With the rush to capitalize on the gen AI boom, datacenters have never been hotter. But before signing that multi-billion dollar purchase order on GPUs, Cadence Systems suggests using a few of them to simulate whether that fancy new bit barn of yours can actually handle the heat.

At the AI Infra Summit this week, simulation savant Cadence [1]added Nvidia's biggest iron, the GB200 NVL72 Superpod, to its digital twin fire. At roughly a megawatt, the compute cluster is among the GPU giant's most complex compute platforms to date.

Each Superpod [2]consists of eight 120 kilowatt NVL72 racks containing more than 500 Blackwell GPUs and 288 Grace CPUs capable of churning out a combined 11.5 exaFLOPS of the lowest-precision compute money can buy.

[3]

However, getting the most out of the systems requires a datacenter specifically designed not only to handle the intense thermal load that comes with packing a megawatt of compute into eight racks, but one whose power draw can jump from idle to 100 percent and back in a fraction of a second.

[4]

[5]

The last thing bit barn operators want to do is find out the hard way that they've under-specced their facilities and can't fully utilize the machines their customers have borrowed billions to buy.

Instead, Cadence suggests using a few spare GPUs to fire up a game of Datacenter Tycoon and see how those plans play out in the digital world before hitting send on that multi-billion dollar purchase order.

[6]

After years of peddling his Omniverse digital twin and data visualization tech, it seems that Nvidia CEO Jensen Huang has finally found the platform's killer app: Designing better bit barns – or AI factories as he likes to call them. Of course, Nvidia would find a way to use GPUs to gamify the process of building datacenters so it can sell more GPUs.

Nvidia has been working with Cadence to build out this capability since March, but, only this week, added support for Nvidia's most sought-after GPU system.

Cadence's Reality Digital Twin Platform — we're afraid Datacenter Tycoon isn't what it's actually called, but it sort of does work that way — is an infrastructure visualization platform that leans heavily on Nvidia's Omniverse tech. The software vendor claims that users can drag and drop datacenter components, like racks, coolant distribution units, and power delivery shelves, around in a virtual environment.

[7]

It includes 3D models of more than 14,000 components from more than 750 vendors, which enables customers to build and run physics sims like computational fluid dynamics to predict power, space, and cooling demand.

[8]Oracle boasts $455B backlog from AI boom, but not all its new friends will live to pay up

[9]Nvidia's context-optimized Rubin CPX GPUs were inevitable

[10]French datacenter biz signs 12-year nuclear pact with EDF

[11]Bring your own brain? Why local LLMs are taking off

And if Cadence doesn't have a model of a part you'd like to deploy, they'll make one for you, because the software required to run these sims presumably ain't cheap either.

Once the purchase order has been sent that doesn't mean that the game — or rather digital twin — is over. Cadence contends that customers can continue to benefit from the platform by using it to simulate infrastructure changes and failure states while optimizing performance.

We imagine this capability will be quite useful when Nvidia's 600kW Kyber racks start [12]shipping in 2027, and 120kW racks start looking reasonable by comparison. ®

Get our [13]Tech Resources



[1] https://www.cadence.com/en_US/home/company/newsroom/press-releases/pr/2025/cadence-expands-digital-twin-platform-library-with-nvidia-dgx.html

[2] https://www.theregister.com/2024/03/18/nvidia_turns_up_the_ai/?td=rt-9c

[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aMH09E2sOwwjdIpMmXawRQAAAAM&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0

[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aMH09E2sOwwjdIpMmXawRQAAAAM&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aMH09E2sOwwjdIpMmXawRQAAAAM&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aMH09E2sOwwjdIpMmXawRQAAAAM&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aMH09E2sOwwjdIpMmXawRQAAAAM&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[8] https://www.theregister.com/2025/09/10/oracle_cloud_llm_cash/

[9] https://www.theregister.com/2025/09/10/nvidia_rubin_cpx/

[10] https://www.theregister.com/2025/09/08/data4_edf_nuclear_deal/

[11] https://www.theregister.com/2025/08/31/local_llm_opinion_column/

[12] https://www.theregister.com/2025/03/19/nvidia_charts_course_for_600kw/

[13] https://whitepapers.theregister.com/