REDEFINING ENTERPRISE AI
SCROLL TO READ
Jump to:
Redefining the AI Data Center
Shell
01
Shell
Redefining the AI Data Center
AI and other data-intensive digitalization are driving booming demand for servers and data center space — challenging budgets, system performance, and carbon goals. Shell Group’s Houston facility is a masterclass in infrastructure engineering for a denser, cooler, cost-effective future.
Scale a major data center to boost performance for AI, High Performance Computing (HPC) supporting mission-critical digital initiatives. Advance the company’s journey to net-zero emissions by 2050.
Challenge
Unlocking New Performance and Efficiencies
Hover to view
advances at Shell Houston Data Center 2017-2023
Enabled by 1,728 4th Gen AMD EPYC™ 9654 processors on 864 dual-socket systems
Increase in chip power
2017
140 watts
360 watts
2023
Increase in chip power
Increase in rack power
30 kw
2017
Increase in rack power
60 kw
2023
1000 kw
Increase in compute power-per-rack
2017
Up 3.3x
2023
Reduction in # of racks
2017
70%
2023
Cooling Efficiency
Air
2017
Liquid
2023
Immersion
+1000x improvement in heat absorption
David Baldwin, High Performance Computing Manager, Shell IT
“The price, performance, and memory bandwidth of AMD EPYC™ CPUs are perfect for our needs. They are world class… We’ve managed to make our money go further… We’ll continue to push the bounds of what’s possible...”
Operating Benefits of STMicroelectronics Silicon Strategy
Click to view
Performance uplift per core
1st Gen AMD EPYC processor
How Criteo handles far greater volume faster with fewer servers
Hover to view
Enabled by 17,500 single-socket machines running AMD EPYC™ and 8004 series CPUs
2nd Gen AMD EPYC processor
3rd Gen AMD EPYC processor
6
%
Value per core
25
%
Dual-socket
Power reduction per core
30
%
Enabled by 3rd Gen AMD EPYC™ processors
Reduction in data center electricity usage
Enabled by 3rd Gen AMD EPYC™ processors
33
%
2020
Bid volume / day
Server count
Latency
2025
318.7M
5.1B+
2020
60,000
30,000
2025
2020
100ms
-20ms
2025
MAJOR PERFORMANCE AND ENERGY GAINS END-TO-END
Click to reveal
AI Laptops
Data Center / Cloud
Workstations
On-premises Server
Using 7,500 high-performance laptops to run and share 3D models between engineers, designers, and remote field technicians at solar and wind farms significantly reduces power use and emissions.
AI Laptops
25
Enabled by AMD Ryzen 7 7840HS processors
Power reduction per laptop
%
360
tons co2 / year
Emissions reduction
Enabled by AMD Ryzen 7 7840HS processors
Accelerating demanding simulation, modeling, and CAD workflows significantly speeds up wind farm viability analysis.
Data Center / Cloud
3
YEARS
Time to complete wind farm viability analysis
Enabled by AWS with 4th Gen AMD EPYC™ processors
6
MONTHS
Hover to view
Key to design of electrical substations and the entire electrical delivery pipeline — from wind farm to grid — regulatory simulations now run in half the time and support twice the complexity, thanks to increased compute power.
Hover to view
On-premises Server
Regulatory Simulation Performance: Speed and Scale Gains
Enabled by 3rd Gen AMD EPYC processors
60-80
mins
8-32 variables
20-30
mins
64 variables
Conduct image analysis of drone inspections of railroad tracks.
Workstations
8-12
hours
Processing
Enabled by AMD Ryzen 9, Threadripper™ processors
60
mins
CAD simulation
Enabled by AMD Ryzen 9, Threadripper™ processors
Hover to view
20
mins
4
hours
Ready to lead your enterprise into the future of AI today?
Start here
1st Gen AMD EPYC processor
12
%
2nd Gen AMD EPYC processor
25
%
3rd Gen AMD EPYC processor
Single socket
Single socket
30
%
Dual-socket
Latency
Server count
Bid volume / day
Latency
Server count
Bid volume / day
Scale AI faster
Modernize infrastructure for AI faster
Run AI workloads faster
Unified Infrastructure from Edge to Core
As AI moves into wider deployment, organizations face increasing pressure to build a solid foundation that can accelerate AI’s business impact.
Leading companies worldwide are adopting a unified, end-to-end infrastructure strategy. They’re rewriting the rules on cost, performance and energy consumption at every layer of the enterprise stack.
Keep scrolling to see how bold initiatives are powering AI growth at Fortune 2000 companies eager to:
R&D at Enterprise Velocity
02
STMicroelectronics
RealTime AI at Cloud Scale
03
Criteo
Transforming End-to-End Performance
04
Acciona
Increase CPU density with 165,888 processor cores packed into immersion-cooled racks running at 100kW each, delivering best-in-class performance and improving critical power usage effectiveness (PUE).
Solution
STMicroelectronics
R&D at Enterprise Velocity
The new ROI in AI infrastructure isn’t just speed — it’s speed per watt. STMicro’s modernized R&D data center improves CAD simulation throughput by 25%, unlocks real-time design feedback loops, and cuts electricity use by 33%. Result: faster, economical delivery of new chips to customers of Europe’s largest semiconductor company.
As part of regular 27-month server refresh, deploy a fleet of 550 core-dense HPE servers running 1,000 AMD EPYC™ processors, bursting as needed to Azure cloud.
Solution
Boost data center performance to keep pace with market velocity, growing complexity in design and manufacturing, and shrinking chips. (Halving the size of transistors requires 4x processing power and data storage.)
Challenge
“We start with the most important thing for us in the server, which is the CPU… We always need the fastest processor with the most cores… Better performance means faster time to market... We’re able to design a chip faster and run more designs in parallel than before.”
Olivier Joubert, Senior Infrastructure Architect, STMicroelectronics
Criteo
Real-Time AI at Cloud Scale
The Paris-based digital global ad firm now handles a 16x increase in online bids from 100 countries in less than 20 milliseconds each — with half the number of servers used in 2020 and no performance penalty.
Refresh and modernize servers in seven online and offline data centers around the world supporting compute, storage and AI for 100 countries.
Solution
Handle an ever-increasing volume of real-time requests for programmatic display advertising, now exceeding millions per second.
Challenge
“We display more than 5.1 billion ads per day. We have roughly six times the number of products on the Amazon Marketplace. The criteria in the selection process are all about raw performance plus power consumption.”
Matthieu Blumberg, Senior Vice President of Engineering Infrastructure, Criteo
Acciona
Transforming End-to-End Performance
What happens when your business, compute and sustainability strategy converge? Big gains for each at Acciona, a multinational public infrastructure company.
The firm’s Australia Region supercharged edge-to-cloud performance, cutting simulation times by 66%, power consumption by 50% while accelerating project delivery times.
Mission-critical upgrades of upgrades of high-powered, energy efficient laptops, workstations, servers and cloud services deployed in the field, offices, data centers and onsite computing at remote wind and solar farms.
Solution
Cost-effectively speed up and transform critical design, build, maintenance workflows supporting large, sustainable public infrastructure and renewable energy projects.
Challenge
“AWS AMD EPYC™ CPU Compute, network bandwidth and storage performance are the important things for cloud, and [AWS AMD EPYC CPU] excelled at all three.”
Marc Mendis, ANZ Cloud Services Architect, ACCIONA
AMD
END-TO-END FOUNDATIONS FOR ENTERPRISE AI
AMD is emerging as the rare vendor offering a full portfolio spanning all parts of enterprise infrastructure. From the data center and AI PCs, to software, networking, and cloud partnerships, AMD helps IT teams expanding mission-critical AI workloads balance:
Performance and energy efficiency
Optimized production and innovation
Tap the blocks below to learn more.
Tap the buttons below.
Increase in compute power-per-rack
Reduction in # of racks
Cooling Efficiency
Air