Building the next generation of generative artificial intelligence is not an easy thing, and no single cloud provider is able to provide all of the hardware, software, systems, frameworks and expertise required to support its development.

Fortunately, that isn’t necessary, as Google Cloud has instead turned to its wide ecosystem of partners to create what it says is an expansive and open platform for AI development. In an update today at Google’s Cloud Next ’24 event in Las Vegas, Kevin Ichhpurani, corporate vice president of global ecosystem & channels at Google Cloud, explained how the company has been pursuing a partner-led approach to delivering the services and capabilities necessary to support the most advanced generative AI models and applications.

Expanded AI infrastructure

That approach has paid off handsomely, Google Cloud claims, since it now can offer one of the widest choices to customers when it comes to the infrastructure to support their AI needs. Its infrastructure portfolio includes both its own Tensor Processing Units, including the newly announced TPU v5p, plus an extensive selection of Nvidia Corp. graphics processing units that will soon incorporate the new Blackwell GPUs when they launch later this year.

“This broad selection … ensures that our partners can offer customers the greatest choice of how to build and run their AI workloads,” Ichhpurani said.

Compute is not the only infrastructure requirement for AI, as companies must also get to grips with their data estate to train and fine-tune the models they’re developing. To that end, Google Cloud goes further than offering its own data warehouse service, BigQuery, offering access to dozens of the industry’s leading third-party data platforms, such as Confluent, Databricks, Elastic and MongoDB.

These platforms can all be accessed through Google Cloud’s Vertex AI platform for AI development, which essentially acts as a network for connecting AI models and applications to various data warehouses and databases, enabling the flexibility and freedom customers’ desire, Ichhpurani explained.

Expanded choice of foundation models

Google Cloud has also extended its hand to the developers of many of the best foundational AI models in the business. Vertex AI’s Model Garden now provides access to a staggering 130 foundation models, including open-source ones such as Meta Platforms Inc.’s Llama 2 and Google’s own Gemma, plus popular proprietary models from partners such as Anthropic PBC, Cohere Inc., Mistral AI and others. In the last few weeks, Google Cloud added new models, including Claude 3 Sonnet and Claude 3 Haiku, as part of its mission to keep expanding customers’ access to the most innovative developers in generative AI.

In addition, Google Cloud recently partnered with Hugging Face Inc., which operates the largest repository of open-source models.

Of course, Google Cloud has its own selection of foundational models too, with its Gemini family. According to Ichhpurani, it’s simpler to access the Gemini models now, thanks to partnerships with technology providers such as Salesforce Inc. and SAP SE, which have made them available directly through their own platforms. For instance, the Gemini models are now available in SAP’s Generative AI Hub, for SAP customers to quickly get started building generative AI applications that utilize SAP data.

AI developer tooling and applications

Ichhpurani said he understands that developers may not want to use the Vertex AI platform, and in that case they can instead access a range of alternatives via Google Cloud, including Colab, VS Code, JetBrains, Replit and Stack Overflow. In addition, they can find various other tools and services for every other layer of the AI stack through Google Cloud Marketplace, including ArizeAI, Cast AI, Gretel, OctoAI and Writer.

“We continue to see momentum from the world’s largest software and data companies co-selling with us,” Ichhpurani continued. “In the last six months, many of our partners have introduced new AI apps and solutions in Google Cloud Marketplace, including Broadcom, Deloitte, Five9, Kinaxis, MicroStrategy, NielsenIQ, Quantiphi, ServiceNow, Shopify, Box, Trend Micro, Twilio and UiPath.”

Third-party enterprise apps and software

In addition, Google Cloud’s ecosystem of partners now offers numerous generative AI-powered applications and services built atop Vertex AI and the Gemini models. Those apps range from tools that can summarize documents and write job descriptions, to ones that can parse difficult contractual language and more.

At Google Cloud Next, dozens of the company’s partners are launching new AI-powered capabilities in their products, Ichhpurani said. For instance, Palo Alto Networks Inc. named Google Cloud as its preferred AI provider, and will use Gemini as the basis of its new threat analysis and incident summary tools on the Cortex XSIAM platform. Then there’s CrowdStrike Holdings Inc, which is integrating its Falcon platform with Google’s AI services, and Exabeam Inc., which is tapping Google Cloud AI to improve security outcomes for its customers.

AI implementation

Ichhpurani acknowledged that many Google Cloud customers rely on expert consultancies and partners to implement AI into their business operations, and the company aims to support this by extending a hand to dozens of well known services providers. He said Google Cloud’s systems integrator and consulting partners, including Accenture Plc, Capgemini Inc., Cognizant Inc., Deloitte Touche Ltd., PriceWaterhouseCoopers and others, have so far committed to training more than 200,000 experts to use the company’s AI offerings,

Google Cloud has facilitated this by providing those partners with the resources, rewards and upskilling opportunities required to support customer success. And now the company is expanding its support for these partners, with a new Generative AI Services Specialization and new Generative AI Delivery Excellence and Technical Bootcamps. In addition, the company has launched a new user experience within its Delivery Readiness Portal to simplify access to the most important resources.

Support for AI startups

Google Cloud’s extensive partner ecosystem is one of the main reasons why its AI cloud infrastructure and services are used by more than 60% of all funded generative AI startups, including almost 90% of generative AI “unicorns,” which are those startups valued at more than $1 billion.

“At Google Cloud Next, we’re announcing many new and expanding startup customers who are using Google Cloud’s AI-optimized infrastructure, models, AI and ML tooling and routes-to-market like Google Cloud Marketplace,” Ichhpurani said. “We’re also showcasing new businesses being built with AI from the ground up, like Ohalo Genetics, and vertically aligned startups who are finding exciting ways to solve industry-specific issues with AI, like Atomo Health, Rad AI,and Queenly.”

But Google Cloud doesn’t plan to rest on its laurels anytime soon, as it’s also partnering with Nvidia on a new Google for Startups Accelerator: AI-First program, which is focused on early-stage companies building AI-powered applications and services.

Kicking off this program is the inaugural cohort of AI-First North American Startups program, which caters to startups in the U.S. and Canada. Later this year, it will launch similar programs for AI startups in Europe, India and Brazil, Ichhpurani said.

The accelerators will provide 10 weeks of hands-on mentorship and technical project support for early-stage firms that use AI in their core services or products. As part of the program, they’ll gain access to various AI leaders and experts at Google, as well as credit to use Google Cloud’s most advanced AI tools and infrastructure.

At the conclusion of the program, the selected 15 startups will graduate to become members of the Google for Startups Accelerator community, which provides access to ongoing support and connections with mentors, like-minded founders and investors.

Image: Vector_Corp/Freepik