Jensen Huang, CEO of Nvidia, has publicly announced that he has identified a completely new market: CPUs designed specifically for AI agents. The estimate is $200 billion. This is a strong signal for the entire tech industry.
However, the news isn't just for the big players. In fact, Italian B2B SMEs considering AI infrastructure investments need to understand how this paradigm shift will affect costs, availability, and architectures in the next 18-24 months. Consequently, infrastructure choices made today could prove premature or, conversely, strategically advantageous.
In this article, we at SHM Studio Let's analyze what has changed with this announcement, what immediate impact is expected on the market, and what operational moves are worth considering for medium-sized Italian companies working in the B2B or advanced retail sector. Therefore, reading is recommended for those who manage technology budgets or oversee corporate digitalization projects.
Huang's announcement: a market that didn't exist yet
During a public event in May 2026, Jensen Huang announced he had identified a "completely new" market for Nvidia. It is not about GPUs, the company's traditional playground. It is about CPUs specifically designed to support AI agents. The estimated value shared is 200 billion dollars. According to TechCrunch , Huang described this segment as distinct from and complementary to Nvidia's current offering.
Therefore, the statement isn't just a financial projection. It's a precise strategic signal: AI agents require a different type of processing compared to pure generative models. They need CPUs optimized for low latency, sequential reasoning, and autonomous task orchestration.
Plus, this positioning puts Nvidia head-to-head with Intel and AMD in a totally new arena. The enterprise CPU market is already worth hundreds of billions. Adding the AI agent piece completely reshapes it.
Why AI agents need dedicated hardware
An AI agent isn't a chatbot. It's a self-driven system that can plan, run, and fix actions step by step. So, the computing load is totally different from a model just answering single prompts.
GPUs excel at massive parallelism. However, AI agents often work in serial mode: they read context, decide, act, verify. This flow requires CPUs with high cache, low latency, and efficient context memory management. In particular, architectures like those described in recent reports by Gartner on enterprise AI confirm that 2026 will mark a turning point in the adoption of agents in structured business processes.
As a result, Nvidia is anticipating a demand that B2B SMEs will start expressing in the next 12-18 months: cloud or on-premise hardware optimized not for training models, but for running agents in production.
Immediate impact on the cloud market and providers
The announcement has direct implications for major cloud providers. AWS, Google Cloud, and Microsoft Azure will need to update their compute instance offerings to include CPUs optimized for agent workloads. This process takes time. Therefore, in the short term, SMEs already operating on public clouds will not see immediate changes.
However, prices for AI-oriented instances could face downward pressure in the medium term. In fact, Nvidia's entry into this segment increases competition. Similarly, hybrid infrastructure vendors will need to update their product roadmaps.
According to the analyses of McKinsey on the AI market , companies that put money into the right AI setup see operational returns 20-30% higher than those going for off-the-shelf stuff. This matters even more when you talk about rolling out agents at scale.
What Italian B2B SMEs need to do now
First off, we need to tell apart two business scenarios. The first involves SMEs that are still figuring out whether to use AI tools. The second involves those that already have pilot projects or live rollouts going.
For companies in the first group, the Nvidia announcement suggests not rushing into standalone hardware investments. In fact, the ecosystem is being redefined. Relying on managed cloud solutions remains the most flexible choice for the next 12 months. The services of AI consulting by SHM Studio specifically include evaluating the architecture best suited for the specific business context.
For companies in the second group, however, it makes sense to actively monitor the cloud providers' roadmaps. Following the Nvidia announcement, AWS and Google Cloud will update their offerings of agent-optimized instances. Therefore, those with active projects should plan an infrastructure review by the end of 2026.
In addition to this, B2B SMEs using CRMs, ERPs, or marketing automation platforms need to check if their vendors are integrating agent-based features. This directly impacts the strategies of Digital marketing and of LinkedIn campaigns based on advanced automation.
The open construction site: what Nvidia hasn't said yet
Huang's announcement is deliberately high-level. No product names, launch dates, or technical specifications have been communicated. Therefore, the $200 billion estimate remains a total addressable market projection, not a declared revenue plan.
Several questions remain open. How will this new CPU line position itself against the ARM architecture, which currently dominates AI inference chips? What will the licensing models be for cloud providers? To what extent will SMEs be able to access these resources without enterprise intermediaries?
Despite this, the strategic signal is clear. Nvidia is building a vertical ecosystem that goes from the GPU for training to the CPU for agent execution. This vertical consolidation is comparable to what Apple has achieved with its M-series chips in the consumer segment. The implications for the enterprise market are significant.
For those managing projects of web development with integrated AI components, or campaigns of Google Ads optimized by predictive algorithms, understanding the evolution of the underlying hardware is not an academic exercise. It is a concrete operational variable.
18-month outlook: what to expect by the end of 2027
By the first half of 2027, it is reasonable to expect the first concrete product announcements from Nvidia in this segment. Similarly, major cloud providers will integrate CPU-agent instances into their managed offerings. The market will start differentiating prices between generative workloads and agent-based workloads.
For Italian SMEs, this means that infrastructure decisions made today will directly impact competitiveness in 2027-2028. In particular, those who start structuring AI agent-based workflows now—even in a simple form, through no-code or low-code platforms—will have an advantage when dedicated hardware becomes available at affordable costs.
Finally, it is worth remembering that adopting AI agents isn't just about IT. It impacts the SEO Strategy , the production of content , managing ad campaigns, and even website structures. We at SHM Studio we are already guiding SME clients through this strategic adaptation journey. Anyone who wants to dive deeper can check out our section Blog or contact us directly from the page contacts .
Related articles
Discover more articles exploring similar topics, selected to offer you a more complete and stimulating perspective. Each piece of content is carefully chosen to enrich your experience.
