1 of 27

Miguel Soto

Cloud Solutions Architect, LATAM

Intel Xeon 6

The future of AI Powered Cloud

1

2 of 27

2

“What do all of these things have in common?” (Pause for audience guesses: AI, GenAI, cloud, apps…)
It’s not just that they’re intelligent or transformative.
The real answer?
They all run on CPUs—specifically, Intel Xeon.
From GenAI inference to traditional ML, from cloud-native platforms to enterprise apps—CPUs are quietly powering the workloads that define modern business.
As AI scales across environments, compute infrastructure must adapt—and accelerated CPUs are delivering the flexibility and performance enterprises need.

(Top Left) Bridging the Radiology Gap: Xeon Speeds Cancer Care Amid Staff Shortages: https://newsroom.intel.com/artificial-intelligence/bridging-the-radiology-gap-xeon-speeds-cancer-care-amid-staff-shortages
(Bottom Right) WaitTime Crowd Management Platform: https://www.intel.com/content/www/us/en/content-details/846106/waittime-crowd-management-platform.html
(Top Right) Netflix Boosts Seamless Streaming Experiences: https://www.intel.com/content/www/us/en/customer-spotlight/stories/netflix-streaming-customer-story.html
(Bottom Left) Audi Precision Manufacturing: https://www.intel.com/content/www/us/en/customer-spotlight/stories/audi-automated-factory.html#:~:text=Figure%201.,and%20to%20other%20use%20cases.%E2%80%9D
(ALT) Shell’s Immersion Cooling Advances Sustainability: https://www.intel.com/content/www/us/en/customer-spotlight/stories/shell-customer-story.html

3 of 27

3

The majority of enterprise �AI projects run on �Intel^® Xeon^®, built for scalable, �general-purpose AI workloads.

3

4 of 27

4

GenAI is shifting from GPU-heavy LLM training �to smaller, more targeted models, inference �optimization, and agentic AI.

And, GenAI is only part of the enterprise AI story. �For decades, enterprises have relied on general-purpose compute to power AI – from data analytics and machine learning to forecasting and fraud detection.

4

5 of 27

Customer service teams leverage churn prediction models to proactively identify and retain high-value clients.

Operations teams optimize supply chain decisions using predictive analytics to improve inventory accuracy and reduce delays.

The finance team uses continuous fraud detection to monitor all transactions in real-time and trigger instant alerts.

Product developers accelerate designs with rapid prototyping and GenAI-powered simulation tools.

5

Enterprise AI, in reality, looks like this:

5

The reason for this shift is simple: your business isn’t one-dimensional - and neither is your AI strategy.
Every department - from product development to finance to operations- can benefit from AI, but each has different needs, constraints, and compute profiles.
Picture this:

Customer service teams leverage churn prediction models to proactively identify and retain high-value clients.
Operations teams optimize supply chain decisions using predictive analytics to improve inventory accuracy and reduce delays.
The finance team uses continuous fraud detection to monitor all transactions in real-time and trigger instant alerts.
Product developers accelerate designs with rapid prototyping and GenAI-powered simulation tools.

Each of these scenarios is powered by AI - but each demands something different from the infrastructure.
That’s why flexibility matters.
CPUs provide the adaptability needed to support a wide range of workloads—without over-engineering or overspending.
In enterprise AI, success comes from matching the right compute to the right workload—and that’s where accelerated CPU platforms like Xeon 6 shine.

6 of 27

6

MATURE AI

EMERGING AI

Data & Feature Engineering

Classical Machine Learning Training & Inference

Deep Learning Inference

Generative AI Fine-Tuning / Training

Generative AI Inference

Agentic AI Orchestration

Edge & Embedded AI

Ingestion, transformation, orchestration, vectorization

Native execution; optimal for training and inference; ensemble models, batch processing

Small/mid-size DL models, transformer inference, real-time and batch execution

Multi-GPU orchestration, memory/I/O coordination, fallback compute

Executes small/mid-size model inference with low-latency response; orchestrates RAG pipelines and optimizes MoE routing

Task routing, tool execution, hybrid model coordination

Orchestrates low-latency GenAI inference at the edge, with fallback compute, RAG execution, and agentic coordination.

CPU-INVOLVED?

AI WORKLOAD CATEGORIES

CPU ROLE

PROOF POINTS

CPUs handle ~70% of �the AI pipeline workload — including ingestion, transformation, and vectorization

CPUs have powered Mature AI for 30+ years

SLMs and Agentic AI expected to drive 40%+ of enterprise AI deployments by 2027

Inference will dominate 80% of AI cycles by 2028

6

Let’s step back and look at the full enterprise AI landscape—from traditional ML to GenAI and Agentic AI. Across this spectrum, CPUs aren’t just relevant—they’re essential.
In the data pipeline, CPUs handle the majority of tasks—ingestion, transformation, vectorization—making them the backbone of AI readiness.
For mature AI, which still accounts for over 60% of enterprise AI spend, CPUs deliver scalable, reliable compute for ML, analytics, and forecasting.
As we move into emerging AI, accelerated CPUs are optimized for emerging workloads like SLM inference and agentic orchestration—where modularity, logic, and responsiveness matter most.
And in GPU-based systems, the host CPU plays a critical role. It’s the orchestrator—feeding data to the GPU, managing memory, coordinating I/O, and ensuring the entire system runs efficiently. Without a capable host CPU, even the most powerful GPU sits idle.
CPUs are not just part of the AI stack—they are the foundation.
Whether powering mature AI, orchestrating GenAI deployments, or enabling agentic workflows, accelerated CPUs—especially Intel Xeon —are vital across the entire enterprise AI environment.
They deliver the flexibility, scalability, and orchestration needed to support mixed workloads, hybrid deployments, and real business outcomes.
In a world of diverse AI demands, the CPU is the constant.

7 of 27

7

The foundation is already there. Your business, your developers, �your operations have operated in this CPU environment for 30 years.

CPUs provide flexibility across your enterprise workloads and deployment environments.

CPUs are still the ideal compute for mature AI and AI data architectures.

CPUs are required in accelerator-based AI systems to serve as the central orchestrator.

CPUs are foundational to secure AI infrastructure, enabling trusted execution and enterprise-grade data protection.

Because…

Why CPUs now?

CPUs are optimized for emerging AI workloads like SLM inference and Agentic AI.

7

So Why CPUs Now?

Starts with trust and familiarity.

Your business, your developers, your operations have operated in a CPU environment for 30 years. That foundation isn’t just legacy—it’s leverage. It’s what makes CPUs the most pragmatic path forward for scaling AI.

Flexibility across environments.

CPUs deliver the agility to run diverse workloads across cloud, edge, and on-prem. They’re deeply integrated into enterprise infrastructure—trusted, scalable, and ready to support hybrid deployment models.

Ideal for traditional AI and data-centric architectures.

From classical ML to data preprocessing and model serving, CPUs remain the optimal compute for mature AI workloads. They’re tuned for the data-heavy, latency-sensitive tasks that underpin enterprise AI.

Central orchestrator in accelerator-based systems.

Even in GPU-heavy environments, CPUs are indispensable. They manage data flow, coordinate compute, and anchor orchestration. Without CPUs, accelerators don’t scale—they stall.

Security as a strategic enabler.

CPUs are foundational to secure AI infrastructure. They enable trusted execution environments, enforce enterprise-grade data protection, and support zero-trust architectures. In a landscape of rising ransomware and AI-targeted threats, CPUs are the anchor of secure deployment.

Optimized for emerging workloads.

As we move into SLM inference, Agentic AI, and multi-modal orchestration, CPUs are showing up where flexibility and control matter most. They’re enabling the next wave of AI—not just supporting it.

Bottom line: CPUs are already everywhere. They’re not just part of your infrastructure—they are your infrastructure. And that makes them the most strategic, scalable, and secure path forward for enterprise AI.

8 of 27

8

With CPUs

the future is more accessible than you think.

8

9 of 27

9

10 of 27

10

Why you choose our accelerated CPUs:

Trusted Compute Foundation

AI-Optimized Architecture

Deployment & Workload Flexibility

10

11 of 27

11

TRUSTED COMPUTE FOUNDATION

When we say Trusted Compute Foundation, we’re talking about a deep commitment to keeping your business safe, secure, and resilient - right from the heart of your infrastructure.

With Intel Xeon, security starts inside the chip itself. Technologies like SGX and TDX enable confidential computing and Zero Trust architectures at the hardware level - protecting data, applications, and memory even while in use.

Xeon is a proven foundation - trusted across enterprise workloads for decades. It integrates seamlessly into your existing environments, so you can modernize securely without retraining teams or rearchitecting systems.

100% of the Intel processor vulnerabilities addressed in 2024 were discovered through internal security research.

Intel scored 82.2, ranking #1 across the silicon industry for product security assurance maturity.

Intel reported 4.4x fewer firmware vulnerabilities in root-of-trust and 1.8x fewer in confidential computing technologies than AMD.

^100%

#1

4.4x

11

When we talk about Intel Xeon’s Trusted Compute Foundation, we’re talking about more than just security - we’re talking about resilience, trust, and readiness for AI-driven enterprise transformation.
Security isn’t just software layered on top. With Xeon, it starts at the silicon level. That means your applications, data, and memory are protected from the inside out - built-in defenses that operate below the OS, shielding you from threats that software alone can’t catch.
For example, Xeon’s confidential computing capabilities protect data even while it’s being processed - so sensitive information stays encrypted and isolated, even in use. And with hardware-based boot integrity, Xeon verifies your system software before it runs, ensuring only trusted code executes.
This foundation is purpose-built for Zero Trust architectures - where no user, device, or workload is trusted by default. Xeon helps enforce that model at the hardware layer, giving you visibility and control across your infrastructure.
And it’s not disruptive. Xeon is the backbone of enterprise compute, trusted for decades. It integrates seamlessly with your existing tools and workflows, so modernization doesn’t mean starting over - it means accelerating securely.
Why does this matter? Because AI is reshaping the threat landscape. In 2024, the global average cost of a data breach reached $4.45 million, according to IBM. As AI workloads grow, so do the risks. Xeon gives you the hardware-level protection needed to run AI safely - whether on-prem, in the cloud, or at the edge.
And Intel doesn’t just respond to threats - it anticipates them. In 2024, 100% of processor vulnerabilities were discovered internally by Intel’s own security researchers. That’s the kind of proactive assurance that sets Xeon apart.
Intel is consistently ranked as a cybersecurity leader. With Xeon, you’re building on a foundation that’s proven, trusted, and ready for what’s next.

Sources:

Intel: https://www.intel.com/content/www/us/en/content-details/846149/2024-intel-product-security-report.html
ABI Research: https://go.abiresearch.com/lp-embracing-security-as-a-core-component-of-the-tech-you-buy
IBM Data Breach: https://www.ibm.com/reports/data-breach

12 of 27

Xeon 6 handles up to 69,000 concurrent queries, up to 3.2× higher concurrent prompts than AMD EPYC 9965

69k

Xeon 6 with AMX delivers up to:

2.59× higher vector search throughput for RAG systems vs AMD EPYC 9575F

1.93× higher DLRM performance vs AMD EPYC 9654

1.85× faster BERT-Large inference vs AMD EPYC 9654

17× faster ResNet-50 batch inference vs AMD EPYC 9654

12

AI OPTIMIZED INFRASTRUCTURE

When we say AI Optimized infrastructure, we mean giving your business the tools it needs to run enterprise AI smarter, faster, and easier – on the Xeon platform it already trusts.

Xeon accelerates AI workloads with built-in instruction sets that deliver intelligent performance out of the box, without changing your apps or retraining your teams.

Integrated accelerators handle data movement, analytics, and security - streamlining the AI pipeline and freeing up compute for what matters most.

And with high core density, advanced memory, and energy-efficient design, Xeon scales to support any AI workload - from single models to enterprise-wide deployments.

12

When we say AI-Optimized Infrastructure, we’re talking about giving your business the power to run enterprise AI smarter, faster, and easier — on the Xeon platform it already trusts.
Let’s start with AI instruction sets. These are specialized capabilities built directly into the Xeon chip that accelerate complex AI tasks — like language understanding, image recognition, and recommendation engines. Because they work with the software your teams already use, there’s no need to rewrite applications or retrain staff. You get performance gains without disruption.
Next, integrated accelerators. Xeon 6 includes built-in engines that handle routine tasks like data movement, analytics, and security. These accelerators take pressure off the CPU, allowing your AI pipelines to run faster and more efficiently — especially under heavy workloads.
And finally, scalable performance. Xeon is built for the enterprise — not just one project or department. With high core density, advanced memory, and energy-efficient design, it scales to support your AI workloads. Whether you're deploying a single model or running hundreds across a cluster, Xeon adapts to meet your needs — reliably and sustainably.
In short, AI-Optimized Infrastructure means Xeon is ready to accelerate your business from day one — with no compatibility issues, no rework, and no compromise.

Sources:

Concurrent prompts: https://community.intel.com/t5/Blogs/Tech-Innovation/Artificial-Intelligence-AI/Accelerating-vLLM-Inference-Intel-Xeon-6-Processor-Advantage/post/1717082
Intel official processor claims: https://edc.intel.com/content/www/us/en/products/performance/benchmarks/processors/
Performance and TCO results based on Intel internal testing and modeling using the Intel Xeon Processor Advisor Tool. See Intel.com/PerformanceIndex and XeonProcessorAdvisor.intel.com for configuration details. Your results may vary.
Common Conditions Across All Benchmarks:

Software Stack:

PyTorch 2.4.0
Intel Extension for PyTorch (IPEX) 2.4.0
OneDNN v3.4.2

Memory Configuration:

Xeon 6: MRDIMM at 8800 MT/s
AMD EPYC: DDR5 at 6400 MT/s

Acceleration:

Intel AMX used for INT8 acceleration across all workloads

Normalization:

All results normalized to AMD EPYC performance = 1.0

Test Environment:

Ubuntu 24.04 LTS
2x NVIDIA H100 NVL PCIe GPUs
Xeon 6776P and 6980P used as host CPUs

Individual Benchmark Qualifiers:

2.59× Vector Search Throughput

Workload: Wiki-45M dataset
Method: Intel SVS IVF BF16
Compared To: AMD EPYC 9575F
Xeon SKU: 6776P (64C)
AMD SKU: 9575F (64C)
Library: robin-map, fmtlib, eve, tomlplusplus, pybind11, MKL, spdlog, catch2, GSL
Test Date: March 2025

1.93× DLRM Performance

Workload: Criteo Terabyte dataset
Precision: INT8
Compared To: AMD EPYC 9654
Xeon SKU: 6980P (128C)
Batch Size: Optimized for throughput
Test Date: April 2025

1.85× BERT-Large Inference

Workload: SQuAD1.1
Precision: INT8
Compared To: AMD EPYC 9654
Xeon SKU: 6980P (128C)
Batch Size: Optimized for throughput
Test Date: April 2025

17× ResNet-50 Batch Inference

Workload: ImageNet
Precision: INT8
Compared To: AMD EPYC 9654
Xeon SKU: 6980P (128C)
Batch Size: Optimized for throughput
Test Date: April 2025

13 of 27

Xeon 6 delivers up to 50% lower TCO and stronger performance than AMD EPYC 9005 - across diverse enterprise workloads.

Supports predictable scaling and operational autonomy across deployment models.

Intel–NVIDIA collaboration aligns �x86 stack with CUDA architecture — simplifying deployment across GPU-accelerated AI workloads.

50^%

x86

13

DEPLOYMENT & WORKLOAD FLEXIBILITY

When we say Deployment & Workload Flexibility, we mean �it’s built to handle any business computing challenge - wherever and however you need it.

Xeon supports diverse workloads on a single platform, so you don’t need separate systems for different jobs.

Deployments are simplified and future-ready thanks to Xeon’s ability to run in public, private, sovereign, and edge environments - from data centers to retail stores to remote sites.

Its open ecosystem works with leading software stacks and avoids vendor lock-in, making it easy to launch new solutions �and reduce investment risk as your business evolves.

13

When we say Deployment & Workload Flexibility, we’re talking about infrastructure that’s built to meet business demands wherever and however they arise.
With Intel Xeon, organizations can run diverse workloads—AI, analytics, traditional compute—on a single platform. That means fewer silos, simpler operations, and more agility across environments.
Whether you're deploying in public cloud, private cloud, sovereign environments, or at the edge—like retail stores or remote sites—Xeon adapts. It’s designed to support predictable scaling and operational autonomy, no matter the deployment model.
And because Xeon is built on an open ecosystem, it works seamlessly with leading software stacks and avoids vendor lock-in. That gives clients the freedom to launch new solutions without being boxed into proprietary infrastructure decisions.
Intel’s collaboration with ecosystem partners—including NVIDIA—ensures Xeon is optimized for hybrid environments and GPU-accelerated workloads, helping simplify deployment and integration across AI use cases.

Sources:

Performance and TCO results based on Intel internal testing and modeling using the Intel Xeon Processor Advisor Tool. See Intel.com/PerformanceIndex and XeonProcessorAdvisor.intel.com for configuration details. Your results may vary.

14 of 27

Matching GPU Price Performance Using Amazon Instances With Intel® Xeon® Processors

Storm Reply helps its customers deploy large language models (LLMs) and Generative AI solutions.
Needed a cost-efficient, high-availability hosting environment to build its LLM-based solution to serve a major company in the energy sector.
A solution developed for the Amazon C7i-family (shared with M7i and R7i) supported by 4th Gen Intel® Xeon® Scalable processors, Intel libraries, and Intel’s open GenAI framework proved an ideal hosting environment for Storm Reply’s LLM workloads.
LLM inference on instances with Intel Xeon Scalable processors was on par with GPU instance price performance. Intel libraries also provided an average response time of 92 seconds, contrasting the 485 seconds required without the Intel library.1

1 For more complete information about performance and benchmark results, visit https://www.intel.com/content/www/us/en/customer-spotlight/stories/storm-reply-customer-story.html

Products and Solutions 4th Gen Intel® Xeon® Scalable Processors oneAPI Toolkit Intel® Extension for Pytorch	Industry IT Services & IT Consulting	Organization Size 201-500	Country Italy	Partners AWS	Learn more Case Study

14

Among many other services, Storm Reply helps its customers deploy large language models (LLMs) and Generative AI solutions. Storm Reply needed a cost-efficient, high-availability hosting environment to build its LLM-based solution to serve a major company in the energy sector. After a thorough evaluation, a solution developed for the Amazon C7i-family (shared with M7i and R7i) supported by 4th Gen Intel® Xeon® Scalable processors, Intel libraries, and Intel’s open GenAI framework proved an ideal hosting environment for Storm Reply’s LLM workloads.

After optimization, Storm Reply determined that LLM inference on instances with Intel Xeon Scalable processors was on par with GPU instance price performance. Intel libraries also provided a significant benefit. Storm Reply’s testing found that the same machine (running Llama 2-13b in bf16 on the same set of questions and same parameters) had an average response time of 92 seconds, contrasting the 485 seconds required without the Intel library.1

15 of 27

Solving problems �from AI/ML and Analytics �to Database and HPC

15

16 of 27

Achieve Cost-Performance for the Workloads that Matter�with new AWS 8^th Gen EC2 Instances powered by custom Intel® Xeon® 6 CPUs

16

Best Cost-Performance with new AWS 8^th Gen Instances powered by Xeon 6 CPUs

**Amazon Elastic Compute Cloud (Amazon EC2) M8i and M8i-flex�instances are the next-generation general-purpose instances powered by custom to AWS, Intel Xeon 6 processor. Amazon EC2 M8i and M8i-flex deliver the highest performance and fastest memory bandwidth among comparable Intel processors in the cloud. The M8i and M8i-flex instances offer up to 15% better price-performance, and 2.5x more memory bandwidth compared to previous generation Intel-based instances. They deliver 20% better performance than M7i and M7i-flex instances, with even higher gains for specific workloads. They are up to 30% faster for PostgreSQL databases, up to 60% faster for NGINX web applications, and up to 40% faster for AI deep learning recommendation models compared to M7i and M7i-flex.

https://aws.amazon.com/ec2/instance-types/m8i/

**Amazon Elastic Compute Cloud (Amazon EC2) R8i and R8i-flex instances are next-generation memory optimized instances powered by custom Intel Xeon 6 processors, available only on AWS. Amazon EC2 R8i and R8i-flex deliver the highest performance and fastest memory bandwidth among comparable Intel processors in the cloud. The R8i and R8i-flex instances offer up to 15% better price-performance, and 2.5x more memory bandwidth compared to previous generation Intel-based instances. They deliver 20% better performance than R7i instances, with even higher gains for specific workloads. They are up to 30% faster for PostgreSQL databases, up to 60% faster for NGINX web applications, and up to 40% faster for AI deep learning recommendation models compared to R7i.

https://aws.amazon.com/ec2/instance-types/r8i/

**The C8i and C8i-flex instances offer up to 15 percent better price-performance, and 2.5 times more memory bandwidth compared to C7i and C7i-flex instances. The C8i and C8i-flex instances are up to 60 percent faster for NGINX web applications, up to 40 percent faster for AI deep learning recommendation models, and 35 percent faster for Memcached stores compared to C7i and C7i-flex instances.

**C8i-flex instances are a lower-cost variant of the C8i instances, with 5 percent better price performance at 5 percent lower prices. These instances are designed for workloads that benefit from the latest generation performance but don’t fully utilize all compute resources. These instances can reach up to the full CPU performance 95 percent of the time.

https://aws.amazon.com/blogs/aws/introducing-new-compute-optimized-amazon-ec2-c8i-and-c8i-flex-instances/

17 of 27

Delivering through �our strong partnership

17

18 of 27

Intel and AWS Partnership Dates Back To the First EC2 Instance with Intel® Xeon® Processor

“At AWS, we’re committed to delivering the most powerful and innovative cloud infrastructure to our customers. By co-developing next-generation AI fabric chips on Intel 18A, we continue our long-standing collaboration, dating back to 2006 when we launched the first Amazon EC2 instance featuring their chips. Our continued collaboration allows us to empower our joint customers with the ability to run any workload and unlock new AI capabilities.”� –Matt Garman, CEO at AWS

https://newsroom.intel.com/corporate/intel-strategic-collaboration

18

19 of 27

Hardware optimization

AWS instances:

M6i, C6i, R6i…

AWS instances:

M5, C5, R5…

AWS instances:

M1

Intel Xeon

(2006)

Intel Xeon Scalable 4^th Gen

(2021)

Intel Xeon Scalable 2^nd Gen

(2019)

Intel Xeon Scalable 3^rd Gen

(2021)

AWS instances:

M7i, C7i, R7i…

Intel Xeon Scalable 5^th Gen

(2024)

Intel Xeon Scalable 6^th Gen

(2025)

AWS instances:

I7ie, G7, P6

AWS instances:

M8i,M8id,…

Price:

M1.2xlarge:

$255

Price:

M5.2xlarge

$280

Price:

M6i.2xlarge

$280

Price:

M7i.2xlarge

$294

Flex: $279

Price:

N/A

Price:

M8i.2xlarge

$309

Flex: $293

*Prices based on AWS public calculator.

19

20 of 27

Intel® Architecture Instance Types on AWS

General purpose instances provide a balance of compute, memory and networking resources, and can be used for a variety of diverse workloads.

Habana Gaudi

1^st and 2^nd Gen Intel® Xeon® �Scalable processors

General

Purpose

Compute-�Optimized

Memory

Optimized

Accelerated�Compute

Compute Optimized instances are ideal for compute bound applications that benefit from high performance processors.

Memory optimized instances are designed to deliver fast performance for workloads that process large data sets in memory.

Accelerated computing instances use hardware accelerators, or co-processors, to perform functions more efficiently.

2^nd Gen Intel® Xeon® �Scalable processors

Intel® Xeon® Scalable

processors

Intel® Xeon® v4

Processors

Storage optimized instances are designed for workloads that require high, sequential read and write access to very large data sets on local storage.

Storage�Optimized

Intel® Xeon® v3

Processors

DL1

M5(d)n

M5zn

R5(d)n

R5b

T2

C5(d)

R5(d)

HMI

T3

M5(d)

C5n

z1d

P3dn

I3en

M4

X1

R4

P2

P3

G3

F1

H1

I3

C4

X1e

D2

X2idn

X2iezn

I4i

HPC6id

3^rd Gen Intel® Xeon® �Scalable processors

HPC

Optimized

Ideal for applications that benefit from high-performance processors including large, complex simulations and deep learning workloads.

4^th Gen Intel® Xeon® �Scalable processors

R7iz

M7i

M6i(d)

M6i(d)n

C6i(d)

C6in

R6i(d)

R6i(d)n

X2iedn

D3

D3en

G4dn

P4d

C7i

R7i

Features Intel® AMX

U7i

C7i-Flex

M7i-Flex

^5th Gen Intel® Xeon® �Scalable processors

^6th Gen Intel® Xeon® �Scalable processors

M8i

M8i-Flex

R8i

R8i-Flex

I7i

I7ie

See https://aws.amazon.com/ec2/instance-types/ and speaker notes for details.

C8i

C8i-Flex

20

21 of 27

Instance Analysis

Current	$	6i	$/ Perf *	7i	$/ Perf *
c5.xlarge	0.2620	c6i.xlarge	20%	c7i.xlarge	33%
c6a.2xlarge	0.4716	c6i.2xlarge	3%	c7i.2xlarge	29%
m5.xlarge	0.3060	m6i.xlarge	20%	m7i.xlarge	33%
m5.2xlarge	0.6120	m6i.2xlarge	20%	m7i.2xlarge	33%
m6a.2xlarge	0.5508	m6i.2xlarge	3%	m7i.2xlarge	29%

Public prices, São Paulo Region

* Estimated

21

22 of 27

Resource optimization

22

$ 0.9157

4th Generation Intel Xeon Scalable

5:3�Instance Consolidation

2nd Generation Intel Xeon Scalable

C5.xlarge

5 instances

3 instances

m7i-flex.xlarge

$ 0.2620

*Hourly price per instance in São Paulo

*Pay as you go

*Linux

Total for 5 instances

$ 1.31

$ 0.3052

Total for 3 instances

30% potential savings

48 GB RAM

40 GB RAM

22

23 of 27

ITAU Unibanco – Caso de Estudio

El banco más grande de Brasil y América Latina, con operaciones en todo el mundo que atiende a unos 55 millones de clientes.

23

24 of 27

Reto: Transformar y modernizar sus aplicaciones, reduciendo costos, aumentando las ganancias y mejorando la escalabilidad.

ITAU Unibanco – Caso de Estudio

24

25 of 27

Solución: Migró el 99% de su nube privada y el 20% de su plataforma distribuida a AWS. +19.000 servidores

ITAU Unibanco – Caso de Estudio

25

26 of 27

Resultados: Reducción del 99% en el tiempo de entrega de la plataforma. Satisfacción del cliente

De

3.5x – 6.4 veces

Rendimiento de la infraestructura mediante instancias de AWS basadas en

Intel Xeon de ^4.ª generación.

ITAU Unibanco – Caso de Estudio

26

27 of 27

27