Google Cloud's second chance at the enterprise

Welcome to Runtime! Today: Google Cloud makes its pitch to developers and CIOs as the best place to build enterprise AI apps, the meteoric rise of MCP hits a snag, and the latest enterprise moves.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

AI is a mulligan

LAS VEGAS — Looking back 15 years ago, Google was just as well positioned as Amazon — and much better positioned than Microsoft — to take over enterprise computing, given the incredible array of data-center breakthroughs it made while building the world's search engine. Obviously, that's not how things worked out, but if enterprise AI becomes the next great platform that runs the world economy, Google Cloud is in a very interesting place.

Around 30,000 people showed up in Las Vegas this week for Google Cloud Next, up from last year's total of 20,000, and you could feel the increase walking around the halls of the Mandalay Bay Convention Center. The main topic was, of course, AI, and while most in attendance have been bombarded with pitches for generative AI for nearly three years from just about every single company that sells enterprise tech, Google Cloud has one of the most comprehensive stories to tell.

CEO Thomas Kurian got right to it in his opening keynote speech on Tuesday, arguing that Google Cloud has three distinct advantages over its competition.
Those include a cloud infrastructure service with "leading price, performance, precision, and quality," a commitment to building services that work across multiple clouds and on-premises data centers, and a software-development platform that "enables you to adopt AI deeply while addressing the evolving concerns around sovereignty, security, privacy and regulatory requirements," he said.
Over the next 90 minutes, Kurian and a parade of Google executives introduced new products and services such as expanding its private wide-area network to all customers, a new version of its GPU-like TPUs (tensor processing units), and new capabilities for Vertex AI, the set of tools and services that customers use to build generative AI applications.

One of the major reasons why Google fell behind AWS and Microsoft among the first generation of cloud adopters was its lack of empathy for the not-quite-Google-scale problems that most enterprises encounter when trying to build their digital infrastructure. The greenfield generative AI opportunity has allowed the company to reintroduce itself.

"The goal isn't to be all or nothing on one end of the spectrum; like, [selling only] Google-originated innovation or simply just taking orders from customers and just fulfilling them. We have to create a blend in our roadmap," said Will Grannis, vice president and CTO, in an interview.
AWS became an enterprise tech powerhouse thanks to its pioneering cloud services and Microsoft leveraged deep enterprise relationships to transition into its cloud era, but the Google of a decade ago seemed to expect that CIOs would assume anything that was good enough for Google was good enough for them.
This time around, Google is doing a much better job figuring out how to package the technology it develops internally to send you down a YouTube rabbit hole in a more digestible format for enterprise customers.
"The basis of how [customers] thought about it in 2007, ’08, ’09, was all about, 'how do they help me either go faster by building apps or reduce my cost of data centers by allowing me to lift and shift workloads?' Now, they’re thinking about it in a different way. They’re looking at it as, 'Can I use AI to transform my business? Who’s got the best platform and tools to help me do that?'" Kurian said in an interview with Stratechery.

Needless to say, AWS and Microsoft are not sitting idly by while Google Cloud chases those customers, who are under the gun from their AI-pilled bosses to articulate a strategy for their companies.

But there's no guarantee that AWS's huge cloud user base will want to build AI apps on its platform when it's never been easier to run digital infrastructure across multiple clouds.
And Microsoft's prescient move to sign an exclusive deal for OpenAI's models confers less and less of an advantage every month as new boundary-pushing models come out from a variety of companies, including Google and Amazon.
"Google is building for a unique moment; we're investing in the technology and the ecosystem to power your growth and transformation," Kurian said Tuesday.
No matter how it all turns out, when there is fierce competition for the future of enterprise tech it's a great time to be a buyer.

Oh yeah, security

The hype around Anthropic's Model Context Protocol has really taken off over the last month as developers realize the implications it could have for making agentic AI applications easier to build, and Google Cloud not only embraced it this week but also introduced its own service called Agent2Agent that was designed to work in concert with MCP. But as more developers start to play with any emerging technology, the shortfalls start to appear.

Simon Willison highlighted some of those problems Wednesday, referencing a blog post by Elena Cross entitled "The 'S' in MCP stands for security," a headline that made enterprise tech news hacks instantly jealous. Like a lot of technology based around large-language models, MCP is susceptible to prompt injections, through which a malicious hacker can cause all kinds of chaos.

"The great challenge of prompt injection is that LLMs will trust anything that can send them convincing sounding tokens, making them extremely vulnerable to confused deputy attacks," WIllison wrote. As is the case with any promising new technology, MCP backers will have to solve some problems to make it go mainstream, according to Equixly co-founder and CTO Alessio Dalla Piazza: "its current security posture exhibits concerning weaknesses reminiscent of early web application security challenges."

Enterprise moves

Matt Kraning is a new partner at Menlo Ventures, joining the firm after several years at Palo Alto Networks with plans to invest in "AI, enterprise SaaS, national defense, and cybersecurity companies."

Stacy Dillow is the new chief people officer at HPE, following several years at Fluor Corporation.

The Runtime roundup

"We weren't breached! We weren't breached!" Oracle continues to insist as it slowly shrinks and transforms into a corn cob.

"The US Secretary of Education referred to AI as 'A1,' like the steak sauce," TechCrunch reported, and there's a reason Runtime comes out just before happy hour.

Thanks for reading — see you Saturday!