When Google Cloud held its annual Cloud Next event this week, the company’s message was less about the novelty of AI agents and more about the operational reality that follows once they multiply. In a blog post published on 22 April 2026, Google framed the moment as a turning point: the conversation has moved from building a single agent to managing them at scale.
That shift matters for any business that has already experimented with AI agents or is planning to. The tools and platforms announced at Cloud Next suggest that the next phase of enterprise AI will be defined less by what agents can do individually and more by how organisations govern, secure, and optimise them collectively.
Refer & earn: 30% of a client’s first month, then 10% of every month after that, for up to 24 months per client. See the partner programme →
From experimentation to infrastructure
Google reported that its first-party models now process more than 16 billion tokens per minute via direct API use by customers, up from 10 billion the previous quarter. That figure, while abstract to many business leaders, signals a rapid increase in the volume of AI work being handled programmatically. It also underscores why infrastructure decisions are becoming more consequential.
To support that growth, Google said that in 2026 just over half of its overall machine learning compute investment is expected to go towards the Cloud business. In practical terms, this means the company is directing a majority of its AI infrastructure spending toward the services that external customers and partners use, rather than keeping it solely for internal projects.
For businesses, the implication is straightforward: the cloud provider you choose is increasingly also your AI infrastructure provider, and its investment priorities will shape what you can build and at what cost.
The agent management challenge
Google introduced Gemini Enterprise last autumn as an end-to-end system for what it calls the agentic era. At Cloud Next, the company reported that in Q1 it saw 40% growth in paid monthly active users quarter-over-quarter for that product. That metric, while specific to Google’s own platform, reflects broader enterprise adoption of agent-based tools.
But growth brings complexity. Google acknowledged this directly, noting that every employee in every organisation can become a builder, and that this shift comes with challenges. The company posed the question that many IT and operations leaders are now facing: how do we manage thousands of them?
To address that, Google announced the Gemini Enterprise Agent Platform, describing it as a secure, full-stack connective tissue for building, scaling, governing, and optimising agents. It is positioned as a mission control for the agentic enterprise. For business readers, the key takeaway is that agent management is being treated as a distinct platform category, not just a feature of existing cloud services.
This matters because without centralised management, agents can proliferate across departments, each with its own data access, security posture, and performance characteristics. The result can be duplicated effort, inconsistent compliance, and difficulty in measuring return on investment. A platform approach aims to bring those disparate efforts under one governance framework.
Security as an agentic problem
Google also used Cloud Next to highlight new agentic solutions for threat detection, part of an AI-powered cybersecurity platform that combines Google’s Threat Intelligence and Security Operations with Wiz’s Cloud and AI Security Platform. Additionally, the company launched Wiz’s new AI Application Protection Platform, which it said provides autonomous protection from code to cloud to runtime across multicloud, hybrid, and AI environments.
The framing here is notable: AI can increase security risks, but Google argues that cloud customers now have AI on their side to protect their organisations. For businesses, this reinforces that security considerations must be part of any agent deployment from the start, not an afterthought. As agents gain access to more systems and data, the attack surface expands, and traditional security tools may not be sufficient.
Hardware tailored for AI workloads
On the infrastructure side, Google announced that this year it is bringing the eighth generation of its Tensor Processing Units with a dual chip approach. These will be offered to Cloud customers as a core part of its selection of compute processors, alongside a portfolio of NVIDIA GPU instances.
For business decision-makers, the significance is less about the technical specifications and more about choice and availability. Having multiple processor options means organisations can match workloads to the most appropriate and cost-effective hardware, whether that is Google’s own TPUs or NVIDIA GPUs. It also signals that Google is committed to offering flexibility rather than locking customers into a single architecture.
Google also noted that it aims to be customer zero for its own technologies, testing and scaling them internally before offering them to cloud customers. It cited Bigtable, its database service, and its TPUs, which have been important in training and powering its Gemini models, as examples. This approach, if it holds, could mean that products arrive with more real-world validation than typical early releases.
What this means for your business
If you are already using AI agents or planning to, the Cloud Next announcements suggest several practical considerations.
First, agent management is becoming a distinct capability. As the number of agents grows, having a central platform to govern them will likely become as important as the agents themselves. Businesses should assess whether their current tools can scale from a handful of agents to hundreds or thousands, and whether they provide the necessary oversight.
Second, security must be designed in. The integration of threat intelligence and AI security platforms indicates that security vendors are adapting to the agentic era. Businesses should ensure that their security posture evolves alongside their agent deployments.
Third, infrastructure choice matters. The availability of TPUs and GPUs from a single cloud provider gives organisations more options to optimise for performance and cost. It also means that infrastructure decisions should be revisited as new generations of hardware become available.
Finally, the pace of change is unlikely to slow. Google described the pace of technological change since last year’s Cloud Next as never faster, and its momentum in cloud and AI suggests that the competitive landscape will continue to shift. For businesses, staying informed and being willing to adapt will be essential.
Google also mentioned that it will share more at Google I/O on 19 May, which may bring further announcements. For now, the message from Cloud Next is clear: the agentic era is here, and managing it well is the next challenge.
Sources
- Cloud Next ‘26: Momentum and innovation at Google scale, Google, published 22 April 2026.
Disclosure: Faida Workforce provides consulting and implementation services for AI and agent-based systems. This article is for informational purposes and does not constitute a recommendation of any specific vendor. AI tools assisted in the production of this article; a human editor is accountable for its content.
30% of their first month, then 10% every month after that
Refer a business to Faida Workforce and you earn 30% of the client’s first paid month, then 10% of every month for up to 24 months per client — rising to 15% from the accrual after you pass 10 active clients. Paid by PayPal, batched on the 1st.
10 Pro clients = $2,598.70 over 24 months — $3,748.13 at the 15% tier.
Figures are potential earnings worked from the published rates, assuming every referred client pays for the whole cap — not a promise, not an average. The $1 trial is never commissionable.
