Google Cloud CEO sees sunny prospects thanks to AI demand • The Register

Google Cloud CEO Thomas Kurian says the Chocolate Manufacturing unit’s rental computing enterprise has $106 billion of unfulfilled contracts, and he expects Google Cloud will be capable of understand about half of that in income inside two years.

Talking on the Goldman Sachs Communacopia and Expertise Convention in San Francisco on Monday, Kurian said: “So our remaining efficiency obligation, or backlog [as it] is typically referred to, is now at $106 billion. It’s rising sooner than our income. Greater than fifty % of it can convert to income over the subsequent two years.”

In July, Google father or mother Alphabet reported [PDF] $13.6 billion income for Google Cloud in Q2 2025, up 32 % from the identical interval a 12 months earlier. If Kurian is appropriate – and his statements had been preceded by the boilerplate warning concerning the limits of forward-looking statements – that means Google Cloud could have an additional $53 billion to mess around with by 2027.

For the sake of comparability, Microsoft for the primary time in July reported Azure’s annual income and annual development charge, $75 billion and 34 % respectively for its fiscal This autumn 2025. That is about $19 billion on a mean quarterly foundation. AWS, which reported income of $30.9 billion in Q2 2025, grew at a charge of 17.5 % that quarter.

Addressing the funding group, Kurian made a case for ongoing income development based mostly on the potential for on-premises IT to maneuver towards cloud providers.

“Cloud computing continues to develop as the first automobile via which enterprises deploy their core info expertise methods,” he stated. “Cloud, though it has grown, it is nonetheless in its early section as a result of plenty of machines and functions nonetheless run on-premises and haven’t but moved.”

Kurian stated the main target of cloud clients has shifted from software or website hosting to distributors who can present expertise and options to assist companies rework their operations by making use of synthetic intelligence.

Google Cloud, he argued, is well-positioned to seize that enterprise via product differentiation in efficiency, price, reliability, and effectivity in AI infrastructure.

Citing Google’s eleven years of constructing AI methods and chips, Kurian stated Google’s AI methods are closely optimized for coaching and inference.

“For instance, should you’re working a big scale cluster, we have now two occasions the facility effectivity, that means you get two occasions the flops per watt,” he stated. “And with energy now, the scarce useful resource, you get much more capability. We’re usually seeing a couple of fifty % efficiency delta between us and different gamers.”

By way of whole capability on a single system, he stated, you will get 118 occasions extra throughput on Google’s methods in comparison with “the subsequent participant.”

Kurian went on to tout Google’s AI storage optimization and its high-bandwidth optical networking that may be simply shifted from coaching to inference with out downtime.

“For those who evaluate us to different hyperscalers, we’re the one hyperscaler that gives our personal methods and our personal fashions, and we’re not simply reselling different individuals’s stuff,” he stated, seemingly ignoring Microsoft’s latest debut of its own in-house models and Amazon Nova. “The amount of tokens we course of is twice [compared to] different suppliers in half the time – so roughly 4 occasions the quantity.”

Kurian went on to element the 5 methods Google is monetizing AI – one thing of a priority for traders given the billions hyperscalers are spending to construct out information middle capability for AI providers.

Some clients pay based mostly on consumption of tokens in AI fashions, he stated. Some pay for subscriptions to providers like Google Agentspace or Workspace.

Some monetization comes from elevated product utilization, Kurian added, pointing to how utilizing Google’s cybersecurity agent includes risk evaluation AI. “For instance, we’re over 1.5 billion risk hunts, we name it, utilizing Gemini. And that drives extra utilization of our safety platform. Equally, we see development in our information cloud.”

And, Kurian stated, Google monetizes a few of its merchandise via value-based pricing. So clients utilizing the corporate’s customer support system might pay by deflection charge – assist requests not routed to an individual. And clients utilizing Google’s artistic instruments would possibly pay based mostly on promoting conversion charge.

Lastly, there’s the upsell – the place Google Cloud clients transfer to pricier service tiers for bigger quotas, extra succesful fashions, and different niceties.

“Sixty-five % of our clients are already utilizing our AI instruments in a significant means,” he stated, including that those that achieve this use 1.5 occasions as many merchandise as these not utilizing AI instruments but.

As to how companies are utilizing AI, Kurian described 4 main domains. Corporations are utilizing AI to assist construct digital merchandise, as Google did when it helped Warner Bros. rework “The Wizard of Oz” for the Las Vegas Sphere. Others are utilizing AI for customer support and point-of-sale help.

“Others are utilizing it to streamline the core of the corporate and their again workplace,” stated Kurian, pointing to how Dwelling Depot makes use of AI for its HR Assist Desk to reply worker questions on advantages and the like. After which there are firms utilizing AI for IT functions, like writing code and scanning for cyber threats.

In brief, if you’re within the enterprise of renting cloud infrastructure, you’ll be able to earn much more hire by promoting companies on AI providers that rely upon cloud infrastructure. ®