Disclosure: this article contains affiliate links. If you buy through them we may earn a commission at no additional cost to you. Our recommendations are based on the evidence recorded alongside this article, and prices were verified on the date shown.

This guide is for anyone deciding whether to act on Nvidia's newest CPU announcement today or wait for firmer terms. Short version: the shipping claim is real and sourced directly from Nvidia; the price, the spec sheet, and which cloud providers carry it are not — verify those before you commit anything.

What changed and why it matters

Nvidia published a blog post on its own site with the headline "Delivering Vera: Nvidia's First CPU Built for Agents Is Shipping Now". The claim comes directly from Nvidia, not a leak or a secondhand summary, and the headline states two things at once: this is Nvidia's first CPU designed specifically for agents, and it is shipping now — not "coming soon" or "sampling to select partners."

That distinction matters for anyone renting compute for agentic workloads, because "shipping now" from a chip maker and "available to rent from a cloud provider today" are two different milestones, often separated by weeks or months. Nvidia's post confirms the first one. It does not, in the material available for this guide, confirm pricing, which providers carry it, region availability, or a spec sheet — and that gap is exactly what this guide is built to help you navigate.

What to look for

Until a cloud provider publishes its own listing, price, and spec sheet for Vera-based instances, the smartest move is to know what to check for and in what order, rather than treat any number you've seen elsewhere as current.

1. Primary confirmation of the CPU itself. Start with Nvidia's own announcement rather than secondhand summaries — it's the only sourced fact available right now: the CPU is named Vera, it's described as Nvidia's first CPU built for agents, and the headline states it's shipping.

2. A provider's own listing, not a rumor. Treat any specific price, instance type, or region as unverified until you see it on the provider's current pricing page. Announcements from a chip maker rarely ship with a full retail rollout on day one.

3. A published price per hour or per unit. If a page doesn't state a number, don't estimate one — wait for the provider to publish it rather than filling the gap yourself.

4. What workload it's actually built for. Nvidia frames Vera as built specifically for agents; if your workload isn't agentic inference or a closely related pipeline, that framing alone shouldn't be the deciding factor.

5. Availability by region and account tier. New silicon typically rolls out to specific regions or account tiers before general availability, so confirm your own account and region are covered before you plan around it.

Recommended options

There isn't yet a verified, apples-to-apples comparison to make here — the only sourced detail is the announcement itself, and no pricing, spec sheet, or provider lineup for Vera has surfaced in the material used for this guide. Rather than inventing options to fill a table, here's what's actually confirmed versus what still needs verifying:

Best for: buyers who want to track the announcement and act the moment a provider posts real pricing. Bookmark the primary announcement and check your preferred provider's instance list directly rather than a summary of it.

Skip it if: you need agent-workload compute in production this week and can't wait on unconfirmed availability. Build with what's already listed and priced on your provider's current page instead of planning a deployment around a headline.

Common mistakes

The most common mistake is treating a manufacturer's shipping announcement as proof that every cloud provider already has the product listed and priced. A chip vendor's "shipping now" describes the chip leaving the vendor's own pipeline, not that every reseller or cloud has stocked it, priced it, and opened orders.

The second mistake is repeating a price or spec seen on a forum or a secondary write-up as if it came from Nvidia or from a provider directly. If a number isn't on the primary announcement or the provider's own page, it isn't confirmed — treat it as a placeholder until it is.

The third is skipping the step of checking your own account and region before assuming access. New instance types roll out unevenly, and the same account that has access in one region may not have it in another.

Before you act on any number tied to this announcement, See current pricing at the merchant so you're deciding from what's live today, not from what a headline implied. That's the cheapest way to avoid budgeting for compute you can't actually access at the price you expected.

The decision is simple: if you need agent-ready compute now, buy against whatever price and instance type is already published on your provider's page, not against what a headline implies. If you can wait, treat everything except the shipping announcement itself as unverified until a provider posts real numbers.