Microsoft will deploy AMD’s Helios rack-scale AI accelerator ‘at scale’ on Azure – Radeon Instinct MI455X and Epyc Venice power will be available through Redmond’s cloud infrastructure ⁠62%

By Jeffrey Kampman⁠46%

7/20/2026, 6:05:00 AM

BS Summary: This article contains 20 faulty reasoning types, including Framing Effect, Biased Writer Voice, and Optimism Bias, with Attempt to Sell a Product or Service as the most egregious example at 28.1% saturation with 138 hits. Analysis detected 1,045 faulty-reasoning hits from 491 analyzed words, generating a BS Score of 57.2% and a BS Rank of ⁠62% (8,452 of 21,887 articles). This article is worse (more manipulative) than 61.40% of the article peer group.

The demand for AI compute is already insatiable, and it seems only poised to grow in the wake of the introduction of frontier-class open models like Kimi K3 that anybody can potentially fine-tune and serve. 
Against this backdrop, Microsoft and AMD are teaming up to get Redmond more AI FLOPS for both internal and external use. 
The two companies announced this morning that Microsoft will commit to adding AMD's Helios rack-scale AI accelerator in volume to run frontier-model workloads in its own data centers, as well as for Azure AI infrastructure customers and services. 
(Image credit: Microsoft) Photonics and high-speed data movement is the next big AI bottleneck 
The data center cooling state of play 
Massive AI data center buildouts are squeezing energy supplies 
Ultra Ethernet: The data center interconnection of tomorrow 
The partnership makes next-gen AMD AI compute available to Azure customers like AI labs for AI training and inference serving workloads, and it’ll also underpin managed compute for enterprise customers looking to deploy AI workloads through Microsoft Foundry. 
The two companies didn't indicate the exact size of Microsoft's Helios deployment in either watts or dollars, but the commitment would seem to be another major win for AMD as it seeks to grab data center GPU share from Nvidia. 
AMD has struck massive partnerships with OpenAI and Meta in the past year with gigawatts of compute installations and hundreds of billions of dollars potentially hanging in the balance. 
For a quick refresher, the Helios rack-scale accelerator will take the fight to Nvidia’s Vera Rubin NVL72 system when it arrives later this year. 
Helios joins together 72 next-generation Instinct MI455X GPUs with an aggregate of 31.1TB of HBM4 memory capacity across the system. 
Those GPUs offer as much as 1.4 exaFLOPS of FP8 compute and 2.9 exaFLOPS of FP4 for AI models using those OCP AI data types. 
AMD is targeting 260 TB/s of scale-up bandwidth within the rack, on par with Nvidia’s Vera Rubin NVL72 rack-scale system, and 43 TB/s of scale-out bandwidth using UALink over Ethernet, or about twice that of Vera Rubin, although the performance of UALink over Ethernet in practice remains to be seen. 
Microsoft and AMD also announced that Azure will add two new VM series built on AMD's upcoming sixth-gen Epyc Venice CPUs : the HDv2 series for "agentic AI and data pipelines," and the HXv2 for semiconductor design workflows. 
Microsoft will also leverage its existing deployment of AMD Pensando DPUs to integrate that hardware into its Azure Boost offerings to accelerate networking and storage processing operations. 
Tom’s Hardware will be on the ground at AMD’s Advancing AI event this week, where we expect to learn more about AMD’s AI ambitions for the second half of this year and beyond. 
Stay tuned for our coverage from that event. 
Article reasoning-pattern comparisonThis article: 8.1%Jeffrey Kampman: 8.4%Tom's Hardware: 3.8%Confirmation Bias8.1%This article: 10.2%Jeffrey Kampman: 1.7%Tom's Hardware: 2.3%Anchoring Bias10.2%This article: 8.8%Jeffrey Kampman: 2.7%Tom's Hardware: 3.5%Availability Heuristic8.8%This article: 0.0%Jeffrey Kampman: 1.1%Tom's Hardware: 1.1%Representativeness Heuristic0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.5%Hindsight Bias0.0%This article: 5.1%Jeffrey Kampman: 4.2%Tom's Hardware: 3.5%Overconfidence Bias5.1%This article: 22.4%Jeffrey Kampman: 7.7%Tom's Hardware: 9.2%Framing Effect22.4%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.9%Loss Aversion0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.7%Status Quo Bias0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.3%Sunk Cost Effect0.0%This article: 14.3%Jeffrey Kampman: 11.2%Tom's Hardware: 6.0%Optimism Bias14.3%This article: 0.0%Jeffrey Kampman: 1.2%Tom's Hardware: 2.0%Pessimism Bias0.0%This article: 12.8%Jeffrey Kampman: 2.1%Tom's Hardware: 6.5%Negativity Bias12.8%This article: 0.0%Jeffrey Kampman: 0.9%Tom's Hardware: 1.3%Self-Serving Bias0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.6%Fundamental Attribution Error0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.2%Actor-Observer Bias0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.6%In-Group Bias0.0%This article: 0.0%Jeffrey Kampman: 0.6%Tom's Hardware: 0.3%Out-Group Homogeneity Bias0.0%This article: 7.7%Jeffrey Kampman: 3.9%Tom's Hardware: 2.8%Halo Effect7.7%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.0%Horn Effect0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.0%Dunning-Kruger Effect0.0%This article: 6.7%Jeffrey Kampman: 1.8%Tom's Hardware: 2.0%Recency Bias6.7%This article: 7.7%Jeffrey Kampman: 0.7%Tom's Hardware: 0.6%Primacy Effect7.7%This article: 0.0%Jeffrey Kampman: 0.4%Tom's Hardware: 0.1%Blind-Spot Bias0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.2%Ad Hominem0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.2%Straw Man0.0%This article: 2.9%Jeffrey Kampman: 5.6%Tom's Hardware: 5.4%Appeal to Authority2.9%This article: 0.0%Jeffrey Kampman: 0.9%Tom's Hardware: 2.0%False Dilemma0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 1.0%Slippery Slope0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.1%Circular Reasoning0.0%This article: 13.0%Jeffrey Kampman: 4.7%Tom's Hardware: 6.1%Hasty Generalization13.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.3%Red Herring0.0%This article: 0.0%Jeffrey Kampman: 1.3%Tom's Hardware: 1.1%Bandwagon0.0%This article: 7.7%Jeffrey Kampman: 1.8%Tom's Hardware: 3.0%Appeal to Emotion7.7%This article: 0.0%Jeffrey Kampman: 0.7%Tom's Hardware: 0.8%Begging the Question0.0%This article: 7.3%Jeffrey Kampman: 1.4%Tom's Hardware: 3.5%Post Hoc (False Cause)7.3%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.1%Tu Quoque0.0%This article: 8.1%Jeffrey Kampman: 0.7%Tom's Hardware: 0.8%Burden of Proof8.1%This article: 0.0%Jeffrey Kampman: 0.4%Tom's Hardware: 0.3%Appeal to Nature0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.6%Composition/Division0.0%This article: 0.0%Jeffrey Kampman: 0.6%Tom's Hardware: 1.6%Anecdotal0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.0%No True Scotsman0.0%This article: 10.2%Jeffrey Kampman: 2.1%Tom's Hardware: 3.4%Ambiguity (Equivocation)10.2%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.0%Gambler’s Fallacy0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.2%Middle Ground0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.0%Personal Incredulity0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.2%Special Pleading0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.1%Genetic Fallacy0.0%This article: 7.7%Jeffrey Kampman: 2.5%Tom's Hardware: 2.4%Unattributed Quote7.7%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.9%Quote-first Misdirection0.0%This article: 22.2%Jeffrey Kampman: 13.4%Tom's Hardware: 7.1%Biased Writer Voice22.2%This article: 1.6%Jeffrey Kampman: 0.1%Tom's Hardware: 1.3%Indoctrination1.6%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.1%Politically Left Leaning Bias0.0%This article: 0.0%Jeffrey Kampman: 0.0%Tom's Hardware: 0.0%Politically Right Leaning Bias0.0%This article: 28.1%Jeffrey Kampman: 4.7%Tom's Hardware: 3.3%Attempt to Sell a Product or S…28.1%

491 words analyzed.

Speakers

No attributed speakers were identified in this analysis.

Loading…
Loading…
Loading…

Analysis

Hover over highlighted words in the article to view the associated bias or fallacy analysis.