging trend, as vendors including Microsoft and ServiceNow (NYSE: NOW) have already announced pricing for packages that will shortly be available. The relationship with Workday gives Google Cloud the opportunity to integrate its AI assets in a leading SaaS property, creating revenue growth opportunities but also awareness within a large customer base.


Google Cloud and SAP also had new joint announcements at the event, building on a long history between the two vendors. As Kurian noted when presenting at SAP’s Sapphire event earlier this year, “Google Cloud has had an amazing collaboration with SAP going back five years.” Google Cloud’s parent company, Alphabet, runs on SAP for core systems and is a vocal advocate for SAP S/4HANA. At the Google Next event, it was announced that their joint relationship would be extended yet again, as SAP will utilize Google foundation models via Vertex AI and SAP data to develop solutions for customers that advance sustainability initiatives and to improve safety and manufacturing processes within the automotive industry.


Beyond the big-name alliance announcements, Google Cloud is looking to become a hub of the AI ecosystem and cited progress in that regard. Google Cloud now has more than 100 foundation models from third-party and open-source contributors in its model garden. Google Cloud’s model garden includes open-source contributors like Confluent (Nasdaq: CFLT), DataRobot, Elastic (NYSE: ESTC) and MongoDB (Nasdaq: MDB) as well as more commercial participating organizations like Acxiom (Nasdaq: ACXM), Bloomberg, CoreLogic (NYSE: CLGX), Equifax (NYSE: EFX), NIQ and TransUnion (NYSE: TRU).


The announcements at Google Next advanced partner initiatives, following a consistent strategy to build the partner ecosystem. From a strategic perspective, Google Cloud is focused on profitability and sustainable growth, and partners are the key to achieving those goals. Heading into the event, Google Cloud sustained a profit for two consecutive quarters, with partners allowing the company to develop multifaceted relationships that maximize clients’ GCP consumption.


In the months leading up to the event, Google Cloud reinforced its commitment to partners through the updated Partner Advantage program, which offers product-specific tracks, financial incentives and new certifications to support partners as the opportunity around GCP expands, particularly in the area of GenAI. Additionally, Google Cloud promised to provide partners with greater access to its internal resources, training programs and best practices, suggesting the company is becoming more hands-on with its GSI partnerships, from presale to postsale, to ensure initial migrations and custom projects are successful.