Selling systems to drive AI training and inference is not as easy as embedding AI into products to sell them. Ask Cisco. $CSCO nextplatform.com/2024/08/15/ai-…

AMD needs system engineers so badly to design future AI and HPC clusters that it is willing to spend $4.9 billion to buy a hyperscale server maker with 1,100 of such people. $AMD $NVDA nextplatform.com/2024/08/19/why…

Supermicro doubled its revenues in fiscal 2024, and it plans to double it again next year. $SMCI $NVDA $AMD nextplatform.com/2024/08/20/sup…

Arista Networks wants its fair share of back-end AI networks, and can sell front-end networks as it ramps the Etherlink products starting now and into production early next year. $ANET $CSCO $AVGO nextplatform.com/2024/08/21/ari…

Andy Bechtolsheim, a guru of both systems and networking, outlines how we boost XPU performance for AI workloads by 100X by 2028 – and how interconnects can keep up. $ANET $NVDA $AMD $INTC nextplatform.com/2024/08/26/bec…

IBM is getting serious about native AI processing and DPUs on its big iron systems. $IBM nextplatform.com/2024/08/27/ibm…

Immediately after the numbers come out, we have exclusive sit-down with Nvidia CFO Colette Kress to talk about planning for explosive growth and how software will be an increasing part of Nvidia’s financial future. $NVDA

How much of Nvidia’s business is driven by software, and how will that change over time? The Next Platform asks Nvidia CFO Colette Kress after Nvidia reported its Q2 F2025 numbers. $NVDA nextplatform.com/2024/08/28/in-…

As long as Nvidia can make Hopper H100 and H200 GPUs, it can sell as many as it can make and the Blackwell delay hurts customers a lot more than it hurts Nvidia. $NVDA nextplatform.com/2024/08/28/nvi…

Leap frog: We do a deep dive on MLPerf inference performance and cost for AMD MI300X and Nvidia H100, H200, and B200 GPUs. $NVDA $AMD nextplatform.com/2024/09/03/the…

TACC fires up Vista, a hybrid CPU-GPU machine based on Nvidia compute engines with 4.1 petaflops of CPU and 40.8 petaflops of GPU performance. $NVDA nextplatform.com/2024/09/04/tac…

Hope springs eternal, and HPE is banking on enterprises buying its AI server wares at a profit as they also buy more traditional servers. $HPE $NVDA $AMD $INTC nextplatform.com/2024/09/05/pro…

Now that AWS has a managed Slurm workload manager for ParallelCluster, moving HPC applications to the cloud just got a whole lot easier. $AMZN nextplatform.com/2024/09/09/aws…

The big AI upstarts are pivoting from training to inference as the commercialization of GenAI requires more compute for a lot less money. $NVDA $AMD $INTC nextplatform.com/2024/09/10/the…

Sometime next year, Oracle will fire up the biggest GPU cluster in the world, which Elon Musk might have rented for xAI but chose not to. $ORCL $NVDA nextplatform.com/2024/09/11/gre…

It is easy to criticize those who are trying to track AI spending and calculate ROI, but it is hard to do a better job. Here’s why. nextplatform.com/2024/09/13/the…

Europe’s first exascale supercomputer, Jupiter, will be contained in a modular datacenter. Here’s why. $NVDA nextplatform.com/2024/09/16/thi…

Intel has to do business with the hyperscalers and cloud builders, but they are frenemies at best right now. $INTC $AMZN nextplatform.com/2024/09/17/the…

The CINECA HPC center in Italy wants to spend $31.3 million to get at least 165 nodes with eight GPUs each to boost the Leonardo supercomputer’s AI performance. That upgrade would need very unrealistic price cuts. $NVDA $AMD nextplatform.com/2024/09/23/a-g…

Intel launches high-end “Granite Rapids” Xeon 6 server CPUs – not too many SKUs, and probably pretty high, but mostly fair, prices. We have to guess because Intel didn’t give out prices. $INTC $AMD nextplatform.com/2024/09/24/int…