# Clips from the theCUBE Supercomputing 2026: Dell, Vultr, Penguin campaign

Five clips posted for theCUBE Supercomputing 2026: Dell, Vultr, Penguin have 20K views between them. The biggest, let’s cook’s X clip at 6.2K views, pays about $12 at the campaign’s $2.00 per 1,000 views; the typical clip here, at 2.8K views, about $6.

## Every clip the platform lists

- [What if AI models could share memory instead of constantly copying data between GPUs? 🤯  At #SC26, Preet Virk of Marvell explains how shared memory could eliminate expensive KV cache transfers, improve AI inference performance and allow multiple GPUs and AI agents to access the https://t.co/f6eyhgZrzo](https://x.com/jacklegenddd/status/2123309584496431470) on X by let’s cook: 6,203 views, pays about $12
- [AI models are growing so fast that even hundreds of GPUs can't hold all their parameters.  At #SC26, Marvell and Penguin Solutions explain the AI memory bottleneck and why scaling memory is becoming a major challenge for AI infrastructure.  @nysewired @theCUBE  #AI https://t.co/IMpENDrskq](https://x.com/jacklegenddd/status/2123307754412532083) on X by let’s cook: 5,937 views, pays about $12
- [The future of AI depends on more than just powerful technology. Infrastructure is becoming a major part of the conversation.  Ihab Tarazi, SVP &amp; CTO of Dell ISG, shares his perspective at SC26.  @nysewired @theCUBE @ihab https://t.co/qMIL5HHZPr](https://x.com/adam79977/status/2123155203885093156) on X by @Musharaf: 2,832 views, pays about $6
- [AI infrastructure just took an exciting step forward.  At #SC26, John Furrier (@furrier) sits down with Andy Mills of Penguin Solutions to showcase the world's first photonics-based HBM-class memory appliance, developed alongside Marvell.  By bringing fiber optics closer to the https://t.co/NdxCnuyp3E](https://x.com/donsam_samuel/status/2123175620590916056) on X by @finewelding: 2,570 views, pays about $5
- [AI just flipped from training to inference, and most GPUs are still idle. Preet Virk (Marvell) breaks down the real problem: low GPU utilization from data movement not arriving on time. Wasted FLOPS. The bottleneck isn’t more chips it’s getting bits to the GPUs faster. https://t.co/NF5c7YF8fB](https://x.com/greatkllip/status/2123147842189754647) on X: 2,242 views, pays about $4

## Who posted them

- let’s cook: 2 clips, 12,140 views, about $24
- @Musharaf: 1 clip, 2,832 views, about $6
- @finewelding: 1 clip, 2,570 views, about $5

## How this is counted

Content Rewards publishes each campaign’s top clips with their view counts. We read them at every check and keep each one with a working link, so this is the platform’s own list, not every clip ever posted for the campaign. A pay figure applies the campaign’s posted rate for that app to the clip’s views, after its minimum views and payout cap. It is what the rate pays for those views, not a confirmed payment, and the campaign’s remaining budget can cut it. One clip comes without the poster’s handle. Views are as the platform last reported them, 19 Nov, 3:07 pm ET.

Page: https://clipcpm.com/campaigns/reelsbuilder/thecube-supercomputing-2026-dell-vultr-penguin/clips. Campaign: https://clipcpm.com/campaigns/reelsbuilder/thecube-supercomputing-2026-dell-vultr-penguin.
