惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

D
DataBreaches.Net
V
Vulnerabilities – Threatpost
C
CERT Recently Published Vulnerability Notes
Google DeepMind News
Google DeepMind News
GbyAI
GbyAI
Y
Y Combinator Blog
T
Threatpost
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
Project Zero
Project Zero
Engineering at Meta
Engineering at Meta
MongoDB | Blog
MongoDB | Blog
MyScale Blog
MyScale Blog
Security Latest
Security Latest
T
Threat Research - Cisco Blogs
量子位
I
Intezer
Simon Willison's Weblog
Simon Willison's Weblog
C
Cybersecurity and Infrastructure Security Agency CISA
L
Lohrmann on Cybersecurity
L
LINUX DO - 最新话题
The Register - Security
The Register - Security
T
Tailwind CSS Blog
爱范儿
爱范儿
Google DeepMind News
Google DeepMind News
T
Troy Hunt's Blog
Stack Overflow Blog
Stack Overflow Blog
Cloudbric
Cloudbric
S
Secure Thoughts
The GitHub Blog
The GitHub Blog
T
The Blog of Author Tim Ferriss
L
LangChain Blog
Recorded Future
Recorded Future
小众软件
小众软件
www.infosecurity-magazine.com
www.infosecurity-magazine.com
T
Tor Project blog
人人都是产品经理
人人都是产品经理
F
Full Disclosure
O
OpenAI News
Webroot Blog
Webroot Blog
A
Arctic Wolf
TaoSecurity Blog
TaoSecurity Blog
P
Privacy & Cybersecurity Law Blog
Jina AI
Jina AI
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
Exploit-DB.com RSS Feed
Exploit-DB.com RSS Feed
雷峰网
雷峰网
Microsoft Security Blog
Microsoft Security Blog
H
Heimdal Security Blog
B
Blog RSS Feed
Vercel News
Vercel News

The Next Platform: In-depth coverage of high end computing

Uncle Sam Awards $2 Billion-Plus To Quantum Companies, But Wants A Cut Oak Ridge Starts Weaving Together A Quantum, Classical HPC, And AI System Stack Dell Bulks Up Hardware As AI Infrastructure Shifts To On-Premises Cisco Wins Over AI Customers With Merchant Silicon And Optics With Its IPO Done, Cerebras Can Get Back To Pushing The AI Envelope HPE Throws VM Users A Lifeline, Unifying Containers And VM Management In Cloud Stack OpenAI, Microsoft And Friends Build A Better, More Scalable Ethernet Compute And Memory Price Hikes Drive IT Spending Way Higher Sometimes, Air Is The Only Way For AI Systems To Keep Their Cool Arista Rides AI Scale Out Networks, Moves Into Scale Across, And Awaits Scale Up If You Can Make A Compute Engine, You Can Sell A Compute Engine Cleveland Clinic Simulates Large Proteins With Quantum-Centric Supercomputing Broadcom Helps CPU And XPU Makers Go Vertical With Compute Microsoft Committed To Doubling AI Infrastructure In Two Years Google Is A Full Stack AI Player, And Is Playing Well AWS Will Be An OEM, Just Like Google And Maybe Microsoft New Google Networks Tuned Up For GenAI Inference And Training Microsoft And OpenAI Remain Friends, Are Looking To Hook Up With Others AI-Driven CPU Shortage Saves Intel’s Financial Cookies The GenAI Battle Shifts From Frontier Models To Agentic Platforms With TPU 8, Google Makes GenAI Systems Much Better, Not Just Bigger Cisco Scales Out Quantum Systems With A Quantum Network Switch The Second Time Will Be The IPO Charm For Cerebras Imagine An Army Of AI Minions Handling Incident Response AI Will Soon Drive A Third Of TSMC’s Business Bechtolsheim & Friends Breathe Life Into Pluggable Optics One Last Time How HPC And AI Digital Twins Accelerate Quantum Error Correction The Embrace Of AI In Design Transforms Cadence And Its Customers Nvidia Brings The Power Of Open Source AI Models To Quantum Computing Building The Imperfect Beast For Enterprises, GPUs Need Virtualization As Much As CPUs Ever Did CoreWeave Takes As Much Financial Engineering As It Does Datacenter Design Contemplating Meta’s Homegrown MTIA Compute Engine Roadmap Most Neoclouds, Sovereigns, And Enterprises Will Buy, Not Build, Their AI Stacks Broadcom And Google Benefit Mightily From Anthropic’s Meteoric Growth Rebellions AI Rings Up The Money To Rack Up AI Inference Systems Nvidia Software Pushes MLPerf Inference Benchmarks To New Highs Broadcom Makes Its Pitch To Run Kubernetes On VMware VCF The $2 Billion Nvidia Deal With Marvell Is About A Lot More Than NVLink Fusion Classiq Says Quantum Is On Its Way, But Patience Is Needed Demonstrating The Scientific Usefulness Of Quantum Systems We Need Servers – Lots Of Servers. . . . Arm Comes Full Circle With Homegrown, AI-Tuned Server CPU Riding The Memory Boom And Trying To Avoid The Bust Data Analytics Helps Make The Mighty Lionesses Roar Driving Down The AI System Roadmap With Nvidia The Open Agentic AI World According To Nvidia Nvidia Finally Admits Why It Shelled Out $20 Billion For Groq Nvidia Says OpenClaw Is To Agentic AI What GPT Was To Chattybots IBM Unrolls Blueprint For Quantum-Classical HPC Computing Women Get Data-Driven Health Boost As The FA Tackles Sports Science Four Months Into Its Comeback, Zapata Stakes Its Claim In Quantum Software Eridu Cuts To The AI Networking Chase With High Radix Switch System We Need A Proper AI Inference Benchmark Test How AI Is Boosting Gender Equality In High Performance Racing Custom Compute Engine Biz Growing More Than Marvell Ever Hoped Broadcom May Become The Biggest Counterbalance To Nvidia Ayar Labs Gets $500 Million To Ramp Photonics Into 2028 AI Systems With Cisco Outshift, Agentic AI Is Teed Up For the Internet Of Cognition Nvidia Sees The Light On Silicon Photonics And Maybe Optical Switching AI Servers Finally Dominate Dell’s Systems Business VAST Data: What Controls The Data Is More Important Than What Stores It So Far, Nobody Turns Tokens Into Money Like Nvidia SambaNova Pits Its Engineering Against Nvidia For Agentic AI Some More Game Theory, This Time On The AMD-Meta Platforms Deal AMD Says “Helios” Racks And MI400 Series GPUs On Track For 2H 2026 CPU-Only Compute Still Matters To A Lot Of HPC Centers Taalas Etches AI Models Onto Transistors To Rocket Boost Inference Some Game Theory On That Nvidia-Meta Platforms Partnership AI Eats The World, And Most Of Its Flash Storage The Current AI Networking Wave Will Be A Tsunami Of Money By 2027 The Memory Crunch Pinches Cisco’s Profits Only A Few AI Platforms Can Survive The Greatest AI Show On Earth Cisco Doubles Up The Switch Bandwidth To Take On AI Scale Out And Eventually Scale Up Datacenter Spending Forecast Revised Upwards – Yet Again The Twin Engine Strategy That Propels AWS Is Working Well With GenAI Turbochargers, Google Is Shifting Its Cloud Into A Higher Gear AMD Finally Makes More Money On GPUs Than CPUs In A Quarter Dassault And Nvidia Bring Industrial World Models To Physical AI TACC Explores Mixed Precision And FP64 Emulation For HPC With Horizon Robotics Will Break AI infrastructure: Here's What Comes Next Oracle’s Financing Primes The OpenAI Pump Gartner Takes Another Stab At Forecasting AI Spending Microsoft Is More Dependent On OpenAI Than The Converse Big Blue Poised To Peddle Lots Of On Premises GenAI Microsoft Takes On Other Clouds With “Braga” Maia 200 AI Compute Engines Nvidia’s $2 Billion Investment In CoreWeave Is A Drop In A $250 Billion Bucket Intel Is Still Struggling In The Datacenter, But It Could Get Better Is Nvidia Assembling The Parts For Its Next Inference Platform? TSMC Has No Choice But To Trust The Sunny AI Forecasts Of Its Customers Cerebras Inks Transformative $10 Billion Inference Deal With OpenAI By Decade’s End, AI Will Drive More Than Half Of All Chip Sales Startup Quantum Elements Brings AI, Digital Twins To Quantum Computing D-Wave Makes Gate-Model Power Move With Quantum Circuits Buy Building The Future Of Software In The AI-Native Era Arista Modular Switches Aim At Scale Across Networks, Hit Scale Out, Too NextSilicon Takes Aim At CPUs And GPUs With “Maverick-2” Dataflow Engine How HPC Is Igniting Discoveries In Dinosaur Locomotion – And Beyond Oracle First In Line For AMD “Altair” MI450 GPUs, “Helios” Racks
HPE Works Harder And Smarter To Chase Datacenter Profits
Timothy Prickett Morgan Timothy Prickett Morgan · 2026-03-11 · via The Next Platform: In-depth coverage of high end computing

The compute engine and networking ASIC makers are profiting mightily from the GenAI boom, and so is Taiwan Semiconductor Manufacturing Co, the world’s largest and most important foundry. Everyone else downstream from them has a very tough time squeezing profits from sales. Even if business is up, costs for memory and flash are going up faster, and companies like Hewlett Packard have to work harder to find and select profitable deals. And they have to walk away from deals that do not make sense economically.

Every quarter for many years now, we look at the financials of maybe two dozen key, publicly traded IT suppliers to try to ascertain how their datacenter businesses are doing so we can give you a better sense of what is going on out there as you navigate the churning seas of the GenAI boom and try to negotiate around some rocks below the water line.

Our summary chart of the “core systems” business at Hewlett Packard Enterprise is usually the last thing we show you in our financial analysis of the company, but this time around we are going to start here because it frames the conversation. Take a look:

You can see how the GenAI boom and then the acquisition of Juniper Networks in the past two quarters have helped push up the revenue line for the servers, storage, switching and now routing, security, systems software, and financing that HPE sells into the datacenter. These numbers do not include branch, edge, and campus networking revenues from the newest HPE or other corporate investments.

This particular dataset has only been cast backwards to the time when Antonio Neri took the helm as the company’s chief executive officer, which was after the company bought HPC supplier Silicon Graphics but before it bought its bigger competitor Cray.

This has been a tough row to hoe for Neri & Co, hasn’t it? Kudos to Neri and his predecessors for staying in the game and trying to make some money while building complex systems for some of the most important organizations in the world.

As you can see, the HPE datacenter systems business is now trending along at about $8 billion a quarter or so, and operating income is trending around $1 billion, which sure beats half of that and which explains why, in part, why HPE shelled out $14 billion in cash to buy Juniper. To be more specific, we think this core systems business drove $7.81 billion in Q1 F2026, up 16.2 percent, and that operating income for this datacenter stuff was $997 million, up 63.4 percent. The networking profits are helping, clearly.

Looking ahead, HPE wants to use Juniper to capture more networking revenue. Speaking generally, the AI systems market might grow to $1 trillion a year by 2030 or so, and $200 billion of that might be for networking. Even if you are a pessimist and say only a quarter or a third of that network spending will be done by traditional large enterprises, that is still a hell of a lot more total addressable market than either HPE or Juniper had been chasing with their respective networking businesses over the past decade. The trick will be to get GPU and maybe XPU allocations to build the systems that allow HPE to drive revenues with AI servers and drive profits with AI networking.

That is how Broadcom is doing it with its nascent custom AI systems business, and that is how Marvell is doing it, too.

HPE did not provide revenue figures for AI servers and AI networking in its financial presentation for the first quarter of fiscal 2026, which ended on January 31, which would make life easier, but we can make some pretty good guesses about how traditional servers did and then subtract it out to get AI server revenues. With this methodology, what we can tell you is that AI server sales were not great, even though most of them were to enterprise customers (which is good if you want AI to mainstream) and not to service providers with bigger budgets.

By the way, when HPE says AI servers, it means AI servers and HPC servers, which includes big clusters (and little ones) for simulation and modeling workloads. HPC’s share of the accelerated system pie changes radically from quarter to quarter, as it did for SGI and Cray.

If my model is right, and I obviously think it is pretty close, then AI/HPC server sales – what HPE used to call APU system sales – were down 39.2 percent to $517 million, and that was down from a pretty weak $850 million in AI/HPC server sales in the year ago period and down 40.4 percent sequentially from a pretty weak $867 million in Q4 F2025 ended in October last year. In the trailing twelve months, HPE’s AI/HPC sales are off 14.5 percent to $3.85 billion in fact.

But this may be on purpose as HPE is walking away from bad service provider deals and focusing on more profitable enterprise and sovereign deals. Neri did say on the call with Wall Street analysts that the $5 billion backlog it has for AI/HPC systems is primarily comprised of deals with enterprises and sovereigns.

Our best guess is that traditional ProLiant servers drove $3.72 billion in sales in Q1 F2026, up 8 percent year on year and up 1.9 percent sequentially from a pretty good $3.65 billion in Q4 F2025. Maybe it is a good thing that HPE didn’t do a lot of AI/HPC system sales in the quarter because it would have hurt profits, particularly with DRAM and flash memory prices skyrocketing.

Here how Neri characterized the impact of memory and flash shortages and price spikes on HPE:

“The IT market is facing a sharp acceleration in supply tightness and increasing component costs, most notably in DRAM and NAND. We expect elevated prices to persist well into 2027. We are taking a series of distinct actions to address the current industry dynamics.”

“First, we are focused on securing supply. We have expanded our long-term multiyear agreements with our key silicon and memory partners to secure the capacity needed to meet customer demand. Second, we are protecting our margins. We have adopted an agile pricing posture with price adjustments across the entire portfolio with shorter quote commitment cycles. We have amended our quoting terms with a right to reprice existing orders for commodity cost increases between quoting and shipment. And third, we are proactively communicating with customers and channel partners, providing lead time and cost visibility along with alternative configuration recommendations to shape demand.”

“DRAM and NAND now make up over half of the bill of materials cost of a traditional server, and the share will continue to rise as component costs increases. As a result, we expect higher average unit prices in both our server and storage products. Networking is more insulated, with memory comprising a significantly smaller portion of the bill of materials. Given the supply dynamics, our fiscal 2026 strategy prioritizes higher margin product orders, which have an impact on our AI systems revenue growth rate for the year.”

Here is a full breakdown of the revenues and profits of the new HPE groups and divisions, which were announced this week:

Let’s ponder this new Networking division. Neri said that networking, including HPE and Juniper products, now accounts for nearly 30 percent of the company’s revenue and more than half of the company’s profits. Obviously, over the next two quarters, as the Juniper numbers are added to the financials, it will grow by around 2.5X year on year, but it is not clear how much the combined businesses are actually growing. Basically, this buys HPE two more quarters of easy compares and then we will know how Networking is really performing. (I still don’t know what will happen to the “Rosetta” Slingshot Ethernet interconnect for HPC systems, now that I think on it. . . .)

HPE’s datacenter switching business was up 4.8X to $444 million. HPE had a teeny tiny routing business (probably based on Broadcom chips) that only drove $1 million a year ago, but after the acquisition, it is up 77,900 percent to $780 million in Q1 F2026.

Servers are still the revenue driver, as you can see, and brought in $4.23 billion in the period, but it was down 2.7 percent because of (purposely) sluggish AI/HPC system sales. Storage in the aggregate doesn’t change much at HPE quarter to quarter and year to year, and was up six-tenths of a point to $1.06 billion this time around. Financial services, which also trundles along, was up three-tenths of a point to $876 million in the quarter.

Add it all up, and HPE had 9.3 billion in sales, up 18.4 percent, with operating income of $470 million, up 8.5 percent, and net income of $452 million, off 27.9 percent and bolstered by some tax accounting stuff. The company exited Q1 F2026 with $4.84 billion in cash and $17.71 billion in long term debt, most of which came through the Juniper acquisition.