Readying Enterprise Networks for Artificial Intelligence

Based on a survey of 269 North American IT professionals, EMA’s report, "Readying Enterprise Networks for Artificial Intelligence," examines how early AI adopters are evolving their networks to support the demanding requirements of AI training and inference.

The research provides actionable guidance for IT leaders on how to rethink, re-architect, and upgrade data center and WAN infrastructure to enable AI-driven operations.

Check out these research slides from Shamus McGillicuddy, EMA Vice President of Research, to get key findings and insights to help your organization prepare its network for the future of AI.

1. @ema _resea rch
| @ema_research
Sponsored by:
Readying Enterprise Networks
for Artificial Intelligence
Network Infrastructure and Operations
Shamus McGillicuddy
Vice President of Research
Enterprise Management Associates (EMA)

2. @ema _resea rch
| @ema_research 2
Watch the On-Demand Webinar
Readying Enterprise Networks for Artificial Intelligence On-Demand Webinar
https://info.enterprisemanagement.com/networks-for-ai-webinar-ss
© 2025 Enterprise Management Associates, Inc.

3. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Featured Speaker
Shamus is a Vice President of Research at EMA,
where he leads the network infrastructure and
operations practice. He has nearly two decades of
experience in the IT industry. His research focuses on
all aspects of managing enterprise networks,
including network automation, network observability,
multi-cloud networking, and WAN transformation.
© 2025 Enterprise Management Associates, Inc. 3
Shamus McGillicuddy
Vice President of Research
EMA

4. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Sponsors
© 2025 Enterprise Management Associates, Inc. 4

5. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Research Methodology
© 2025 Enterprise Management Associates, Inc. 5
Survey of 269 IT professionals involved in preparing
their networks for AI applications and traffic
Corporate AI strategy
was already in progress
74% had at least some AI
applications in production
Subject matter
experts
Middle
management
IT executives

6. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Job titles
47% CIO/CTO
33% IT/related director
8% IT-related manager/supervisor
5% IT-related president
5% IT or network architect
3% IT or network engineer
1% IT/network infrastructure analyst
IT groups
65% IT executive suite
13% IT security/cybersecurity
9% IT architecture
7% IT/Network operations
5% Cloud operations/engineering
1% Network engineering
Top industries
24% Banking/Finance/Insurance
19% Manufacturing – Heavy/Industrial
13% Retail/Wholesale/Distribution
10% Manufacturing – Consumer goods
9% Health care/Pharmaceutical/Biotech
6% Professional services not related to IT
Company size (employees)
48% Midsized enterprise – 1,000 to 4,999
29% Enterprise – 5,000 to 9,999
23% Large enterprise – 10,000 or more
Demographics
© 2025 Enterprise Management Associates, Inc. 6

7. @ema _resea rch
| @ema_research
AI Strategy and the
Network

8. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
AI Adoption Activity
© 2025 Enterprise Management Associates, Inc. 8
42% of AI enterprises have an
AI Center of Excellence to
lead strategy
AI technologies in production by end of 2025:
58% Proprietary LLMs
51% Machine learning
34% Open source LLMs
32% Agentic AI
18% Retrieval-augmented generation

9. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
AI Workload Distribution: Enterprises Must Update Data Center and Wide-Area
Networks
© 2025 Enterprise Management Associates, Inc. 9
Where will your training workloads reside by 2028? Where will your inference workloads reside by 2028?

10. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Complexity of workload distribution across sites
Latency between workloads and data at WAN edge
Complexity of traffic prioritization
Network congestion
Security risk
Cost/Budget
Rapid technology evolution
Networking team skills gaps
39%
34%
33%
29%
Integration between AI networks and legacy networks
Bandwidth demand
Coordinating traffic flows of synchronized AI workloads
Latency
43%
41%
38%
36%
AI Networking Challenges
© 2025 Enterprise Management Associates, Inc. 10
Data center
networking
issues
42%
39%
36%
33%
WAN issues
Business Concerns

11. @ema _resea rch
| @ema_research
Network
Infrastructure
Preparation

12. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Network protocol adoption plans
67% 64% 33%
Only 49% Say Their Data Center Networks are Ready for AI Traffic
© 2025 Enterprise Management Associates, Inc. 12
Planned infrastructure investments
High-speed Ethernet
(800 GbE)
Hyperconverged
infrastructure
SmartNICs/DPUs
75% 56% 45%
Ethernet RoCE NVMe over Fabrics InfiniBand
42%

13. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Only 48% Say Their WANs are Ready for AI Traffic
© 2025 Enterprise Management Associates, Inc. 13
Essential solutions
73% High-performance cloud interconnects
56% Private, dedicated AI backbone networks
53% WAN overlay solutions (SD-WAN/SASE)
WAN optimization/acceleration requirements
61% AI-aware bandwidth efficiency techniques (deduplication, compression)
59% AI-aware data transmission acceleration (AI protocol optimization)
49% AI-aware WAN remediation
49% AI-aware traffic prioritization/shaping
64% are reducing WAN latency by deploying AI workloads in edge computing footprints

14. @ema _resea rch
| @ema_research
Securing AI
Investments

15. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Top Security Concerns
© 2025 Enterprise Management Associates, Inc. 15
Data privacy and compliance risk
Vulnerabilities in AI APIs/third-party integrations
Targeted attacks against AI models
Data leakage
Enterprises that
expect less
AI networking success
perceive more
risk from targeted
attacks
60%
48%
47%
43%

16. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Top Protective Measures
© 2025 Enterprise Management Associates, Inc. 16
Which steps have you taken or plan to take to secure your AI applications and data?

17. @ema _resea rch
| @ema_research
Observability of
AI Networks
17

18. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Only 47% Believe Their Network Observability Tools are
Ready to Manage AI Networks
© 2025 Enterprise Management Associates, Inc. 18
Data collection adjustments:
67% Real-time network metric monitoring
51% Real-time flow monitoring
40% Broader coverage
33% Increased packet capture scalability
Reporting and analysis adjustments:
59% AI application recognition/intelligence
46% Predictive congestion analysis
42% Anomaly detection for AI-related traffic patterns
34% GPU-cluster traffic pattern analysis
68% strongly believe their tools should leverage AI to manage AI

19. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Final Thoughts
© 2025 Enterprise Management Associates, Inc. 19
67% believe they will be completely successful with prepping networks for AI
Optimism came from IT execs
Technical personnel and middle management were pessimistic
Potential best practices:
Hire AI experts now
Focus on connectivity to third-party networks (LLM providers, etc.)
Automate AI traffic prioritization across networks
WAN acceleration: look for AI-aware data transmission acceleration and WAN
remediation techniques
Leverage edge computing to address latency

20. @ema _resea rch
| @ema_research
@ema _resea rch
| @ema_research
Learn more about this new report:
https://bit.ly/ema-networking-for-ai
© 2025 Enterprise Management Associates, Inc. 20
In-Depth Insights in the Full Report, Available from Sponsors