The Next Platform

You don’t need to wait for Ultra Ethernet to build a better Ethernet, and Microsoft and OpenAI proved it. $MSFT $NVDA $AMD $AVGO nextplatform.com/connect/2026/0…

How – and why – Microsoft will double its datacenter capacity in two years. $MSFT nextplatform.com/cloud/2026/05/…

@dasoulmade1 Godzilla. Sheesh.

Apparently if you’re right-brained you see a rabbit, if you’re left-brained you see a turtle. What do you see?

Google Cloud is now the same size Google itself was a decade ago, and it is 25 percent more profitable. $GOOG nextplatform.com/cloud/2026/05/…

The Annapurna Labs chip business at Amazon has an annualized revenue run rate of $20 billion. Amazing. $AMZN nextplatform.com/cloud/2026/04/…

The Boardfly interconnect topology and Virgo datacenter network make the forthcoming Google TPU 8 AI accelerators do a lot more work that prior TPUs. $GOOG $NVDA nextplatform.com/connect/2026/0…

Our first pass deep dive on Google new Sunfish TPU 8t and Zebrafish TPU 8i. $GOOG nextplatform.com/compute/2026/0…

History repeats. Instead of scaling up quantum computers, maybe we need to scale them out. $CSCO nextplatform.com/connect/2026/0…

Why does Cerebras need so much money? Because OpenAI, and maybe AWS, needs so much iron. $CRBS $AMZN nextplatform.com/compute/2026/0…

AI compute and networking plus HBM controllers will soon drive a third of TSMC’s revenues. And it won’t be long before it is half. $TSM nextplatform.com/compute/2026/0…

Don’t count pluggable optics out just yet in the datacenter. XPO looks like the last hurrah before co-packaged optics takes over. $ANET $NVDA $AVGO nextplatform.com/connect/2026/0…

Two shiny new AI models: One you can’t have and the other you can’t have open source. $META nextplatform.com/ai/2026/04/13/…

I finally had time to circle back to the Meta MTIA AI compute engine roadmap. $META $NVDA $GOOG $AMD nextplatform.com/compute/2026/0…

Nvidia’s investment in Marvell could be a lot more interesting than it looks. $NVDA $MRVL nextplatform.com/connect/2026/0…

Selling $125 billion of servers in a quarter is nothing special. It’s the new normal. $DELL $HPE $NVDA $ARM $AMD nextplatform.com/compute/2026/0…

A deep dive into Arm’s first full-fledged datacenter CPU. $ARM $NVDA $TSMC $AMD $INTC nextplatform.com/compute/2026/0…

Nvidia has spiced up its 2028 AI hardware roadmap with reveals about the future Rosa-Feynman systems. $NVDA nextplatform.com/compute/2026/0…

@Matt_Pinner Humanity. <Smile>

In your opinion, what's the biggest threat to humanity?

Agentic AI will require all kinds of compute engines, including CPUs, GPUs, DPUs, and now LPUs. $NVDA $GOOG $AMD $INTC nextplatform.com/ai/2026/03/17/…

Is AI networking upstart Eridu working on a high radix, waferscale switch ASIC? Maybe. Just maybe. $NVDA $AMD $AVGO $MRVL nextplatform.com/connect/2026/0…